Re-building Audio Mirror amps for 6C33 triodes
Copyright 2011 © Troels Gravesen

What if?

12-01-2011
My PSE 6AS7 Audio Mirror monoblocks have served me well for more than four years now and from time to time the idea of rebuilding the amps for 6C33 tubes was on my mind, not least due to a friend's Amplifon WT40, a 40 watts PP design. The sound from this WT40 more than suggested the 6C33 tubes might deliver a sound at least equivalent, if not superior to the 6AS7 tubes, and getting more than twice the output doesn't hurt either. We may choose flea powered triodes for a lot of reasons, but bass grip or dynamics is not their middle name.

Based on a lot of web search, the 6C33 tube has a lot of aficionados and the discussions on how to use the tubes best are endless. Dos and don'ts by the numbers and people are much divided when it comes to how these tubes should be treated. The 6C33 tubes are often used by OTL builders, but also PP designs seem popular. For SET usually only one tube is used, where PSE versions seem rarer.
The Audio Mirror OPT and PSU are dimensioned for 6C33 tubes, but also take 6AS7 by replacing fixed bias with auto bias. Removing the cathode resistors leaves quite some space as can be seen from the pics below.

So, I mailed Vladimir at Audio Mirror to hear if he would provide the schematics for the 6C33 amp and not only did I get a prompt reply, but also the schematics for an improved version where fixed bias is replaced my an N-channel power mosfet circuit, that is: Auto bias.

Vladimir:
I will suggest you to try auto bias for the 6C33C. I recently did that and I am impressed. It depends what kind of sound you like of course. With negative voltage (grid biasing) to the grids, these tubes sound more dry, a little harsh, with extended mids, but not very good low end. All this is compared to autobias version. Autobiased version is very, very dynamic. Very fast! I can't believe what bass I've got from tubes with output transformer. It is unbelievable. I also got smoothened mids and highs versus the little hard and dry mids and highs from the grid biased version. Also no more bias checking, re-biasing or hard to bias or out of range tubes. I really recommend you the autobias version.
About teflon sockets - surprisingly they are no good at all. I had a few amps with the same sockets installed by customer request. All of them failed. They have badly machined metal inserts for the tube pins. I don't know what kind of metal they use, but I had hard time to desolder (after the amps has been sent back for replacing the tube sockets with ceramic ones) the wires from pins. Also they looked like the metal burned, dark oxide film over them. I am still using with no problems the Chinese ceramic ones. (either regular or gold plated). One more thing, be prepared for a lot of heat. Twice the heat you have now coming from 6AS7 amps.

If you read the tread here: http://www.goodsoundclub.com/Forums/ShowPost.aspx?postID=6140#6140, there are a number of things here that shouldn't be done, not least auto bias, but I decided to take my chances with Vladimir's suggestions.

Having done the mods according to Vladimir's recommendation I can say that this is the most transparent amp I've ever had running my speakers. Next, as Vladimir says, the bass grip and dynamics are phenomenal. Bass transients have a punch not even the JungSon's 80 class A watts can do much better.

Audio Mirror Website

The thing about this amplifier is heat and hum. The latter I'm sure I'll overcome; a minor residual hum that isn't audible from the listening seat, but noticeable when standing next to the speaker. 95 dB speakers are not forgiving.
Heat is the primary caution and installing a heat sink for the two mosfets is next - as soon as I find one that fits into the chassis. Measuring temperatures with an infrared thermometer proves stainless steel not to be an efficient heat conductor and I would like to be able to solder the cathode wires to the mosfet's drain pin without having the solder melting after a few minutes.
What's seen below has now been running for many hours and appears so stable I even let my wife alone with the amps. Initially I had a 6C33 tube arching after a few hours and despite a spectacular view, something that takes many hours of careful surveillance before confidence is restored. I guess I ran the amps too hard before they had really fully burned in.

30-01-2011
There's a 2R4/5W resistor in the mosfet circuitry and initially I used 2R2 as 2R4 is not available here in Europe. To reduce the current flow a little, Vlad suggested a higher value like 2R7 or 3R3. The voltage drop across this resistor (2R4) should be around 620 mV. This was what I had with 2R2, hence drawing some (0.62/2.2) 280 mA. Vladimir also suggests not going below 200 mA, so I tried 2R8 (2x5.6 ohms in parallel) and measured ~600 mV = 214 mA. Much better! The mosfet only reaches some 155^oC compared to 185^oC before. I think I'll settle for this and leave out the planned heatsinks.

