Widowmaker Amp Part III: Resolution

The widowmaker amp: no transformer, no fuse, AC neutral chassis grounded, and a two pronged un-polarized plug. A technician killer.

This will be a brief post. My others on the subject of this amp were not. I, as I usually do, have gone ahead in my ingorance instead of stopping to do a little research on who has conquered this territory before my. And so I have wasted much time again, as I so often do. I’m not stupid, just ignorant and too enthusiastic.

Here are the schematics I pulled from the net. Thanks to the author! The 170k resistor – .047µF cap network must be removed but only after carefully following Rob’s clear instructions on his site.

Rob Robinette, of robrobinette.com, explains it all beautifully. And with graphs and stuff. I’ll just take one of his images and give a brief explanation. If this is interesting, please visit robrobinette.com/widowmakers.htm to read more from the man himself about the widowmaker amp. Below is a schematic he uses to explain how the grounding system worked in these amps.

What tripped me up was the RFI AC Ground Capacitor, and “return resistor.” In a previous post, I actually considered omitting the “R-C network in series with the ground.” But I left it. Big oops. That R-C network is there so that if you plug in the two prong AC cable backward, the shock you receive is limited to roughly 3ma at 120v AC. A good shock but not enough to kill you. That capacitor was called the death cap by technicians because a short across it would give the user the fill 120v at 15A from the fuse or breaker, making their guitar potentially deadly. Removing this R-C net makes the amp behave normally, once you’ve added isolation, and grounded the chassis to PE ground.

Big shout out to Rob Robinette, and not just cuz I share his first name. He’s a tube educator who deserves much praise and respect. I’ve learned lots about making hifi amps by reading about what guitar amps do from him. And lots about tube math too! Thanks, Rob! The noise in the amp is still a little present, but the whole amp functions how a modern amp should. No shocking the user, and no “return” R-C network inside to add noise to the ground rail. It’s not perfect, but it’s a damn sight better than it was. Thanks again, Rob!

I need to slow down and do research to get to where I’m trying to go, so I can start where others have finished and use my time the most wisely. It’s hard, man: I just wanna be great at something now. Cheers. Hope this was interesting again.

Widowaker amp part III: 60Hz Hum Issue

I have been working on this lil amp over the last week. The owner was kind enough to pay me yesterday, which was awesome, but he also asked “what’s with that hum?” When he turned it on. So I drove home, thinking about how to eliminate this 60Hz hum. D’oh!

By now I should know: if I notice it, it’s going to be a problem for the owner. Oops.

To go through the list of modifications I made in the previous posts:

-New Metal Film Capacitors in place of old paper-in-oil ones.

-Chassis is Protective Earth grounded and no longer grounded to neutral mains

-Transformer added: Triad Magnetics toroidal isolation transformer

-New AC power cable added, and mains fuse

-Cleaned input jack and it’s short to ground switch.

Out of the list above, there a few distinct possibilities that could be adding 60Hz hum to the output. One is that the toroid’s location, which is near the input tube, is not ideal and the toroid actually should be placed outside the chassis or near the rectifier instead of near the input tube. The toroid is so close to the ground of the input tube that it could be coupling noise in via proximity to it’s magnetic field. Hopefully just by re-routing the wires around the input tube I can fix the problem, but it might require transformer relocation in the chassis to fix. Ugh.

It could also be the fuse holder. I know sometimes a loose or crappy quality fuse holder can cause the 60Hz noise from the mains to vibrate into the PSU, causing a 60Hz hum to be audible on all power rails.

Another possibility is that the ground rails in the circuit need to be lifted from the chassis. In the original design neutral was the ground, and it was connected to the chassis. It’s possible that having the neutral AC mains touching the PE grounded chassis could be causing some kind of ground loop.

Another thing I’d not considered was that maybe by using metal film capacitors they could actually be picking up the magnetic field from the transformer, and coupling noise into the circuit that way. Maybe paper-in-oil caps were used because paper and oil don’t interact with magnetic fields? Also I should measure wether I have a 130kΩ or a 170kΩ resistor in parallel with the .047uF, and maybe try to swap it out for the other value. I’ll also change the value of the bypass cap for the 50C5, because I put a 47uF in instead of a 33uF which is closer to the original value, and will lower the gain a lil bit in the last stage.

The last thing I’m going to try is to add a little capacitance to the power supply. Why? In a previous post I criticized the previous service done to this amp because there was a 330uF 200V cap in the power supply circuit. This cap’s capacitance was too high, and it might have caused problems in the circuit conducting current. What I missed is that the stock schematics show a 100uF cap followed by a 47uF cap, and not two 47uF capacitors which is what I put in. I’ll try adding that 100uF cap, and see if it helps. It’ll likely help the overall noise level, but not the hum. I think the hum is caused by the toroid, AC mains routing, or maybe the fuse.

In reality, I may have to set this whole chassis up on my bench, open side up, and do some measurements and testing with it belly-up. Normally, with Fender and similar style amps, the chassis drops out of the top of the amp, and shows it’s guts to the technician pretty easily while remaining plugged in to the speaker for load and power for testing. That makes servicing top-cabinet chassis designs easy on the bench. Making their chassis fit the bottom of the amp makes the Harmony amps like this one much trickier to do benchtop testing and measurement with

First thing’s first: move the transformer to the other side of the chassis and see if the hum goes away. Since there was no room inside to do it, I chose to try it on the outside, with the wires going round the back of the chassis. When I flipped the switch, I was displeased to hear the same damn humming that I’d heard before. No audible difference. Shit. This shoots some holes in my solution. I thought that perhaps I could move the transformer, and that the field it produces wouldn’t be coupling noise into the preamp tube anymore. Clearly, nothing about that assumption is true. Ok.

Next I’m going to add a 100uF cap in place of the 47uF I put into the circuit, and see if that reduces the noise. After that, I’ll probably do the fuse holder to make sure it’s not vibrating. It may come down to using an oscilloscope to trace the noise, but I’m hoping I don’t have to do that. It’s also possible that the rectifier or one of the other tubes could have a bad insulator between it’s Cathode and it’s filaments, which could push 60Hz noise though the audio path. I suppose if it continues to be a problem, I could swap tubes. That just means I have to buy more (ugh.)

Maybe it was the capacitors? Metal film interacts with magnetic fields. But that doesn’t explain why the hum doesn’t change when the transformer is moved to the other side of the chassis, and from bottom to top side. If the caps were coupling noise from the field, moving the toroid would have made a difference. Unless the coupled noise is coming from the magnetic fields of the filament transformer, which is located under the chassis… Hmmmm…

After the 100uF cap was added we can hear a reduction of the hum, but not by enough. This points me to the rectifier again, or at least the power supply. It’s only a half wave rectified design. That means the noise level is likely to be higher than it would be with a full wave diode bridge to rectify the voltage instead of a single tube, which is noisy. There could be a noisy 35W4 rectifier tube, too, which is possible. The getter inside looks a little spent. I’m going to try to bypass the rectifier’s anode-cathode path with a solid state diode to see if it helps the overall noise level. I don’t think the increase in voltage from the SS diode should matter. I have to leave the rectifier’s filaments connected though because they make the math for the filaments add up.

