I have a beat up old Dynaco Stereo 70 that's been pretty well stripped down to its transformers and tube sockets. I'm trying to settle on a parallel-NFB 'Schade-style' circuit to build into this chassis.
Since the chassis is small, I'd like to keep things as simple and compact as possible.
Basically, I'm trying to adapt the DCPP 'Engineer's Amplifier' to push-pull 6L6GC in the original 'ultralinear' configuration with Dyna A470 OPT. with the original Dyna power transformer delivering about 430V B+. So far, I have this:
Looking at the load lines, I don't understand why the voltages I see in LTspice aren't matching those in the 6CB6 data sheet. The voltages LTspice calculates for the 6CB6 LTP do match what's on the DCPP schematic.
LTspice is predicting 0.055% THD at 1W out into 4 ohms, and 25W at clipping from 1.5V peak input.
Once I settle on a decent circuit I'd like to make a set of PCBs the same size as the Dyna ST70 PCBs. I know it will be a tight fit.
What do you think? Did I do it wrong?
Since the chassis is small, I'd like to keep things as simple and compact as possible.
Basically, I'm trying to adapt the DCPP 'Engineer's Amplifier' to push-pull 6L6GC in the original 'ultralinear' configuration with Dyna A470 OPT. with the original Dyna power transformer delivering about 430V B+. So far, I have this:
Looking at the load lines, I don't understand why the voltages I see in LTspice aren't matching those in the 6CB6 data sheet. The voltages LTspice calculates for the 6CB6 LTP do match what's on the DCPP schematic.
LTspice is predicting 0.055% THD at 1W out into 4 ohms, and 25W at clipping from 1.5V peak input.
Once I settle on a decent circuit I'd like to make a set of PCBs the same size as the Dyna ST70 PCBs. I know it will be a tight fit.
What do you think? Did I do it wrong?
The input is unbalanced. The 6CB6 LTP is the phase splitter. The CCS made with M1 (DN2540) provides the high impedance in the tail to enforce balance.
The NFB is around the output tubes. provided by R15 and R16, from plate to grid of each 6L6GC. I can decrease the value of R15 and R16 to apply more NFB.
I can add a global NFB loop if necessary. If I do, the final amp would have 'nested' feedback loops (one overall gNFB loop including the OPT, with the local plate-grid parallel feedback around each 6L6GC, 'inside' the gNFB loop).
The NFB is around the output tubes. provided by R15 and R16, from plate to grid of each 6L6GC. I can decrease the value of R15 and R16 to apply more NFB.
I can add a global NFB loop if necessary. If I do, the final amp would have 'nested' feedback loops (one overall gNFB loop including the OPT, with the local plate-grid parallel feedback around each 6L6GC, 'inside' the gNFB loop).
IMHO for input diff.stage best performance C2 have to be tied to the ground line and not on the junction point of DN2540 drain & 6CB6 cathodes ,or you can try input diff. stage without C2 .
I don't have any comments regarding the design, just want to mention you cannot take THD from LTspice too seriously. It's a great sim tool, but doesn't always get THD correctly. So I would not let that hold me back.
Thanks everybody.
Finally, it seems I get the best looking performance with R5 and R6 at 220k, which is the value from the DCPP circuit. THD goes up and sensitivity goes down if I decrease those to 150k ohms (a lose-lose proposition). 220k is 6.667X the value of the 6CB6 plate load resistors (33k), which is 15% NFB in Schade's terminology. Gain reduction from NFB is only 7dB (-2.25X). I suppose the best thing to do would be to use 220k for R5 and R6 and add perhaps 5dB global NFB from the OPT 16 ohm secondary tap to grid of U2 (6CB6).
Thanks again for your comments.
It seems this circuit likes a common 22 ohm resistor to ground, from junction of R17//C5 and R18//C6. That decreases the predicted THD by half. But...
Yes, I agree. However, I believe LTspice can tell you which version of a circuit will generate less THD than another version. I don't trust the absolute amount of predicted THD, though. It varies quite a bit depending on which model(s) you use, and what parameters you define in the .tran directive, etc.
I tried without C2 and THD increased a bit. Then I tried C2 tied to ground instead of to the 'top' of the CCS 'tail' circuit and that decreased THD by a bit. Understood that this is all in simulation, not measured in real life. However, I thought with pentodes, best practice is to bypass the screen to the cathode, not to ground. Both are possible and will work, but which connection is better?
