• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

Wollcott cross coupled circuit

SY

diyAudio Moderator Emeritus
2002-10-24 10:19 pm
Chicagoland
www.SYclotron.com
One of my correspondents sent me a link to Henry Wollcott's site:

http://www.wolcottaudio.com/WA_whitepapers.htm

wherein Wollcott describes, in as obscure and unreadable manner as he can, a cross-coupled circuit with some unusual feedback paths. The closest thing to a schematic I can find is in Wollcott's US patent 3328711. Drawing attention to Figure 6, one can see the multiplicity of feedback paths which form the core of his invention.

Is anyone familiar with further work that has refined the original concept shown in the patent? Has this idea been used in OTL amps? And can somebody please tell Henry that white type on black background is horrible to read and that a few drawings would make the text much more understandable?
 
From my read of the Wolcott paper, just a couple of changes are required to fig. 6 in patent 3328711. Replace the tube/transistor input pairs with two 6DJ8 White cathode followers and add coupling capacitors from the White outputs, cross coupled, to the grids of the triode gain tubes (#90 and #87 in fig. 6) ( the cathode drive connections to 90 and 87 stay). The second stage gain triodes and the pentode followers after them would be 6GW8 tubes. The 6DJ8 White cathode followers are used as low impedance drives in order to handle the direct cathode input couplings to the 6GW8 gain triodes. (ie. each White cath. follower drives a gain tube grid and the opposite gain tube cathode)

The major key to the operation of the circuit design would appear to be revealed in Wolcott's patent # 3111630 on nested feedback. Fig. 6 (in patent # 3328711) does include this nested feedback topology. The scheme uses positive feedback from the opposite EL34 grids to give very high gain in the 6GW8 triode gain stage. Bootstrapping of the 6GW8 triode's plate load is used to keep its gain linear to avoid high distortion from positive feedback (although a separate CCS load would be better). (The early Philbrick Op. Amps used a similar scheme of positive feedback for a high gain stage)

Overall the scheme is one of conventional global neg. feedback from the tertiary feedback winding on the output xfmr. and the xfmr primaries (EL34 plates), with likely some frequency selective dropout of the tertiary winding feedback at high frequency to avoid instability due to phase shift.
Also not shown in fig. 6 is a current sensing resistor in the 8 Ohm output used for "positive" feedback to null out xfmr secondary resistance.

The major claims to fame seem to be the all symmetrical circuitry (with some rather fantastic and not likely true claims of cancelling out odd harmonics too - at least other than from normal NFB operation), and secondly the use of positive nested feedback to produce an extreme high gain stage (6GW8 triode) for high loop gain for the global NFB loop. The claim that clipping also kills the super gain pos. feedback so as to produce smooth clipping, while still using high loop gain for normal operation is an interesting one (but I don't see why the EL34 grid drives would drop out in amplitude during clipping, so I have some scepticism here). Global negative feedback is also taken from the primary windings to allow low phase shift in the global NFB loop so as to allow the use of high loop gain and provides low output impedance to avoid xfmr magnetization distortion and better xfmr bandwidth too. The claim that the dual input White cathode followers somehow cancel out each other's noise contributions I find rather dubious however.

I would say this was an interesting/clever solution for its time, but it is clearly obsoleted nowadays by either Edward Cherry's nested feedback approach or Hawksford's error correcting ideas.
Cherry's nested feedback does not require a near unstable positive feedback high gain stage and Hawksford EC only requires unity gain around the feedback loop. (although modest conventional global neg. feedback is still required for overall stability in H- EC)

my $.02 take on the circuitry :)

Don
 

SY

diyAudio Moderator Emeritus
2002-10-24 10:19 pm
Chicagoland
www.SYclotron.com
Don, many thanks. It's interesting that Crowhurst wrote extensively about the benefits of nested feedback, including the advantages/disadvantages of a positive feedback stage inside the loops, long before this patent... The Philbrick amps, IIRC, used a cathode-to-cathode resistor in the cascaded gain stages to provide the positive feedback. Dyna did a similar trick in their tubed phono stages.

I am drawn to your statement about noise cancellation. Certainly, thermal and shot noise will be uncorrelated between sections and won't cancel. But hum and ripple will be correlated and should cancel.

Why do you think his claims of odd harmonic cancellation are probably not correct?
 
new and improved model

A fixup to the Wolcott design so that it would definately produce smooth clipping would be to use partial cathode output windings as well as plate windings (ie, partial cathode feedback outputs). This way when clipping occurs, the EL34 cathodes would lose amplitude. By moving the positive feedback pickoffs for the high gain 6GW8 stages to the EL34 cathodes instead of the EL34 grid drives, the gain would drop under clipping conditions. This would then give smooth clipping. Unfortunately, this mod. would probably increase the distortion of the high gain stages (and the whole amp subsequently) significantly during normal operation, since difficult load impedances would lower global loop gain.