Bypassing the tube with a solid state diode makes no difference either, noise wise. It does, however, increase the power flowing through the circuit by about 20v DC, taking the B+ on the plate for the 50C5 tetrode tube from about 120 or so up to about 142 or so. That satisfies the owner’s initial request to “do something to get more power out of it.” But for the moment, I am disappointed, because I thought it would quiet the noise, and again, no luck. I am missing something here… I see that much…

Is the fuse vibrating, and introducing noise into the circuit? Answer: no. I replaced the fuse holder I initially used with a new one and no difference was made. Ok. Again, I am clear that I’m missing something here that seems critical. I’ve made all kinds of changes to the PSU, and to the orientation of the transformer and it’s wires. Is it possible that I’m barking up the wrong tree, and that ground is actually what I should be looking at?

A badly implemented ground rail can cause noise even when the rest of the circuit is well thought out and well assembled. So what do we know about ground in this amp? First thing- it’s a half wave power supply, so the DC ground is actually the neutral AC line. The second thing we know about this amp is that it’s using the chassis as a return path between the output speaker and the input tube. The input neutral is only “grounded” through the 130k resistor and the .047/600v capacitor. Why does that matter? Because we connected the Protective Earth wire to the chassis too, to make it safe. I think the designers of this amp intended it to have a floating chassis ground, otherwise this whole mess of a circuit couldn’t work. OK, if so, then WTF is there an R-C network connected between mains neutral & chassis ground for? Dammit. This amp looks badly designed to me.

What happens when we remove the PE ground lug from the chassis? It makes the chassis a floating “ground” that returns current to the input section of the amp from the output section. That means that we can remove the ground lug from the chassis. Before doing so, I come to a conclusion: the chassis should be PE grounded, and the chassis ground path should be replaced with a wire. Ugh. I actually don’t want to do this, in case I have to “undo” my mod, which could be messy. Lifting the connections from the chassis like this is something I’ve never done in a guitar amp. I could be making more work for myself. This amp is designed strangely.

Unfortunately, changing the chassis grounding has caused the circuit to be unhappy again. Lots of noise and popping. I’m going to go back to how I had it wired initially- as the schematic says, but with the chassis PE grounded. Sorry, Mark! I think this amp is just noisy because of the way it was designed. I’ll tell him to try rolling the tubes out and see if it helps. Sometimes the insulator layer in high filament voltage tubes can be noisy and transfer hum. Super disappointing results for this experiment. I wanted the hum 100% gone, not 60% reduced. The amp is now safe to use, but Mark will have to learn to live with a little hum. The hum is much quieter than the first time he heard it so hopefully he approves.

The only two things I didn’t try were to roll the tubes out to see if that made a difference, and to change the metal film caps out for paper-in-oil ones. The stock caps were PIO, and that may be why they were chosen, beyond the fact that they used to be much more common. Maybe PIO caps don’t pick up magnetic hum. I’ll have to do more research on that front.

The conclusion I come to is this: you can make a badly designed amplifier work safely, but it’s far more difficult and time consuming to make a badly designed amplifier into a well designed one. I’m also acutely aware that I’m deeply ignorant about guitar amplifier architecture and operation concepts. Time to school myself I suppose.

The radioactive green space lizard on the Lucky Charms box was right when he told me that failure tastes great. It ain’t so bad. Seriously, though, I like to publish some failure with my successes I write about here. The reason is that I frequently come up against the limits of my understanding and its good to have a record of things I don’t understand for later reference and thought. It also shows my readers that being bad at it is no reason to stop doing something. Being bad at stuff just means it’s time to do that stuff over & over until you get good at it.

Cheers. I’m going to bong my face off after experiencing the frustration of this hum issue. I kinda have two modes: discouraged and questioning everything I do, or kicking ass and refusing criticism, varmint! I should probably just get a grip. I have to remind myself I did a great job with the modification to make the amp safe, and that’s what matters. The slight hum SUCKS but it’s better than the amp being useless because it was dangerous. At least you can’t hear the humming when it’s played.

At some point, I’ll look back at this and understand it.

Harmony Amp Service Part II: The “Widowmaker” Made Safe

Hi Folks! I am very excited to show off my restoration of the amplifier. In part I, I went through all of the considerations and most of the math necessary to find a transformer for this bad boy, and I mentioned in passing a few things I’d be doing with it when the parts came in. The parts are here! So it’s time for us to install an isolation toroid from Triad Magnetics which cost about 30$ including shipping.

In order to make the space we need to install the toroidal transformer, we need to do some house cleaning inside the chassis. First thing is first: remove what was added by other technicians that doesn’t belong. What, specifically in this case am I removing? Some misguided fool of a technician put a 330uF/250v capacitor on the PSU circuit. Why is this worthy of removal? Isn’t more capacitance always good for power supplies? The answer is mostly extra decoupling capacitance is always good for PSUs, except for when the PSU uses a tube, like the 35W4 used in this circuit as a rectifier (diode.) The tube’s job is to convert AC to DC by only conducting during the appropriate part of the AC cycle.

The tube needs to be “loaded” by another component in order to rectify AC to DC correctly. In HiFi design, there is really only one option (for those of us with the $$$ and know-how) for a tube power supply to be loaded: with a big, expensive, heavy, high voltage low frequency inductor capable of passing all of the current the PSU produces while powered. This is a non-starter for guitar amps in most cases, which are made cheaply to keep them afforable to the customer. There are a few like Fender, Vox and Mesa, where they’ve done designs that use chokes for “clean-tone” amps, but they’re usually not the kind a rookie can purchase to practice with.

While omitting a choke from the PSU doesn’t actually make the quality worse, per se, it does mean that you can’t pass as much current through the tube as you could with an inductive load. Without a choke, most tube rectifiers are loaded with a capacitor instead. This part is mathy: how much current will you draw at idle, and then at peak? Once you know that, you can look at the datasheet for the tube and it will tell the max values for capacitive loading the tube under the conditions where you are drawing X amount of current. You have to choose a capacitor that will discharge somewhat between each AC cycle, thus allowing it to conduct current. So small capacitive loading values are common with tubes (some as low as 4uF!) Too large a capacitance value can actually hamstring a PSU with a tube in it, and prevent it from conducting current properly. Too big a capacitor can also kill your tube.

In this case, 330uF of capacitance is more in that one cap than the whole stock circuit had when it was built! That means someone wanted to make it have a “stiffer” PSU, less likely to sag under load (like a big chord,) and probably thought, “well, it works in solid state amps!” And went ahead and put the cap in. Big mistake- the tube rectifier can’t work properly. The guy working on it was likely trying to compensate for the leaky caps left and right, which adding capacitance to the PSU will not effect. The reason the last owner experienced a “loss of power” playing this lil amp is simply because the old paper-in-oil caps were trashed and leaking. He also told me that the headroom and breakup for the distortion were all F’d up too, which of course, your coupling caps are totally critical to. Let’s get to cuttin’ shall we?

Above you can see the space left by the old multicore capacitor, which I unceremoniously cut out with diagonal cutters. The old capacitor was totally trashed, and one of the three chambers hadn’t even been used by the manufacturer. Pretty funny, because they were paying for the whole thing, they might as well have added it to the PSU. Oh well. The “firecracker” caps, the kind of multicore caps with colored wires instead of solder terminals, are all long ago garbage at this point in the year 2020. Most were rated for “one year in circuit” which I find amusing. So the old one got cut out, and I cut out it’s constituent wires too. I kept the cheapo-ass looking pop rivet in the chassis, as it is better than a hole, but I HATE pop rivets. They look so fucking cheap and crappy. I digress. This space we’ve just made will accomodate our new transformer quite nicely, I believe. The new toroid has rubber pads to prevent it from rubbing against the rivet while the amp is dragged around, after I bolt it in.