Finally, it seems I get the best looking performance with R5 and R6 at 220k, which is the value from the DCPP circuit. THD goes up and sensitivity goes down if I decrease those to 150k ohms (a lose-lose proposition). 220k is 6.667X the value of the 6CB6 plate load resistors (33k), which is 15% NFB in Schade's terminology. Gain reduction from NFB is only 7dB (-2.25X). I suppose the best thing to do would be to use 220k for R5 and R6 and add perhaps 5dB global NFB from the OPT 16 ohm secondary tap to grid of U2 (6CB6).
Thanks again for your comments.
Actually, I think that would be three NFB loops.
- OPT primary UL taps
- Plate-grid shunt NFB on output pentodes (R15 and R16)
- global NFB from OPT secondary to U2 driver (grounded) grid
That looks something like this:
You have two of the three loops that are symmetric (UL and Schade), why don't you try symmetric gnfb as well?
Ground the 4 Ohm tap of the secondary, and connect the 0 Ohm tap to one cathode of the 6CB6, and the 16 Ohm to the other, through resistors.
You added a resistor on the left of R17 and R18, try a resistor on the right side of it too, going to the joined C5 and C6.
Do you have the LTSPICE file of this circuit?
Ground the 4 Ohm tap of the secondary, and connect the 0 Ohm tap to one cathode of the 6CB6, and the 16 Ohm to the other, through resistors.
You added a resistor on the left of R17 and R18, try a resistor on the right side of it too, going to the joined C5 and C6.
Do you have the LTSPICE file of this circuit?
Best connection for C2 is definitely to the ground point , since that 2,2uF capacitor connected from 6CB6 cathodes via R6 and C7 to the ground line will have negative effect on the diff. input stage performance on the higher frequency , simply 6CB6 common cathodes signal have to float on the higher possible impedance AC point .
Some points;:- The 6CB6 was produced in mass for the 1950-70´s TV industry, and like others in the video framework one runs the risk of wild specs. but like so many mass produced types of that era, who cared about thd in video?. Electrode voltages and resultant thd for audio can vary enormously between same makes. I found GE types far more consistent with 3rd harmonic thd than RCA, Sylvania and others which produced masses of 2nd harmonic signatures.. Currently my unfavourite job is sorting my store of 12BY7 into matched pairs for diff driver application., (the tube itself is a high gm flyer) but as others mentioned of using a Spice program to get an idea of thd, which I found hopelessly out. A example, using 12BY7´s, Spice came with with 3% thd on a home made diff driver, whereas on the bench I got 0.5% using an HP distortion analyser.
I can recommend when ordering ex video tubes to request matched pairs; by golly this can make things more predictable.
My experience, ex video tubes can sound harsh on audio; so to circumvent this, invariably I have to use 10-20dB global nfb to tame the spec from misaligned voltages.
The 1W distortion level is somewhat meaningless. See what Spice comes up with at the 20W level, build the design and check it with the real load.
I can recommend when ordering ex video tubes to request matched pairs; by golly this can make things more predictable.
My experience, ex video tubes can sound harsh on audio; so to circumvent this, invariably I have to use 10-20dB global nfb to tame the spec from misaligned voltages.
The 1W distortion level is somewhat meaningless. See what Spice comes up with at the 20W level, build the design and check it with the real load.
What small signal pentodes were made specifically for audio? I can only think of 5879 and EF86.
Would 6AC7 be a better choice than 6CB6?
I'm using LTspice simulations to narrow in on a possible 'best' circuit. I don't believe it can predict THD with much accuracy, but if I make one version of the circuit whose simulation shows 0.03% THD at 1W out into 4 ohms, shouldn't that be better than a different version of that circuit that shows 0.1% THD at 1W out to 4 ohms? Or is even that meaningless?
Would 6AC7 be a better choice than 6CB6?
I'm using LTspice simulations to narrow in on a possible 'best' circuit. I don't believe it can predict THD with much accuracy, but if I make one version of the circuit whose simulation shows 0.03% THD at 1W out into 4 ohms, shouldn't that be better than a different version of that circuit that shows 0.1% THD at 1W out to 4 ohms? Or is even that meaningless?
