RCA OP-6 Analysis

[Winston OBoogie]

Can anyone throw any light on why one end of the OPT primary is connected to the cathode of V3 rather than ground?

The cathode resistor is not by-passed, does this type of connection get around the resulting higher internal resistance?

DaveP

Hi Dave,
I'm very rusty on thinking this stuff through so:  Would it not, as far as the transformer in concerned, act as if the cathode were bypassed without having to actually use a bypass cap? 
The cathode is really the reference the input signal wiggles around not ground after all.  Hmmm?

 

[Winston OBoogie]

To clarify, I mean that the loop is closed around input to output to the same point in which both move together. 
Otherwise, we put a cap on the cathode to fix it and return the output to ground which is, we hope!, also fixed.
Q.E.F.  one less cap in the loop.

 

[emrr]

Sowter sells the V72 input transformer, which can be connected as 1:20.

But most don't want that much gain in the transformer.  It's an easy place to reduce gain for most modern uses.  The occasional person wants to dedicate it to low output ribbons, but all tech requests come with a demand for phantom power. 

To clarify, I mean that the loop is closed around input to output to the same point in which both move together. 
Otherwise, we put a cap on the cathode to fix it and return the output to ground which is, we hope!, also fixed.
Q.E.F.  one less cap in the loop.

Not sure.  (Maybe faulty) memory says it's a 1-2 dB gain, as opposed to the 6dB of cathode bypass, plus the improved noise path.

 

[Winston OBoogie]

Not sure.  (Maybe faulty) memory says it's a 1-2 dB gain, as opposed to the 6dB of cathode bypass, plus the improved noise path.

OK got it.  I suppose I can see that it wouldn't be a free lunch all round as far as gain is concerned, just don't ask me for mathematical proof! 

I can see the noise advantage too since I guess the cathode resistor is out of the equation as far as output is concerned. 
And any circulating ground current in that area....  One more low frequency pole from the cathode cap is gone too I guess.
You've seen way more of this stuff than me Doug. 

 

[DaveP]

The other thing that interests me is the phase difference between cathode and plate, its kind of like a push-pull circuit as the two ends of the primary are out of phase with each other.

DaveP

 

[Winston OBoogie]

The other thing that interests me is the phase difference between cathode and plate, its kind of like a push-pull circuit as the two ends of the primary are out of phase with each other.

DaveP

I already reached my current level of pay grade, I got nothing much more I'm afraid. 
Both ends of the primary would already see opposite polarity regardless of the connection to K or G but...  ? 
Maybe that's the mechanism whereby we get the extra 2dB gain that Doug mentioned?

 

[DaveP]

Well I decided to test it with a 6J7 and a Hammond 156c choke.

Very interesting result:-

With OPT primary earthed at other end gain is 112.8

With OPT primary connected to cathode, gain is 134.4

So it looks like this connection takes the place of a bypass cap as suspected

DaveP

 

[emrr]

Interesting Dave. 

Am I converting correctly to arrive at roughly 1.5dB difference? 

That would not seem to entirely replace a bypass cap.

Why would one be chosen over the other?

Collins 6P is returned to cathode AND bypassed with 50 mfd.  Same with RCA 40-C/D, and the 76 console PGM amp. 


Some interesting stuff here:

http://www.nutshellhifi.com/library/Rosetta_Stone.html

Connecting the return of the primary to the driver-tube cathode gives a further improvement by shortening the AC loop around the audio circuit; in addition, careful ratioing of the cathode-bypass and transformer-return capacitor can essentially cancel most power-supply noise. (A vacuum tube amplifies the difference between the cathode and the grid; if noise that appears at the plate can be reduced in the right ratio at the cathode, it will cancel PS noise appearing the plate.)

http://www.nutshellhifi.com/library/Tube_Fest_Talk.html

 

[DaveP]

Yes it is 1.5dB, thanks for the very interesting WE link...........forgotten skills!

Losing a by-pass cap increases rp, which is not a good idea for an output stage, I will see if I can measure that with different cathode  options.

DaveP

 

[emrr]

Remember in this and many cases with RCA it’s a 50K primary.  Same with lack of bypass cap, regardless of where cathode is returned. 

OP-6’s especially were hung on 600 ohm phone lines.

 

[DaveP]

I tested V3 on its own, no V2 or feedback loop.

With no load output was 88.4V and with a 51k load it was 22.63, so this made an output Impedance of 148K, but it made no difference if the cathode was by-passed or not.  I put this down to the Z of the choke which swamped the rp.

Just for interest I tried V3 as a 6SJ7.
Plate voltage was the same but it drew slightly more g2 current and so Vk was 2.42 instead of 2.33V.

This time the output Z came down to 144k so the rp of the 6SJ7 must be slightly lower as expected.

One interesting point was that in both cases, the waveform was better with a 51k load than without!  The load must alter the angle of the loadline to a more favourable position.  These RCA guys knew their stuff!

DaveP

 

[emrr]

Very cool Dave, nice to see those results. 

 

[DaveP]

So, as the gain of V2 is 100 and that of V3 which is 134.4, we have a open total gain of 13,440

But with 2.2k/560k feedback loop we get a closed gain of 250 with 34.4dB of negative feedback.

34.4dB is a factor of 53.593.  So as our output impedance before feedback is 148k, when this is divided by the feedback factor, we get an output impedance of  2.76K, which is well able to drive  the primary load of 50k of the OPT.

50k:600 OPT's are not easy to find nowadays, so we would probably use a lower rp tube like a triode or even a cathode follower,
with a 15k:600 OPT, but back in the day RCA had no problem using a pentode for all three stages.

DaveP

 

[ruffrecords]

So, as the gain of V2 is 100 and that of V3 which is 134.4, we have a open total gain of 13,440

But with 2.2k/560k feedback loop we get a closed gain of 250 with 34.4dB of negative feedback.

34.4dB is a factor of 53.593.  So as our output impedance before feedback is 148k, when this is divided by the feedback factor, we get an output impedance of  2.76K, which is well able to drive  the primary load of 50k of the OPT.

50k:600 OPT's are not easy to find nowadays, so we would probably use a lower rp tube like a triode or even a cathode follower,
with a 15k:600 OPT, but back in the day RCA had no problem using a pentode for all three stages.

DaveP

As I am all too fond of pointing out, low output impedance is no indicator of drive capability - which is largely down to quiescent current. Also, calculating closed loop output impedance the standard way as you did only applies to small signals. Despite those caveats it is a useful figure to know. Transformer distortion for example declines rapidly with source impedance and of course a low source impedance leads to better LF response.

Rupert Neve's first tube console used an EF86 as the output tube with the output transformer as the plate load. I would not be surprised to find he was familiar with the RCA circuits.

Cheers

Ian

 

[vinyvamos]

Loving this thread guys! Big thanks to DaveP for initiating all of the headscratching and calculations in order to understand this beast .

I have been messing with this design for a while now myself. I have built one pair with 6SJ7, 15K:600 Edcor OP trafos, and 156C chokes (V1 as triode with no fancy gain attenuator) and am now building some that are pretty much replicas of my original RCA OP6. What it is falling down on right now though is inter-winding capacitance at the output trafo (further referred to as IWC here). The original measures just 200pF @ 10KHz between primary and secondary. Every other trafo that I measure that might be close WRT to L and V-ratio seems to have too much capacitance between windings. Even an original V72 OPT, which means that surely sowter's remake of it will have the same issue.

We are talking 1.5nF for the V72 OPT, and then Carnhills VTB2380 has 2nF. VTB2380 has only 115H primary L so C13 should be increased accordingly. And as for the L of the V72 OPT that is up on 400H! Edcor are not even worth trying, as they have huge IWC.

I have found that the VTB2379 is an ideal plate choke, with it's 500H and it's 5900Rdc, plus they are CHEAP! I have compared them to a single 156C and two 156C in series (hum-bucking?) and LF headroom is reduced when only one 156C is used.

So to divulge what I am having an issue with, when testing headroom at various frequencies the HF headroom drops off with output iron that has a high IWC. I am using 10KHz as a spot-point. But it depends how you test it. Normally I am grounding pin 3 of the XLR output, and my scope, VU meter and 600R load sit between pin 2 and GND. This is giving me the worst results, as the AC plate voltage of V3 has a route straight to my common bench ground/mains earth due to IWC, therefore creating attenuation and nasty distortion at HF. If I test it going into a modern electronic-diff. input which we could say has >10KΩ to ground from pin 2 and pin 3, and use the balanced input on my scope, then I get a few more dB before the wave goes from sine to triangle.

RCA's original trafo has just 200pF. Headroom is the same at 10KHz as it is at 1K and 100Hz. Note that the headroom of a V72 is lower than the OP6 (isn't it?). Does anyone more knowledgeable care to discuss this further?

DaveP, in the OP6 unit that you built you had designed a 12 step version of the RCA NFB/volume attenuator. May I ask how you might calculate such an attenuator? I am trying to calculate out the RCA design here on Excel but it is not making sense. I reverse-engineered my original unit so I have all the values.

 

[DaveP]

Not having an original to copy, I started from the idea of a ladder attenuator, which I believe the original is similar to.

I then realised that the final output  resistor should be, say 1Meg as it's the grid resistor for V1.

As the feedback resistor is 1.2 Meg it was possible to estimate the total resistance of the string to get the required feedback level for that resistor.  I did not calculate it as it was far easier to set up the first tube and work out the steps by trial and error.  I literally soldered a long string of resistors together and modified the values to get the dB steps, it took about an afternoon to do .

https://groupdiy.com/index.php?topic=73451.msg929615#msg929615

Hope that helps

DaveP

 

[emrr]

As well, I sent Dave my reverse engineering of the original attenuator several years ago.

 

[DaveP]

Yes, I had a lot of help from Doug as well.

DaveP