Above you can see the Standard brand capacitors which are leaking PCB oil (God I hope it’s mineral oil, but probably not.) I have heard people call that substance “wax.” Wax is a misnomer, though, as it implies that the material seen dripping from these caps is meant to seal the electrolyte inside. In my experience, that “wax” is usually the electrolyte itself, cooled to a cool enough temperature to begin to solidify. Most old capacitors used nasty ass shit for electrolyte, things like PCB oils that cause cancer. In the photo above, the only cap saved from the restoration will be the blue one stretched between the sockets; all the others in this circuit need replacement. So bing, bang, boom. We cut em out, and solder some new ones in.

There we go. The .01uF 400v caps were replaced with .022uF/630v metal film capacitors usually used in powered pickup amplifier circuits for electric guitars. The old Good-All cap was replaced with a new one, also metal film. The electrolytic caps are a burly 400v rating although this circuit likely never sees over 200v. It should make the PSU happier, too, now that it only has 96uF of capacitance in the whole power circuit. The old firecracker style capacitor had a green wire, which was a 30uF cap at 25v. I replaced it with a 47uF/35v cap. The last things to do, after these steps, will be to add the transformer, and then to wire up a new 3-pronged plug that will have the chassis grounded to Protective Earth, and wire the power cable to the transformer so that the on-board on off switch still works properly.

To add the toroid, I simply drilled a new hole in the chassis, making sure not damage the nice blue paint. I measured it’s distance from the edge of the chassis to make sure it fits, and it does. Yay! Toroids are totally fucking awesome compared to EI core transformers: only one hole to mount them, less stray flux bands in the electrical field, smaller, quieter, and far more resilient to over-voltage and over-current circumstances than their older E+I core counterparts. They’re usually expensive, but Triad makes good quality and cheap priced ones for isolation. I chose the model that can supply 440ma.

This amp design is weird: since it was designed to run off of mains with no isolation, the designers had to run the AC through a bunch of the circuit that normally doesn’t see AC in an audio circuit- at least no hifi circuit I’ve ever seen looks like this. One phase of the AC is run through the filaments, cathodes, the screen of V1, and the suppressors of the tubes. Holy Cow. Personally, this amp’s schematic looks like a dog’s breakfast of a mess to me. You can see it below. The isolation transformer I installed was the same as in the drawing. I even forgot to add a fuse, just like the drawing. The 120v mains fuse belongs on one of the two lines going to the transformer from the wall.

Why would the designers take the AC mains neutral, and introduce it to the cathodes and suppressors of the circuits? It acts as a “ground,” although it oscillates with the mains. Ok. That’s pretty straight forward. We only have a single rectified DC phase bouncing up and down to run the DC supply on to begin with, so having this AC neutral ground should be sound OK. They “ground” all of the components in the output stage to neutral. This is ok for the output stage, as it is more immune from transients than the input stage. The input stage “ground” is again an AC secondary, this time from the filament transformer secondary. The two separated ground rails are divided by an RC network comprised of a .047µF/600v cap and a 130k-170kΩ resistor. I assume it’s to attempt to isolate the ground for the input stage, which will be more susceptible to noise from the ground rail.

V1 and V2 also have their screens connected through a 2.2MΩ. The Anode of V1 also feeds the screen of V2 through a resistor, in addition to feeding signal to the grid of V2. Like I said before, it looks like a dog’s breakfast. Lots of noise is possible in this circuit. But hey, I don’t do guitar amps. This may be a perfectly effective solution for signal noise in such a simple circuit. I don’t know because I’ve never had any experience with such a cheap design. I can’t wait to see how it performs.

One end of the isolation transformer’s parallel cores is the red and yellow wire, shown above going to the on off switch on the front panel.

Above you can see the brown wire, which connects to the other end of the isolation transformer’s parallel cores. This way, no AC ever touches the circuit without going through the transformer first.

Above is a photo of the strain relief for the power cable, which is a plastic cable tie and a zip tie working together to prevent catastrophe if someone kicks the power cable out of the wall, as happens at gigs and in studios from time to time.

You can see I opted for the “fool’s strain relief” inside the chassis, by tying a knot in the power cable. The secondary plastic cable tie is a must in this case, to prevent a real problem in case the knot isn’t good enough. Anyhow, all is well with this little 3W amp. We’re ready to put it all back together and see how the owner likes her sound.

I’d like to take this chance to point out something: There are only 7-pin tubes, or septal based tubes, in this amplifier. That is pretty cool. Why? Because I always wanted to know what an amp made with only septal based tubes was capable of, and now I know, at least approximately: 3W for a guitar. If I made a hifi based septal system, I could use solid state rectifiers and get some more efficient power delivery out of the amp, but less wattage due to how Class A amps tend to be much lower power than their guitar amp counterparts. It cuts down on distortion. I could probably do a 1 watt per channel stereo push pull hifi amp with septal tubes and it and still have it rock pretty good. I should design that…

I got distracted and forgot to add the fuse to the unit. Oops. After drilling another hole and securing the fuse holder with a 4-40 screw and nut, we should be good to go. You can see above it’s lookin good.

OK. Now we have the approximate circuit from the schematics above, except we are using a 10Ω resistor where the “7.5Ω fuse resistor” should be, and a real fuse inline with the mains input to the isolation toroid. I’m not sure what the F*&% people were thinking, making a device that uses no isolation and a resistor as a fuse. A resistor is NOT an adequate fuse by any means. A real mains fuse of 125v at 0.5A should do, and although slow-blow fuse would be fine (tube amps are pretty robust to a short circuit condition) I’ll start with 0.5A fast-blow fuse to make sure the amp can handle it before I button it all up again. A 1A fast blow fuse might be necessary to accommodate any current inrush that might occur after the on switch is thrown.

Why might there now be an inrush of current in the circuit power supply if we have removed the large sized capacitor that was swamping the tube rectified PSU down? Shouldn’t less capacitance result in less inrush current? Yes is the answer. But we’re forgetting to account for a critical design aspect of a toroid power supply: the initial magnetizing current needed to magnetize the core. With a toroid, there is more magnetizing current needed to magnetize the core than would be required to magnetize an E+I core transformer. This can cause a sensitive fuse to blow prematurely from the current needed to magnetize the transformer initially. Another alternative is a slow blow .5A, which should be able to withstand the momentary transient current spike of the core magnetizing.

The last thing I would want done to this amp, if it were mine, would be to add an indicator light to show the unit is powered on. I like those. Indicator lights make me warm and fuzzy inside because I know something is turned on and working. An amp without an indicator, too, is likely to end up being left on and forgotten about in the mad studio rush of recording and multitracking. Fortunately, though, tubes like not to be switched on and off, so leaving tube amps on over night (or occasionally for days or weeks,) usually doesn’t “harm” the circuitry at all. It just puts hours on your tube life clock, which is of course somewhat undesirable. You want your tubes to last. But leaving this lil guy on for weeks, even, I can’t see anything inside the chassis failing- the tubes are of questionable lifespan, but they’re dirt ass cheap to replace. I’ve made it all pretty robust, so I’m not worried. Just more things to consider.

Speaking of considerations, I missed the rear wood piece that seals the little cabinet together when I placed my toroid. Fuck. So after I realized the chassis was not going to fit back in the cabinet, I considered cutting a little wood out to fit it. Then I came to my senses, and realized I could move the toroid around inside a little bit, cut out the old ceramic input capacitor and replace it with a polypropylene one soldered in a slightly different place, and toque the transformer down in it’s new location making the cabinet and chassis fit again… Just barely. Phew! So relieved. Now even the input cap is a nice boutique amp cap, so I guess that’s cool, though probably not necessary.

I saw these amps going for about 200$ or so online in a few places. An amp like this, with the “widowmaker” architecture taken out and replaced with a properly isolated power supply, should actually be worth the pain and expense. If we take into account the parts used, it costs roughly 2 hours or so of wiring, stripping, drilling, and assembly in addition to the approximately 50$ for the components used. So if I were to find this amp used for 50-150$, the whole job would add up to about 200$ plus a couple hours. Super cheap for an authentic point-to-point wired amplifier full of glass tubes with a Jensen custom speaker. My conclusion: well worth it. Triad makes other isolation transformers, too, so one has some flexibility with power transformer choice when doing similar “widowmaker” amps. Hope this post was interesting. I really hope the owner likes the new amp, because electrically at least, it is far better designed now than it was when I got it. Cheers. Go play some music, even if you suck at it.

Amp Service: “The Widowmaker,” a Harmony Brand H303B 3W Practice Amp (Part I)

I’ve loved amplifiers ever since I saw my dad’s Marantz growing up. The glow, the metal, the dials, the knobs and meters. It was all so interesting! I also fell in love with guitar amps after I started hangin out with a metal band after I dropped outta college to get high. Good ass times. I highly recommend shirking all responsibility in life for a period of time. You learn a lot. Anyhoo, as I was never much of a musician, I made myself useful to musicians by servicing their amps and other gear, which got me tons of free tickets to shows. I serviced DJ QBert’s Akai Drum machines and I got a ticket to the sold out reunion show of the Invisibl Skratch Piklz in 2018, for example, which was totally fucking phenomenal.

A friend of my dad’s, this guy Mark, is a funny neighbor of my parents. I was at the superbowl party and we were talking and he asked me if I could fix a “widowmaker” amp. I was intrigued. I know he’s a guitarist, but I did not know the WidowMaker brand. He explained it’s not an amp brand, it’s what guitarists call “this kind of amp.” Again, I was intrigued. What could he be talking about?

It’s an amplifier design like many from the early days of guitar amps. Single ended, tube rectified DC Power supply, capacitor coupled, and a total of only 3 tubes. Power transformer? Nada. Choke loaded tube rectifier stage? Nope. No iron cores to be found in any event, until we get to the output transformer. A totally insane design by todays standards. No earth ground in the input wiring either. Apparently these amps were made by the ton for a cheap alternative to Fenders and more expensive brands, and the term Widowmaker is super common among guitarists. I’m surprised I never heard it before.

This Harmony branded H303B model amplifier, which seems to be a 3 watt amp, is called a “widowmaker” amp because there is no galvanic isolation (power transformer) to isolate the player of the amp from the mains in the wall. A transformer isolates the two sections of AC voltage from a direct electrical connection to one another, making it possible for the user of the amp not to shock his band mates among other things. It’s a much better design to use transformers. So why didn’t every amp do it? Cost. They cost and weigh a lot compared to omitting them. Also machining: for more transformers, you need more mounting holes in the chassis. Looking at the photo above, you can see the distinct lack of a power transformer in the chassis- usually they’re a substantial part of the weight of the amp, too, making these lil guys very light weight.

How to fix an amp with no power transformer? Give it one. We need a transformer that does a simple 120v output, with a shield winding (if we can find one) to ground to with the Protective Earth ground from the wall. The new Power Transformer can be from any brand, and must give us a full 120v at roughly 400ma to be able to fully supply the amp without the transformer getting too hot. How do we know that? The silkscreen on the chassis tells us the amp consumes “110-120v AC 0.3A \ Input Power 45 Watts.” From that we can do some simple Power math: 120v X 0.3A= 36 Watts. Hmmmm. Ok. 120v X 0.4A = 48 watts. Now we’re seeing a Wattage close to the 45 quoted on the Silkscreen. So let’s get us a power transformer that is almost certainly overkill: a 400ma isolation transformer.

The Triad Magnetics brand VPT230-220 is a good bet for a nice little isolation transformer. It can do a 115v AC Input to a 115v AC output at 440mA nominal. The transformer is also a toroidal design, making it less likely to pick up or give off hum when compared to traditional EI core transformers. And best of all it’s only 80mm wide or so, so it should fit into this little amp cabinet. Worst case scenario, I mount it on the wood frame and put heat shrinked quick connect terminals on the wiring leading into the chassis for easy disassembly. This transformer choice may however require some NTC resistors in series with the mains connection the the transformer to assure that if, say 125v AC were applied to the isolation transformer, it wouldn’t overload the primary’s voltage ratings as the current inrush limiting resistors would drop the voltage on the mains side slightly. I could also just put a 10Ω resistor in series with the transformer primaries, too, I suppose.

I’d really like to get an R-Core isolation transformer for this amp, but sadly, this project does not really warrant an R-Core. The Chinese ones are unavailable through J&K Audio Design because of the Corona Virus outbreak, and this 3W amp doesn’t warrant an exquisite very pricey R-Core from Japan. Plus the R-Cores from China on eBay are all essentially worthless for tubes, or isolation- unusual voltages and currents available on most of them. A Triad toroid for 30$ should be great for this project. The main PSU Capacitor, seen above, will be replaced by modern caps. The old caps that have leaked fluid will also be replaced, seen below.

The Standard brand is one I’m not familiar with, but all paper-in-oil capacitors degrade and leak electrolyte eventually. I hate PIO caps for that; I always replace them with Mylar or Polypropylene caps just cuz. The amp usually sounds better to my ear with Mylar or Polypropylene caps, but I’ve admittedly never bought a Mundorf Silver-Gold-EVO to put up against any of my 3-10$ capacitors. The highest quality I’ve probably done were Auricaps, with gold leads (which may have been PIO, I can’t remember). I’m not gonna lie, those HiFi caps from Auricap are just sexy warm melted butter sounding, but no need for hifi caps here. It’s supposed to distort! It’s a guitar amp! I’ll be using Sprague Orange Drop caps or something similar. Good guitar amp caps. The Good-All brand caps are usually in pretty decent shape, over 50 years later, but just for good measure, I’ll replace this one too. It looks leaky.

Someone did a “repair” by adding a press-fit through hole radial leaded capacitor to the decoupling circuit for the PSU (I assume) and made the inside of the unit kinda ugly lol. I’ll try to fix that (it’s the big black fat boy that says “+20% +85*C” on it.) The big blue capacitor in the upper right of the photo is the decoupling cap that goes between the two gain stages in the amp, allowing only AC signals through. This cap will give you your “tone” and character in this amplifier, as all audio must pass through that capacitor to get to your ear though the speaker. I super fun experiment would be to add a switch to the circuit (a high-voltage, make-before-break type of switch) to switch between two types of decoupling cap: a Polypropylene and a Mylar or a Paper-In-Oil, for example. Each different capacitor material and value of capacitance would passively impart different character to the amp. A serious modding bad ass would get a rotary knob and give the user different materials and different values of capacitor to pick from. I digress.