I'm finding that if the simulation shows low THD at 1W out, it also shows a lower level of THD at 20W out (THD is proportional to level).
6EW6, 6JC6A high gm IF tubes. 6JC6A is frame grid.
The best 6JC6A tubes for triode mode I've come across were some Magnavox (Japan) and Westinghouse (Japan) 6JC6A ( curves below, mu around 60)
The best 6JC6A tubes for triode mode I've come across were some Magnavox (Japan) and Westinghouse (Japan) 6JC6A ( curves below, mu around 60)
I did collect some 6JC6A pentodes, but I never got any 6EW6.
Those 6CJ6A triode curves sure are pretty.
Those 6CJ6A triode curves sure are pretty.
I'm looking at 6JC6A and 6JD6 load lines. It looks like they like to operate with more plate current than 6CB6 or 6CF6. That could be a problem, since I'm limited in that department to about 270mA total plate current for 4x 6L6 and the four driver pentodes. 12HL7 looks like it needs a lot of current, like a power tube. Too much for this application, I'm afraid.
Here are load lines for 6JD6 and 6JC6A with Rp = 22k and 370V B+. Do these look up to the task?
Here are load lines for 6JD6 and 6JC6A with Rp = 22k and 370V B+. Do these look up to the task?
Last edited:
When Pete first sprung the big red board on the world, I got one stating that I would be highly disappointed if I didn't get 50 WPC. I seriously underestimated the power of the big red board since I eventually saw 250 WPC and 525 watts in mono with both channels in parallel. Back in the long thread on this amp I stuffed virtually every possible 7 pin tube available that could possibly work in the circuit into the board, and every possible output tube regardless of what socket it used. The 6CB6 is an IF amp tube that runs at 45 MHz in a US TV set. As stated, they are all over the place in regard to tolerance of the specs. I had to go through several dozen to find a matched pair plate voltage wise. The 6EW6 is a better choice, but the best 7 pin tube that I found for the DCPP is the 6GU5 which is a hexode. Drop it into the big red board with no changes and get lower THD with more gain, a win-win when driving big tubes to 125 WPC. Pete must have agreed because he used the 6GU5 in one of his mono block DCPP derivatives.
Wow, thank you!
I have 10 of 6CF6, close equivalent to 6CB6, which I was hoping to use. But if those are going to be a huge pain, maybe I should just sell 'em off...
Since 6CF6 and 6GU5 share a compatible pinout, maybe I should build the prototype with 6CF6, get it working, and hunt down some 6GU5 later.
I have 10 of 6CF6, close equivalent to 6CB6, which I was hoping to use. But if those are going to be a huge pain, maybe I should just sell 'em off...
Since 6CF6 and 6GU5 share a compatible pinout, maybe I should build the prototype with 6CF6, get it working, and hunt down some 6GU5 later.
Last edited:
Pete's DCPP boards have balance pots for the 6CB6's (or whatever you're using), so the tubes should balance unless the specs are wildly different, which is entirely possible. So don't see why you couldn't incorporate this feature in your amp. Or did I miss that already.
jeff
jeff
I was just reading some old posts about the DCPP PCB. George mentioned he uses only about 6 to 8 dB NFB combined, and preferred shunt local feedback around the output tubes but didn't like the sound from the combination of shunt-local NFB plus gNFB. I guess I'll need to experiment with this since I'm adding the complication of UL operation.
I wasn't thinking about balance pots, but I can install a couple of 100R trimpots on the PCB to attempt to balance the 6CF6s. No reason not to add this, even if I end up with 6GU5 or whatever.
I wasn't thinking about balance pots, but I can install a couple of 100R trimpots on the PCB to attempt to balance the 6CF6s. No reason not to add this, even if I end up with 6GU5 or whatever.
Reading Pete M's DCPP page a bit more, I see that he derived the -60VDC supply from a 50V-0-50V secondary on the Edcor power transformer he specified. I'm using a Dyna PA-060 power transformer, which has a 55V 'bias tap'. If I use a full-wave bridge on that bias tap, do you think it can deliver the -60VDC at 35mA I'll need for the tail supplies to the two 6CF6 LTPs? Do you think the PA-060 bias tap can stand up to that abuse, or do I need to add a separate, small transformer for that -60VDC supply?
Last edited: