Amplifier #19, Godzilla the Kaiju

“Fuckin’ a, man.” -The Dude

This is a photo of my 19th amplifier to be completed. It’s really something else. Don’t let the 4 tubes all alone on the chassis make you think that this amp is simple- it’s not. It’s a rather wild ass design from Japan, and my first Japanese style build. I actually love Japanese tube amps, because their designs are very transformer centric, and often almost “too simple,” (such as Dr. Sakuma’s amplifiers) or just the opposite, and they are very complicated and extreme like Mr. Michimori’s Eimac 100TH monoblocks he presented at the European Triode Festival a decade or more ago. I love both styles of design. I really like the use of choke based CCS (Constant Current Sources) and transformer coupled stages common in most Japanese designs. I have adopted many of the same methods of driving stages in my amplifiers I’ve made because it makes the amplifier’s math more symmetrical, and loads the impedance of the devices better as volume changes and transient levels change during listening.

Below is the original schematic I used as my jump off point. All credit and praise due to, Sakura Chiba, Japanese tube enthusiasts, and the Japanese Single Ended Amp community. The below design is theirs.

For the inspiration, I started with a schematic on, which is credited there to Sakura Chiba. Mr. Mizushima’s blog is always interesting, even if the blog can be in broken English. This design is called the “Chorke-loaded Cathode Drive” on the blog, but I assume he meant “Choke-loaded Cathode Drive.” The benefit to doing this is that the the choke on the cathode acts as a Constant Current Source for DC and as a basically limitless AC impedance, thus AC decoupling the cathode from ground in the 2nd ECC83 stage. That cathode is then directly driving the 6550 power tube.

The amp I built, while based in many ways on the one shown above, is very different in it’s final form. I had a few issues with the above amplifier. The first is that I wanted more power for my output devices, the 6550’s. So I chose a transformer that was 700v CT instead of 640v CT. I went with a Hammond power transformer and a Hammond power supply choke. People talk shit on Hammond for HiFi use, but I have never had a problem with a Hammond power transformer or choke, and I defy anyone to show me a difference on an oscilloscope between a Hammond transformer and another E+I core transformer of comparable availability. Their chokes also seem to be on par with any expensive HiFi oriented power supply chokes and are often easier to get. I also like the black finish. I also elected to use the Hammond 156C open bracket 150Hy 8ma chokes in place of the 80Hy 10ma chokes in the schematic. I really doubt there are better for cheaper as far as small signal AC decoupling chokes than the Hammond 156C. I’ve used it for doing choke loaded plates and cathodes, and I’ve also used it in parafeed plate and cathode output configurations. The power transformer is Hammond 273BX, which is rated to 201ma, and the choke is 200ma.

Above and below are photos of my hand ass Compu-Bias meter measuring the octals under real-world amplifier conditions. Another lil note: the 6550’s don’t share a cathode resistor, because the amp is not differential. In a differential amp, the shared resistor between the pushing and pulling power tubes is an integral part of how the circuitry works. It also causes the two power devices to share current at idle , which in turn makes each dissipate less heat. In a single ended amplifier, the only real purpose to sharing a cathode resistor between left and right is to cut the overall current traveling through both tubes, a la the Decware Zen’s EL84s. In the Zen, simply adding a second 1% matched resistor to match the first 100Ω cathode resistor on the other EL84 takes the power output from 1 watt per channel to about 5 watts per channel when the tubes do not share a cathode resistor. The other reason not to share cathode resistors is that the sharing of the cathode load between tubes adds cross-talk between channels as the L and R signals’ negative feedback will mix with one another as they degenerate over the cathode load resistor.

Now that I do the math, 201ma was kinda light on the current, as 90 + 90 = 180ma right there. 90ma is a totally average quantity of cathode current for the 6550 to pull too. Oops. I can actually go in and change the power transformer out for another in the future. I’ve already done it with the choke which was originally 150ma, and clearly not robust enough. I then used that choke in my Single Ended Parallel Triode build (post coming up!). When I do the mod I aught to also make room for a discrete power transformer for the heaters in front of the choke, in between the driver tubes. Then I can replace the power transformer with a 375-0-375 CT Hammond at 300ma+ from the 7 series of Hammonds.

In this case though, we can see that even though the 6550 itself can dissipate far more heat than the ~25 watts we intend to get out of each. Why don’t we push our numbers and performance to the limit? Well, the answer is that with tubes the more power your putting through them, generally the more harmonic distortion you’ll see. The second answer is that while our idle current is set at roughly 75ma each tube during idle, peak currents will be higher when the amp is in use. At a plate voltage B+ of 355v or so, 75ma of cathode current is ~26 watts of heat dissipation per tube, which is what the power output transformers are rated for. Thankfully, at 75ma each, that’s only 150ma of current from the power transformer, which is an OK (but still not great) amount of leeway. Also thankfully, the tubes’ biases don’t seem to change when the UL/Triode switch is flipped.