The leaky caps and the lack of a PSU transformer are both contributing to the weak volume and headroom in the amp that the owner complained about. The PSU without the transformer is the reason that the unit is shocking him when he plays and touches a mic stand or another musician. With the additional isolation transformer in the circuit, there is still potential danger from DC shock although the potential for AC shock has been all but zeroed out. Rob Robinette does a great job explaining why in his blog about tube amps (in the post called “the widowmaker” or something similar. God bless Rob Robinette- his page is a wealth of tube information for fools like me. I still have to consult his page on load lines and datasheets for tubes, cuz math is not my core competency. Read his blog post for more electrical info about how to do these kinds of mods. Or you can read my post on restoring an old Silvertone radio with a widowmaker PSU, which I did in 2018.

The new caps will be rated 400v to replace the leaky Standard Brand caps, and 600v for the Good-All cap, which are all the way the amp was designed apparently. I say “apparently,” as the schematics inside the cabinet have been ripped up. The “firecracker” capacitor, the multi-core one traditionally found in radios with the red/ yellow/ green wires, will be removed. The firecracker cap had a voltage rating of only 150v! Super low for a circuit that may see higher than that during peaking events in the mains of the building. The 330uF 400v cap will be removed from the circuit, and replaced with the original spec 100uF/150v electrolytic meant for the tube rectifier. The OPT and preamp stage will sound vastly improved by the use of a power supply transformer. This amp has one major peculiarity I’m not used to seeing: the transformer going between the heaters of the first two tubes with it’s secondary heating the 12v 12AU7 heaters. Below is a photo. The stock 7.5Ω resistor in the original was replaced with this 10Ω Metal Film resistor below. The transformer is used to power the 12v heaters on the 12AU7.

The first two tubes are a 35W4 rectifier and a 50C5 power tube. That means their combined heater voltage is 85v, approximately. They get that voltage from the mains, the same mains line that goes to the anode of the tube rectifier. The two lines are only separated by a 100Ω 5W bathtub-style resistor. The current used by a 50C5 and a 35W4 is .15A for each tube. Since they are in series, I believe they do not require us to add their current together to do our math. The math is: what is the voltage drop over a 100Ω resistor from a 120v AC line that is pulling approximately .15A of current? Ohm’s law tells us that the voltage drop would be roughly 15v at about 2.25w power dissipated through the 100Ω resistor (which is 150ma for the tubes’ heaters.) That leaves us with 105v or so to power the series heaters, which need only 85v. We’re 20v too high! But the 12v tube too uses current. This means the transformer could be a 1:1 or a 2:1 transformer. Let’s assume it’s 1:1, a total draw of 300ma on the line for the heaters with the 12AU7 in series. This makes the voltage across the whole circuit of the filaments drop about 30v over that 100Ω resistor. That gives us 90v, and we need 85 (or a little less). So the math is approximate at best, but I calculate I shouldn’t need more that 500ma or so of power to run this baby in the isolation transformer. That means the Triad unit I bought should be fine.

Another way to do the same math is to mentally remove the 100Ω resistor from the circuit and figure out the math: 85 volts to power the filaments of the rectifier and power tube means a mains voltage of 120 minus the 85 volts we have accounted for in the heaters leaves 35 volts remaining, making the filament transformer for the 12AU7 an approximate 3:1 transforming +/-36v to about 12v for the driver tube. So you can see, adding the 100Ω resistor changes that math slightly, but not the concept. I found this on a search of the net, thanks to the author!


He looks like he’s using a 2A isolation transformer- ugh oh- hope my math wasn’t crazy off, as it tends to be. We’ll see! I can’t see this thing using that much power… Anyhow, look at the link if you’re interested in seeing a better schematic! I’ll try to post a JPEG of it for the next installment.

Well, I think that does it for the analysis and photo session for this cute lil deathtrap of an amplifier. The highlight for me is the UL sticker on the rear- safe my ass. But by the standards of the 50s and 60s, it was safe enough. I’ll also be adding a fuse to the circuit when I add the isolation transformer. Stay tuned later this week for part II where I do the actual work. Cheers!

Modern Grow Light Repair, Part II: TaoTronics 80W LED Fixture (Engineering Reasons This Product Sucks)

HI again folks. Originally, I was going to do this in brief in another post about my repair process. I had some information in there about quality of build, and BIll Of Materials, but I realized this warranted a whole post. Essentially, there is a world of garbage available on the internet, where product producers claim things that they can’t back up. This is one of those cases, and we will soon see how they can offer such a product at under 150$ new. It’s because they’re cutting every corner, and leaving you as the buyer who should have been more wary.

Here is the easiest point to start with. I had to repair it because it was built like crap, and normal wear and vibration caused it to fail. Then he couldn’t get the warrantee department (or whoever) to give up an RMA# for it. Bummer, man…

Poor quality solder, Poor quality layout, mystery LED OEM#s, scary un-certified Constant Current Drivers, barely any Protective Earth wire (very very thin), and a fan driver PSU are all thrown into the light chassis:

Another thing I noticed structurally was that the heatsink could be at least twice the height inside without running into anything. It looks like they just found an off-the-shelf sink and said “good enough.” That’s a theme here. It’s not a bad design ethos, except I like quality driven design better than cost driven design. There’s a place for both, to be sure, but this device takes “designed to maximize profit” to an absurd extreme.

After the small heatsink, feast your eyes on the yellow wire: that’s Earth Ground. Yeah. For an 80W unit that contains 4x separated DC switching supplies going to the same IEC inlet without any additional fusing. Yeah, brotha. No additional fuses. Just 4 PSUs, wired to mains. I like the rust on the metal bands holding the plastic housings of the PSU devices in place. Nice touch. If you compare the two screw heads in the photos below, you can see they’ve been cammed out and they’re not even the same damn screws. One has a flat head, the other doesn’t. Yay, China: “Why even bother to look like we give a fuck?”

Now let’s discuss the circuit: It’s three banks of LEDs, in series. That’s it. Each Constant Current PSU feeds a bank of LEDs in series. Three banks of LEDs, three PSUs inside. The fourth one is for the fans. Crude design, but it’s functional, I suppose. It has many problems.

The PSU for the fans is .8A at 12vDC, constant current, so the fans are sharing the 800ma between themselves. They look like fans that usually have a current of about 1 amp or so at 12v to run in their nominally designed mode. That would mean that the three fans are not spinning at full speed, but rather, they’re slowed by having to share the same Constant Current Source. This is a novel way to keep them quiet. I think this is odd, however: if you gave the 12v DC power to one fan at 1A, it would probably move more air than if you tripled up with 3 fans at roughly 1/3 speed each. I don’t have any good evidence for that, just a hunch. I could be wrong.

Next question: who the fuck is NL Power Supply? I work for Enel North America, and NL Power Supply is confusingly close to my own company’s name. No UL, CQC, IEC, RoHS, FCC, SRRC, CUL, ISO, TUV, or any other Nationally Recognized Test Lab markings on them. Ok. I’ve heard of this: bottom shelf vendors who don’t pay for the certifications, keeping their price to the consumer ultra low. They also risk lawsuit, but as they’re often located in China, lawsuit threats don’t carry much weight. The company making the product using NL Power Supply as an OEM, an Original Equipment Manufacturer, is risking their customers’ safety by using these switching supplies in their products. But they too are Chinese and essentially have no financial reason to care about the people they might hurt in other countries.