Another circuit topology I have been playing with recently would also be a major replacement for the Wolcott design. The Wolcott design uses 12 actual tubes (well 6 tubes with just two EL34 outputs). It only takes two extra tube elements (ie, one dual tube for an LTP diff. amp. stage) to add either Hawksford error correction to the output stage of a conventional amplifier, or to add active ultra-linear operation to an output stage's screens (see recent thread on ultra-linear operation, the LTP controls the screen grids for linear output stage gain instead of using xfmr taps) It is possible to use the same added LTP stage to do BOTH Hawksford EC and active-ultra-linear operation at the same time, since the LTP performs the same error sensing function in either scheme. So, for say a three stage amplifier with 4 tubes, one more dual element tube (5 tubes total) gives linear performance beyond anything ever seen. One would still use the xfmr primary as feedback pickoffs for modest almost global NFB.

Don :)
 
Hi Sy,

Yes, the hum and ripple should cancel in the fully symmetric design. High PSRR.
On the odd harmonic cancellation, these distortions look like either signal compression or signal expansion, ie, they are symmetrical with signal polarity. I don't see where cross coupling, or difference of differences as the Wolcott paper calls it, would fix this problem. But I would be very excited if someone did come up with a solution. Perhaps some simulation of the cross coupling approach would be illuminating to see what it achieves (or doesn't).

If one side were to compress signals by the same amount, whenever the other side expanded signals, we would be onto something here. But I think the problem is that these opposite actions won't track with signal amplitude. Usually one side will become more compressive (or expansive) at a faster rate when the other side is becoming more expansive or (compressive) at a slower rate.

Don
 
Hmmm, I suppose, for small signal levels in class A, the expansion and contraction could.... match up. So there could be some truth to the cross coupling removing odd harmonics. But I think this cancelling will fall apart quickly at higher signal levels. Sort of like a normal LTP where the two sides curvature match up well at low levels, but here we would be matching up expansion/contraction. Then again, the cross coupling is not really the difference (or sum) of two LTP stages, so some simulation would be illuminating on this issue. It would be nice to come up with at least something to cancel odd harmonics.

But in any of these harmonic cancellation schemes, including vacuum tube current mirrors, we are likely to only get the lower harmonic numbers to cancel out. Two (or more) tubes aren't likely to have matching high harmonic distortions from bent or off center grid wires. And the 3/2 power curve doesn't have a linearly varying slope so that we can sum complements to get constant gm. (Mosfets do however, over a limited range, square law curves giving linearly varying gm.)

Don
 
odd harmonic cancelation

Seeing as how odd harmonic cancelation is probably the most interesting issue on the Wolcott or other cross-coupling type designs. I thought I would stick my neck out here and suggest a powerful solution.

Vacuum tube current mirrors nominally can cancel all harmonic distortions for perfectly matched tube characteristics and almost invariably provide significant current gain (20x to 100x). In reality, they will only be capable of nulling out the lower order harmonics well due to inevitable tube mis-matching. But the design of the current mirror inherently is a good fit to the Hawksford error correction technique.

Usually Hawksford EC is only applied to low or unity gain output stages, but the current mirror circuit has two active devices which are supposed to give matching emitter, Oops, I mean cathode signals (small current degenerating or sensing resistors in the cathode circuits of the output tube and input thermionic diode, scaled according to final current gain). This unity match of cathode signals is perfect for an LTP stage to compare and calculate the difference error. The inverted error signal output of the LTP is then fed back to the input as a current summed with the actual input current, with unity gain, so as to cancel the error. This unity gain has to be set accurately with a trim pot. Greater than unity gain becomes an oscillator, unless constrained by an additional global neg. feedback loop in the amplifier with greater than unity loop gain.

The Hawksford EC thus functions to near perfectly match two already closely matched tube characteristics. The vacuum tube current mirror typically uses a thermionic diode and a pentode, and the added LTP will require two more tube elements (a dual pentode would be ideal, 6J11, 6BN11, 6CB11, 6BV11, it may also be possibe to do with just a dual triode LTP by inserting the correction signals into the mirror stage cathodes, not sure about accuracy yet). (Hawksford and/or associated graduate group has previously published on the idea of error correction applied to current mirrors, so it could be patented, I haven't checked.)

Don
 

SY

diyAudio Moderator Emeritus
2002-10-24 10:19 pm
Chicagoland
www.SYclotron.com
Geez, Don, have some coffee today? :D :D

Seriously, thanks for the extremely thoughtful and thorough comments. Regarding clipping, I'm more of the school that feels that overload recovery is more important than how the clipping happens. But it's certainly a debatable issue.

Your point about cancellation in class A is absolutely spot on- any balanced class A stages can beneficially run feedback between polarities, but in AB stages, the feedback needs to be kept within a circuit half to be useful.
 
Stuart and Don,
I initially sent you the Wolcott link and asked if it might make a good thread for discussion.

I was never able to "nut out" the full circuit details from the White Paper despite filling 6 or 8 A3 pages with pencil scribbles.
This coupled with the fact that the design is tending toward a Solid State "Infinite Gain/Infinite Feedback" approach and may therefore sound cold, sterile and boring lead to my interest waning (I put it in the "too hard basket"). How about that for a "juicy rationalization".

I'm going to chase up those patent documents and have another look.