The original called for a tube rectifier. To hell with that- if this thing is supposed to be loud, like 50 watts of loud, then it’s time to use HexFREDs to rectify power. The omission of the tube diodes free up the design to use some serious electrolytic capacitors in the power supply too. Solid state diodes can provide much larger pulses of rectified current than vacuum diodes- amps compared to milliamps. The solid state diodes also “stiffen” the power rails because they don’t sag voltage as bad as vacuum diodes when under a heavy load.

I also elected to use more high voltage (500v) capacitors in my design than the original did. Each one has a 200kΩ resistor at 2W and a 1.5kv 0.01uF RIFA capacitor in parallel with it, unlike the schematics above. It prevents the power rails from breaching the 500v ceiling I have on my capacitors during a cold start, and also discharges them to prevent shock after power off (should the amp need service.) I used Kemet screw terminal caps from the UK, 2x at 220uF 500v and 4x at 100uF 500v. They work super well, and my power rails in this SET are fucking silent as the grave. I’m consistently shocked at how this amp’s finished product has no noise I can discern in the signal path. Most of my other SETs have more noise floor because of the poor power supply ripple rejection at the power tubes in SETs in general. The capacitors were anchored with generic capacitor clamps I got online.

I used a pair of Edcor XSM 600Ω:10kΩ input transformers for the input instead of using a volume knob. I prefer input transformer coupling to voltage dividing. The 600Ω Edcors can be Center Tapped at 150Ω as well, giving the designer the flexibility to add a switch to switch between the two impedances on the input. The fact that the transformers are also 2.5 watts each is also nice, as they can be used for all kinds of impedance matching needs at some decent power levels.

Then the rest of the amp is very similar to the schematics above, until you get to the second ECC83 stage. Visible in the photo above is my personal touch on Mizushima’s brilliance. I used a fully realized parafeed cathode load to drive the 6550 tubes instead of directly connecting the AC decoupled cathode of the ECC83 to the 6550 grids to drive them. I actually somewhat dislike the concept and experience of building directly coupled tube devices. I attempted one early on, and the constant flashover problems inside of the bottles, and the eventually dead tubes, left a bad taste in my mouth. Plus, I’ve seen Loftin White amps explode too, so I’m not a fan of direct coupling. So in this case, I elected to use the cathode choke as a sort of cathode-parafeed interstage by adding an XSM 600Ω:15kΩ as an interstage transformer, with the cathode driving the 15kΩ winding and the grid of the 6550 being driven by the 600Ω winding. Direct-coupling problem solved! And I’d like believe I solved it using the same sort of design aesthetic as a Japanese designer would have, too. The Parafeed gain stage is a very Japanese HiFi style solution to a problem: just use a transformer and a really killer set of capacitors. Everywhere you can.

In this case the coupling capacitors are 1200v .1uF Silver foil Z-caps from Jantzen. Why? Well, honestly, I just wanted to see what the fuss was about. They sound good. I think they may be my new favorite capacitors for their price. I think after buying the Vishay MKP caps from Mouser, it’s a high bar to cross, and since they can be almost comparable in price, I may only actually like the silver ones better conceptually. Silver should sound better, right? The one thing I’ll say for the silver is that they sound smooth. Everything sounds smooth, tight, clear, and very sexy coming through the Silver Z-caps for sure. Its so very hard to tell what component is accounting for what quality of sound in an amp of this rigorous a build. Is it the caps? The signal path? The unusually beefy power rails? The Output Transformers? The tubes themselves? Hard to say, folks. My money is on power supply beef, tube selection, and transformer drive techniques for why this thing sounds so fucking crazy good. Sorry to keep cursing, but this amp is stunningly beautiful sounding.

I don’t like Single Ended Triode style designs, generally: their poor power supply ripple rejection, their fetishization by enthusiasts, the constraints on the transformers, and the power supply factors all make me less than excited about them as a topology for a vacuum tube amp. By copying Mizushima’s 6550 SET design and improving the choke-loaded cathode drive method into a fully realized parafeed-cahode loaded gain stage, the 6550 SET’s design was improved into a new creation. I love the sound quality of this amp, and although it’s only 4 tubes, the Tung-Sol 6550’s with the 3x getter are really impressive looking when paired with the little ECC803S’s and all can be seen to glow in low light. It may be a SET style amp, but it’s THD+N is so low it sounds like it’s not even on when it’s got no signal, and it’s so clear it sounds like a low wattage low distortion push pull- except it’s packing an eye watering 50 watts RMS output.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s