The company would have had an OK shot at a NRTL certificate of some kind if they went with Meanwell switching supplies to drive the LEDs and the fans, making the rest damn near certifiable. Labs usually go much easier on your product if you incorporate “Recognized Components” into your design. In the case of this light, after the PSU devices, the rest would be essentially only a construction review of the chassis and finished product. The reason they didn’t design them to be certified by NRTLs to sell in other markets. I’ll get to that below.

They likely didn’t bother to get certified by any NRTL because they intended to crank out 10,000 fixtures in a year and then scrap the product, and all product support. It’s a good way to liquidate components that may not be useful for anything else. I.E. I believe that they designed it with 3 fans because they needed to use as many as they could. I believe they chose the no-name OEM LEDs to liquidate their stock. I think they chose the heatsink to liquidate their stock. It wouldn’t surprise me if the whole thing, PSUs, wiring, IEC inlet, and even the switch and LEDs are all B-, C-, and D- stock from Shenzhen. That would explain everything I’m seeing inside the unit: designed to be manufacturable, yes, but more importantly, it’s both manufacturable and it’s also going to liquidate a bunch of stocked components and it’ll still be cheap for the end user. Win-win-win. Except that the end product usually sucks.

The Constant Current Drivers for the LEDs are 55-100v DC at 600mA. There are 3 inside, one for each bank of LEDs in series on the PCB layout.

This PSU is essentially exactly what is available from MeanWell Power Supplies, except it has no safety ratings or markings whatsoever. The fact that the PSUs were so cheap likely means that they are using sub quality parts, or that the PCBs inside are badly laid out without proper isolation, and no NRTL would certify them. Another further reason not to get an NRTL certificate is that most arrangements, like with UL for example, involve quarterly inspections of the production facility. If the company had a certification with an NRTL, the company would have to commit to inspections. The Chinese fab where these PSUs are made probably doesn’t want anyone telling them how to make the product either, as doing ISO, UL, and others all call for standardizing parts of the fabrication process.

Why constant current drivers? They’re generally better for LEDs. They save you from the concerns about too much current going through the LEDs and shortening the life of the fixture. This type of LED driver is considered to be the better alternative to a constant voltage driver as the constant voltage driver can lead to having varying levels of brightness in the LED matrix. So far, conceptually at least, we’re looking good. Another added advantage is that with a constant current driver is that there are no series resistors employed to limit current, making it easier to drive the proper current to the LEDs.

Why do I have so many philosophical & logical problems with this fixture? Now I will tell you.

First, it’s unsafe. Straight up. No NRTL logos at all? I simply don’t trust Chinese workmanship unless it’s been verified and inspected (final product, yes, but also process and supply chain) and neither should the reader.

Second, it says CE on the outside label, and I am dubious that the TaoTronics ever did the proper thing required by CE certification. CE requires every product sold in Europe have this mark. The company selling the product in question must identify which CE directives apply to their product, test their product to those directives, and maintain a test record and technical file to demonstrate compliance. I have a really really hard time seeing this as being probable for a company that intentionally went to the least reliable source for their power supplies. Also, I believe that any CE directive applied to these products would require NRTL certified power supplies to drive the LEDs.

Third: LEDs and Architecture. The LEDs used are the definition of no-name Chinese B stock or C stock components. I can’t find any data on them whatsoever. This is a problem for several reasons. The first is that they claim that they’re doing “Red, Blue, White, Infrared, and Ultra Violet” color spectra. How are we supposed to verify that if they don’t tell us who made them, and what series they’re from? We can use expensive color meters, but that’s out of the question for the casual 100-200$ purchaser of these lights. Without the LED OEM#s, it’s difficult to confirm any of the manufacturer’s claims. The Procyon 100 fixtures were easy in this regard because I knew they used Cree XLamp brand LEDs, so I could verify the exact spectra emitted from the lights.

The other consideration for the LEDs is this: they are divided into three banks of LEDS, each with their own PSU. Why is this significant? Each bank of LEDs is not a “matrix” of LEDs, like we saw in the other light fixtures, but in fact a series of LEDs, end to end. Each PSU supplies power to one of the three LED series. This is significant because there is the possibility one of the three PSUs going out, and requiring service. No big deal, maybe inconvenient. It could be convenient too though because one could just service the bad one. The real significance occurs when an LED goes out. This would darken every LED following it in the series, which is not a good design. The LED Matrix advantage is from having paralleled and series diodes, allowing for one to go out without effecting the others too much. The only disadvantage to an LED matrix configuration is that when an LED goes out, it pushes it’s current to the surrounding LEDs it’s matrixed with. This can cause the LEDs around a burned out one to be brighter and eventually ware out faster.

Below you can see the shelves illuminated with the TaoTronics LEDs at the lab where I work. They say “white” spectrum, and, well, it’s not.

The final thing to consider is that the real test of these lights would be their power consumption compared to their PPFD output. I don’t have an Apogee meter to measure the PPFD from my fixtures yet, so at the moment, we’ll have to leave our analysis of µMols/Joule for another day. I’ll be getting a meter soon. Then we can really really get into the math of lighting efficiency.

There you have it. Buy Fluence lights, or at least something designed in America. The Chinese can copy stuff (not very well, either, in my opinion) but designing new concepts and new breakthroughs in the world of what is possible, it takes Californians to do it right. Or Texans, in the case of Fluence.

Cheers, folks. Don’t buy TaoTronics, their products suck ass. Just save a little more money and get something well made.

Old LED Grow Light Repair (Part III: Success)

Oh Gawd, the ugly… The good the bad, and this fucking LED grow light. I had to say it, this photograph is soooooo ungodly gross looking. Ok I’m done trippin. Often when doing electronics repair, one must make a mess before cleaning up. I have thoroughly done that. On the upside, I have removed the 3x red dead LEDs, and each blue one in the same series. I figured if I’m going to replace 6 I might as well do all 8 in series right?

Making a mess of a PCB like the photo above requires special tools. I’m not referring to my penis, although my tool is special. I mean this:

An American Beauty brand 250 Watt soldering iron with a comically large wedge tip. This thing weighs a ton, and it’s mass can easily punch through FR4 fiberglass insulator, so using it is kinda messy. It’s my Nuclear Option for dealing with big ground planes wicking heat away from my little Hakko. When the Hakko can’t hack it, get the American Beauty, and be careful not to scar yourself with 1000* F solder. The Hakko is shown below it for size comparison.

Below we can see the new LEDs from Mouser. All that stuff about baking the LEDs etc can be ignored, as we are not doing this “by the book” but instead we are simply doing it right enough to get away with it for a while.

You can see the pores going to the ground plane, all tinned. The rear of the PCB has heat sink grease on it. The ground plane on the rear is actually electrically connected to the circuit. I should have assumed this, but hey, that’s why I’m repairing it: to learn something. Now I know.

The next part of the process is where things get even more critical not to make mistakes: cleaning up the solder pads. With a large enough heat capacity in your iron tip, removing the old diodes is easy: you just slop some new leaded solder into the solder pads, and hold the 250W iron for a few seconds, and the heat combined with the flux in the new leaded solder makes the whole old LED package come off like it’s barely adhered. What’s left over on the PCB surface is tons of ugly, uneven solder. Each tinned portion of the PCB’s traces that are exposed are known in Electrical Engineering as “Pads.” Pads are what you solder to when replacing components. The “Footprint” is the whole shape on the PCB for each component (a footprint consists of several pads.) The key when cleaning solder is not to heat the board too hot. Over heating the PCB risks removing the pads from the footprint altogether, or ripping up the copper traces from the FR4 insulation material the board is spun out of. Not removing the excess solder could result in a short across the very narrow channels between the diode pads, which would blow shit up when we test it. As my jerk of a President would say, “Not Good!”