For those wanting to look at this design from cold - its basically a Van Syoc (spelling?) front end with white cathode followers (6DJ8) in lieu of the more usual ordinary cathode followers to better drive the cathodes of the cross coupled (2nd stage) triodes - the triode section of 6GW8/ECL86 (same as a 12AX7 section). For the happy hybrid types - chuck the white cathode folowers and substitute a couple of opamps.

The triode loads are bootstrapped for maximum gain. These are followed by cathode follower (the pentode section of the 6GW8)to drive the output tube grids.

Large amounts of local feedback around the output stage (back to the 2nd stage triodes) thus removing any sonic signature of the output tubes themselves (it'll sound the same with any of EL34, 6L6, 6550, KT88 or whatever).

Several other feedback loops including global voltage and current loops.

The current feedback allows tuning of the output impedance to any desired level including negative output impedance if desired.

Its a bit beyond me today BUT hopefully it won't be beyond me next week or next month with a bit more study.

I am also looking at another circuit which bears a liitle resemblance to the Wolcott front end ( in that its an extension of the Van Syoc) - the Kimmel Differential Mu stage.

Thanks for the above - we less experienced tube nutters find these sorts of posts very educational - well I do anyway. After all if we knew it all we'd be one of those lowest of all types, a "Knowitall", a fate to be (easily) avoided!

Cheers,
Ian
 

SY

diyAudio Moderator Emeritus
2002-10-24 10:19 pm
Chicagoland
www.SYclotron.com
Ian, thanks for suggesting that we open up discussion on this. After getting a second note about it, I thought that it might interest more than two of us and so it has. Don't feel bad about the slow wading through the "white paper." That paper is meant more to obscure than to illuminate, so you're to be congratulated for extracting useful content. A friend of mine who is certainly one of the foremost authorities on tube circuits read it and stopped halfway through with a headache.

Are you familiar with the White Powertron? There are definitely echoes here.
 
Stuart,
Thanks - the one main area I could'nt work out was the method of bootstrapping the 2nd stage triodes.

Has he split the anode load and capacitor coupled the cathode follower output to the mid point or is he floating the entire triode section on the cathode follower output or WHAT?

Can you enlighten me (and others).

Also I can't seem to find the US Patent Document 3328711 - do you have a link?

Thanks again,
Ian
 
Jason,
Ta mate. turned into an epic to get AlternaTiff loaded and running but I now have the two patents in front of me.

The bootstrap method is the split load resistor with capacitive coupling from the cathode follower output to the midpoint of the split anode load. Picture is worth a thousand words.

Now if I can jusy find some stuff on the White Powertron I'll be in happy as a dog with two tails.

Cheers,
Ian
 

EC8010

diyAudio Moderator Emeritus
2003-01-18 7:57 am
Near London. UK
Two tails?

The White Powrtron (sic) amplifier is described on pp13-16 of Vol 3 of the "Audio Anthology." The amplifier uses a van Scoyoc phase splitter and although the article doesn't achieve the inpenetrability of the Wolcott piece, it makes a stab at it. If I remember correctly, there is also an updated version of the article floating about on the web.
 

ThorstenL

Previously: Kuei Yang Wang
2002-11-10 5:35 pm
www.facebook.com
Konnichiwa,

SY said:
Is anyone familiar with further work that has refined the original concept shown in the patent?

Without being aware of the Patent and only from hints in Wollcotts Whitepaper I experimented with PFB to the driver stage myself. I also experimented with PFB to the Outut stage screens.

Combining both you can make an amp with astonished measured 1KHz performance, but I felt the sound was much worse than the simpler circuit I used without PFB in both stages.

This may very well be down to my inability to implement it well, but I had a bloody good stab at it.

I filed the Wollcott approach away as something for the Meter Readers.

Still, in an OTL the extra PFB might help more than it hurts.

But why not simply make the OTL suitable in design to work with minimal NFB. An idea of mine would be a IGBT style structure (an N-Channel IGBT can be pictured as compound feedback pair of N-Channel FET and PNP BJT Output) with a Valve and a bank of PNP Transistors in a circlotron.

If you use directly the Base-Emitter Junction to the Anode (no base/emitter resistor) and use a Valve that is effective at sufficiently low Voltages you get a very efficient hybrid that is very close to an "all tube" design and has low distortion and low output impedance without any other trics, of course it still has "local" looped NFB around the compound pair, but so what....?

Sayonara
 

EC8010

diyAudio Moderator Emeritus
2003-01-18 7:57 am
Near London. UK
I thought some of the diagrams in the Patent looked familiar. After much searching through the horizontal filing system, I unearthed a "Low-Distortion Sine-Wave Generato.r" (Arthur R Bailey. Electronics Technology. February 1960. pp64-67) that uses a cathode follower to bootstrap the load of the preceding pentode, enabling the pentode to achieve a gain approaching mu. The large gain is then degerated by negative feedback and the oscillator is claimed to have distortion <0.02% from 10c/s to 100kc/s. My friend who sent me the diagram did so because he couldn't seem to remove squegging from his version...