Click images to view large

Basic principle of Audio Mirror 6C33 PSE auto bias version. Please do not ask for full schematics, proprietory to AM.

Fixed bias version

17-04-2011: To make a long, long story short, the input stage of this design proved seriously unstable and after frying a lot of mosfets, the direct coupled input stage was abandoned in return for the original design including fixed bias. This version may not have the magic touch of the auto bias, but something I can live with. Thanks to Morten for lending me his 6C33 amps for copying.

Once again these chassis' were emptied and cleaned for glue and solder spots.

Even the PSU board was cleaned with acetone. To the right the fixed bias set-up. 
For 6C33 cathode resistors some thickfilm will be installed later. I needed this one running fast and used 20 watt wire-wound resistors.

Now, setting up this amp - although taking quite a few hours - was expected to be an easy task. Not so! Having plans for a second 6C33 amp, I purchased two Lundahl LL1693 OPTs. These can be wired for 300 ohms primary impedance and are to my knowledge the only off-the-shelf OPTs available for this application. Obviously other manufacturers can wind OPTs for two 6C33 tubes if asked.
How to wire the 1693 OPT can be found at JacMusic/Germany: http://www.jacmusic.com. 
Click Techcorner and next schematics and we're into a "4tubes" website. Find "tube schematics with Lundahl transformers": http://www.4tubes.com/SCHEMATICS/LUNDAHL-TRANSFORMERS. Click SE-amplifiers and next 2x6C33 PSE:  http://www.4tubes.com/SCHEMATICS/LUNDAHL-TRANSFORMERS/SE-AMPLIFIERS/2x6C33B-PSE-with-LL1693/. Click schematic. 
Yes, this is well hidden and maybe it's good so, because what we find is wrong. Before wiring the OPTs I asked Lundahl if the shown schematics were correct and had a positive answer. So, I wired the OPT according to Jacmusic's recommendation and turned on power. All voltages were in place in a few minutes and it's really nice to have an identical amp working on the side for reference!
Adding a signal to the amp sounded promising at low level, but turning up the volume produced serious distortion. After a couple of months' agony over the mosfet version, this was not exactly good timing! All possible causes were examined and no errors could be found. Adjusting bias on one of 6C33 tubes had a profound impact on speaker level, although still seriously distorted. However, this made the OPT wiring questionable. The long and the short of it is that the two primary windings of Jacmusic's schematics leaves the two in opposite phase. Suitable for push-pull, I guess. The problem is this:

To the left Jacmusic wiring. Notice the +-marking (encircled in red) has changed compared to Lundahl specs despite the wire numbering being the same. At the bottom the wiring I used leaving the two primary windings in parallel as should be. There are two options here, letting each tube see its own 600 ohms primary - or connecting the two primaries in parallel (300 ohms) and running wires from a central point to tube anodes. I don't think it makes any difference.

So, time to power up the amps again - and sweet music came through the speakers! Finally! The Long and Winding Road...

First of all this set-up has very low hum and trying to get even lower I supplied shielding to the filament supply wires. Minor improvement. The 100 ohms resistors from filament wires to ground is something that has to be tried. I found the shown ground resistors to produce least hum and I have to put the ear to the pulp to hear hum from my 95 dB DTQWT speakers, so no further work in this area for the time being. Maybe a double choke some day (from Lundahl).
For C1 I use 220n tin-foil/teflon and for C2 I use 2.2 uF Mundorf Silver/Gold. They really don't need being this big.
Silver-plated copper in teflon cables were used for general wiring and 0.4 mm pure silver for signal. 

These Lundahl OPTs indeed seem very good. With regard to transparency I don't think there's any difference from the auto bias version. Vlad's comments on sonic differences may be just a tiny bit exaggerated and from pics provided it appears his amps were fitted with Sonicaps, which to my ears are seriously bad and impairs mid and treble performance. With the best coupling caps available, I think the differences are small. The auto bias version may excell in bass performance and a certain holographic magic to the midrange, but for the time being I can live happily with these amps, truly the best I have had running my DTQWTs, OBL-11s, QUATTROs and even the PRELUDEs, the latter really not at thunderous levels.