Removal of excess solder is done with solder wick and a solder sucker. It’s labor intensive, repetitive, and delicate work. our goal is to level all of the solder on the surface, so that when we lay each new LED down, it sits level. After the new diode is in place however, I think the reflowing of the ground pad will probably re-level the diode in place depending on how I do it, so I suppose it’s not that critical. My hand re-work is never going to be exact to how the pick-n-place machine and reflow oven soldered everything initially. Oh well. On with it.

I removed the solder from the pads and then I had to tack one end of the new diode onto the PCB with a soldering iron, and then find a way to heat the ground plane and other diode pad without putting the PCB into a reflow oven. Yaaaaaaaaayyyyyyyy. Fuck. This last part is going to be kinda difficult. I’ll only be out 30$ and a few hours if I botch it and smoke the board, as the light itself is of negligible value so I suppose it’s a good time to give it a shot. I usually doubt myself before I get at it, so this is no exception. Below you can see the result when I was finished after about 30min of soldering and 4 red diodes I accidentally cooked or covered in solder.

To make it easier, I covered the pads on the LEDs with solder before applying them to the heated pads on the PCB. The diodes are marked with polarity by a notch on one side of the LED package when viewed from the top down.

I used the 250W iron to bottom heat the ground plane and pads with the huge tip, and it worked beautifully! Each LED floated briefly on the surface tension of the solder until it fully liquified and the LED settled nicely on it’s footprint. I tested the diodes with a Multimeter; there are no shorts, there is continuity between LEDs, and the LEDs illuminate when tested. Finished-ish! Awwwe shit, son!! I feel big, like not in the sense of gaining weight, but like if I could take on lights this bright I could knock out Mike at a Tyson fight. Word.

Moment of personal criticism: I didn’t do as good a job on this as I’d wanted to. The iron was too hot at 250W, and it burned the FR-4 fiberglass insulation in several places around the pads, resulting in big black unsightly burns in the surface of the nice white solder mask finish. I also removed the solder mask on traces between certain diodes too, which is pretty whack. I meant to make this a better showcase of my skills, but I suppose just having the balls and skills to even attempt this is pretty swagmatic. I digress. Below is a photo of the damage done. Not extreme, no damage to functional PCB parts, but still… ugly.

After I melted the solder through the ground pores from the back the ground plane was all uneven. Any chunks of solder will be undesirable when heat sinking the unit back the the aluminum. So I used a smaller soldering iron to smooth the plane out before reassembling the unit. Below you can see a big solder chunk on the rear ground plane.

This ground plane and the surface of the PCB will both remain a little ugly as the unit is reassembled. It is what it is. Can’t make it perfect. Frankly, I’m surprised it’s gone this well so far. You can see old heat sink grease in the pores of the ground plane.

Above we can see the shiny-ass alcohol and water polished block of aluminum, which has already had new heat sink grease applied. If the grease has separated, as mine had, you must mix it with a popsickle stick or a piece of plastic so that it is consistent again. Then in this case, I determined it was a good idea to take the grease and spread it evenly, without any metal showing, to get a more-or-less even amount spread over the whole block. The finished result is visible below.

Once I had put the new grease on the sink, on I carefully laid the PCB down after I’d cleaned the ground plane as best I could. I then bolted the two torx heads furthest from the heat sink grease first to center the board, and then I went in order to the next bolt until finished, doing the two closest to the camera in the shot below last. That way the pressure was being applied to the PCB against the heat sink from the bottom of the assembly up. I’m hoping that that process of torquing the bolts down eliminated the air pockets that may have formed in the grease under the PCB. Then I cleaned up the excess grease, of which there was not much.

Below you can see I then soldered the fan back onto it’s pads, and the thermal cut-out switch onto it’s pad, and the MOSFET Q1 has been bolted back into place.

Tharr she blows. Now that she’s reassembled, let’s fire her up. Ugh. Anxiety. I spent too much money and time on this. Plus, I feel shitty about the burns and flux on the front of the LED panel but it is what it is.

I almost forgot! Finally, a date code! OMFG- 7/08. As in July 2008. Holy Crap, I never would have repaired this if I knew it was so old. That’s the only purpose to have text on the silkscreen is as a date code: there is no component that number corresponds to.

Oh well. Test time!

Again, I have a knack for shit photography, but I did also repair this grow light. If you look closely, the 8 new diodes are leaving brighter, wider points of light on the iPad’s video sensor die than the other older diodes. I’m actually surprised the iPad can deal with that much light. It’s enough to damage retinas for sure. Huzzah. Holy cow. I still don’t know what I’m capable of.

I’m actually really quite proud of myself. Certifying products for UL and others is boring as shit sometimes, and it’s usually all drama when it isn’t boring. It’s nice to have some time on the work bench while they unfreeze our corporate account this week. It’s nice to know I’m still better at engineering than paperwork.

Cheers. I’ve earned a tremendously thunderous bong toke, and a Mystery Science Theater 3000 episode before I pass out in my underwear. Ahhhh, the glorious life of an engineer continues, varmint!

Old LED Grow Light Repair (Procyon 100, Part I)

As you can see from the photo above, there were and still are plenty of LED lights in the world that use the older technology of diodes tailored to the McCree plant sensitivity curve. This is one of the earliest brands and models that was available back in the day (+/- 8 years or so ago) from homegrownlights.com: the Procyon 100. I don’t know why they called it the “100:” the AC voltage is 120 and the wattage is 125W. So neither spec would make it a 100- anything unit, but that’s neither here nor there. Maybe they thought it would be a good model name because its easy to say and remember. They marketed them as being a replacement for a full 600W HPS light system, each. They are definitely high output: Cree brand XLamp LEDs, at 3 Watts each I believe. I may have an old brochure around. I’ll have to look. Sorry I got stoned after my evening xanax dose and I’m watching The Dark Power with Lash LaRue. It just seems like peppering my speech with the occasional “varmint!” is something that could benefit me. Whipping Native American zombies, like Lash LaRue? Hell no I don’t endorse that. I’m not sure what he was thinking- whipping natives is not a good look.

These units are PCBs bolted to a ceramic substrate that in turn is bolted to the heatsink with the fan. Not a great way to sink heat, and these units generate a ton of it. Why? Great question. Let’s ask that question. Before we do, remember one thing: these were made by hippies in Mendocino a long time ago, and the likelihood that any of them had time enough in the engineering industry to create a better product is low. These people’s mistakes and obvious flaws have given us the lessons that led to our modern horticultural age in which LEDs can beat out any HID fixture in the field. I’m about to be critical, but they did the only thing they needed to do right, well, right: they chose Cree XLamp LEDs. Those babies are bad-ass for sure. Not as cool as the Samsung LEDs used in the Fluence lights, but pretty dope. I’ll grab a datasheet to compare the two OEMs at some point. Below is a close-up of the dies inside the diode structures. The blue dies are white with a swirling pattern etched onto the dies, with bond wires in the center of the swirls. The red dies are beautifully red and look like squares within squares within squares with bond wires at the center, varmint!

Ok. First thing is first. The designers made a critical mistake. They chose to include the PSUs for the LEDs inside of the enclosure, making it only necessary to have an AC cable. Why is this a design mistake? The power supplies driving the LEDs are going to get hot, and lose efficiency. This, in turn, will load the power supplies further, causing more heat. The component life of everything in the PSU has been shortened considerably. And why? To have a “simple” one cable and plug solution? Not ideal for the obvious reason: if your PSU goes out, your have to throw the unit away or repair the whole thing. Having a separate LED driver (Power Supply Unit) is handy too because you can cool it on it’s own, which makes the PSU’s efficiency less influenced by the heat produced at the PCB level by the LED array emitting light. An additional added benefit to having a separated PSU is that if your diodes stop emitting, and changing the driver doesn’t help, you know you’ve got a problem with the diodes, Varmint.

Now, to the fault: Two sets of 4x LEDs out in an opposite symmetrical orientation on each side. I believe this will not turn out to be faults in the diodes, as in I don’t think the diodes are “burned out” or damaged, unless it was by heat I suppose. But heat would have damaged all of them I would think, not just 8. Now that I follow the traces in the above photo, I think I can see why it’s two sets of 4: the LEDs are being driven in a matrix. 4 Diodes in series, and each set of 4 is paralleled to it’s neighbor sets of 4 in series. It appears to be a matrix set up with 4 in series in the center and 3 in series on each end, all paralleled in sets of 4. Ok. Dayum. So now we know two specific sets of 4 diodes in series are not illuminated, which means that 2 of the matrix inputs are not being driven (we can assume, varmint)

So, how is this matrix being driven? It’s just a big ass DC Power Supply Unit in this baby, putting direct current to all of the diode inputs, making them emit constantly as long as the light is plugged in, right? Well, my guess is no. I believe that to lower thermal dissipation in the unit, to extend the life of the LEDs, and to lighten the load on the PSU, the likely thing is that the AC-DC supply on the PCB then supplies pulsed DC to the diodes. I think there will be some kind of driver IC pulsing the DC that has failed to turn those rows of LEDs on. Otherwise it would use too much power and emit too much heat. Being that it’s got a mostly metal enclosure it would have been easily certified for emissions (although these were never certified by anyone, varmint!)

The PSU is a little mystifying. It appears to be at least one switching power supply, maybe two as there are two rectifiers (lower left high current bridge and the smaller lower current bridge in the center of the photo.) The Triad branded device is a common mode choke to remove noise from the power lines. MOV1 is a Metal Oxide Varistor used for safety purposes. The NTC is a current inrush limiter similar to ones I’ve implemented on my home made tube amplifiers, varmint!

I have a hunch IC1 could be our target. D1 is also a good bet to investigate, as it seems to be the current passing device that drives the matrix (due to it’s proximity to the grid.) So I Googled 9910B datasheet, and lo and behold, we have a weiner, I believe: Microchip branded HV9910B open-loop current mode LED driver. At this time, I’d like to point out the number of vias I can see already, which is 6 in this photo, and that makes me shudder a little bit. Power supply rails and feedback senses don’t like little vias when driving lots of current. It squishes the electrons together and creates more heat on the PCB as they travel around the circuit. I design my PSU layouts with no vias. It’s slightly more time consuming, but it makes for a better product in my opinion. There was also a ton of room unoccupied on the PCB, and it was not necessary to make their vias so small, varmint! Upon seeing the image below, I realized this was going to be a pretty straight forward datasheet implementation with a few twists and turns.

The ZVN0545 SOT223 package, the little four-pinned guy next to the electrolytic can capacitor, appears to be a small N-Channel mosfet. I believe that this is used with the 9910B to activate the LEDs. I think the output gate driver for the 9910B drives the gate of this MOSFET. The MOSFET seems to be rated AEC-Q101 for high reliability. That was a good choice for this circuit. It is also Halogen and Antimony free. Indeed, this is a mad decent MOSFET. R3 is a 27kΩ resistor connected to pins 4 & 8 of the HV9910B, which means we have a driver operating in what the manufacturer describes as “constant off-time mode.” R3 is indeed an oscillator timing resistor for the driver IC. Now we hit a snag: the datasheet from Microchip onlt has a typical implementation schematic for the other operating mode for the IC, the Constant Frequency mode. Shit, varmint!

From what I’ve read about digital Power Supply Units, I am aware that PWM controlled buck or boost power supplies usually operate in a PWM mode, constant on-time mode, or a constant off-time mode. That will be for another blog post. Basically, when operating in a current controlled constant off-time mode, the inductor’s ripple current remains constant, which makes the output filter for the power supply have much less expensive requirements (at least with a boost converter.)

OK, varmints. I’m getting dragged into a dogshow here with all this jive. I’d like to reverse engineer this whole power supply and matrix to show what all is going on from an electrical standpoint… But something started to bother me as I was taking bong hits again. The Matrix. Not the movie, man, but our LED matrix. It seemed to me that the only thing that might account for the LEDs being out in the sets of four would be that the driver IC was not addressing those addresses in the matrix. As I thought about it though, I realized we are missing any IC to do digital addressing to the LED matrix. What’s more is that I can’t find any vias between the paralleled sections of the matrix either, suggesting that they aren’t connected to another part of the circuit; they’re there for redundancy! If a diode in a row of 3 or 4 diodes in series goes out, so do it’s companions in the strip. Adding the parallel connections allow the current to travel through other rows of diodes to the parallel busses, powering the other LEDs “around” the failed set. Ugh. Why the fuck not put parallel connections after every LED? It’s the same cost to make the PCB for God’s sake, and it would have isolated the burned out diodes from the start of the analysis. Below you can see a diode working nicely under test.

Below you can see one of our offending diodes. It’s not emitting light because it’s emission junction has turned black from over voltage or over current, or (what I suspect) is a badly applied thermal management technique. If it were over voltage or over current, the driver would be having problems, and everything until the matrix of LEDs itself looks nominal for the design.

So we have a result, which is exactly the opposite of our initial assumption. The initial assumption was that we would not find any burned out LEDs. That assumption caused me to look in the wrong places: clearly the driver architecture and PSU design was still intact; all of the LEDs were lit up (except the bad ones,) without flickering, and without blowing up other LEDs in the matrix. This should have immediately pointed me to do diode tests. However, I had been under the assumption that no diodes were bad because the of their fixture with the same number of hours on it had no burnt diodes. So instead of testing the continuity between the diode packages (checking traces for damage) I should have definitely also checked the diode packages themselves. Upon diode checking with a multimeter, I determined that several diodes are burned out. I also can see the emission junction inside the package turned black (see below, the center of the photo, a red diode with a big black emission gap burn inside the diode itself.)

Concluding for now, varmints! We’ve discovered several offending diodes. Another interesting anomaly I have discovered is that the LEDs on board the unit do not seem to illuminate at the same brightness when checked with the multimeter. This suggests that they either are not matched or that the time spent in circuit has caused them to wear in the amount of flux energy they can produce from a given voltage. Looks like the next part will be concerned with sourcing a decent replacement from the myriad of XLamp products Cree makes, and maybe some matching analysis and datasheet comparison between the Cree brand that Homegrownlights.com used and the Fluence brand’s choice, which are new Samsung LEDs. Then desoldering these motherfuckers to replace them, which I am confident I can do but I am not excited about. See ya then, varmints!