Another go at PP -> IT -> PP

[Max N]

Four of the amps mentioned in the corner of the room.

It certainly seems odd seeing one of the best amps I’ve ever heard sitting there

[Max N]

Why not just add a gain stage in front of the interstage. Solid State maybe balanced in balanced out, or just a simple 5687 LTP. Thats all you are doing by using a DAC with higher level outputs.

Sorry, I was probably a bit defensive earlier - maybe because where I have ended up is sounding really good.
I think the obvious thing for me to do for more gain would be 45PP -> IT -> 6B4GPP. The 45 stage would give me around 10db I think and I have got that to sound pretty transparent in the past.

[Nick]

No sorry needed.

[Max N]

Back to the story....almost up to date now
I spent a few months with the RME DAC before the Sowter 9063s arrived. The RME didn’t give me much more gain than the Focusrite Forte, but I didn’t expect it too. Having lived with it for a while now, I love the features of the DAC. I do use the parametric EQ. The inbuilt ‘bit-perfect test’ is handy. I think I can tell the difference when I switch between some of the filters, now that I know what to listen for, but I doubt if I could in a blind test - it’s very subtle most of the time. I also like the fact that the gain is adjustable, so whatever level your downstream equipment wants, you can always use the full dynamic range of the DAC. Really there are too many features to list, I recommend a look at the manual on-line if you are interested.
What about the sound? It is very transparent, but not in a bad way. It doesn’t sound digital at all. I don’t find myself thinking about whether I should investigate hi def sources, or censoring the playlist to avoid certain music.
Then the Sowter 9063s arrived. They are 100% mu-metal and do not tolerate any dc current. Because the RME is dc-coupled I was a bit worried (although the RME support had assured me the DAC output is very well balanced). Sure enough, measurement revealed about 2 micro amps on one channel and nothing measurable on the other, so all good. The gain of the transformers is perfect, just what I needed. They handle quite high levels, they are rated 24 dBu at 50 Hz, and I have found that I can only saturate them at 5 Hz driving them almost full beans from the DAC, so they are a good match I would say. My first impression when I started to listen was that the bass was tighter, faster more agile, more definition.
But some aspects of the sound had gone backwards. I spent ages trying to figure out where the problem was, and I did find a few problems. But I was still scratching my head when Nick made his ‘simplicity and dogma’ post. He was right of course, so I started retracing my steps.
Swapping the ITs back to the bifilars, the problem was gone but I was now missing the improved bass of the Sowters. So I decided to do what I should have done when the Sowters arrived, and test them. The problem was soon apparent, and I could have kicked myself. The data sheet shows a ruler-flat response, and they were indeed flat up to 20kHz, but I was measuring a big hump above that all the way up to about 90 kHz. Doh, doh, doh - as the data sheet clearly shows, they are designed to work into 10 kOhms. I actually found about 15 kOhms sounded best, but I will probably revisit that.
With the transformers correctly terminated, everything started to come together. I was also now hearing what the DAC is truly capable of, which is quite special.
Finally, I have a sound which I am happy with, everything sounds good, but there is still some magic.

[Max N]

There is one thing I would actually like to try, which is a version of the bifilars on 100% mumetal cores. If I could locate cores of the right size, I could maybe dismantle the bifilars and transfer the bobbins to the new cores?

[Max N]

Gave myself another talking to. I should really have a go at a Lynn Olsen-inspired gain-stage on the front end. I'd be a fool not to try something suggested by Nick....
So, taking a look at Lynn's pages, I was surprised to see that the power supply of my 2A3 output stage is similar to the supply of Lynn's input and driver stages. I'm using an MJ statistical regulator on each channel to feed the centre tap of the output transformer. I'm using 5W zeners in the shunt, and they seem happy taking the full 150mA until the bias control brings the 2A3s in

I don't have the parafeed cap, actually I don't have a bypass cap on the zeners at all. But I have tried temporarily holding a 20uF PP cap where Lynn has the parafeed cap, and I couldn't hear a difference and I couldn't see a difference on the scope.
My cathodes are at ground potential because I'm using negative bias, but I don't think that makes a big difference.....

[Nick]

Just to be clear, The best result I got was not using the original Raven circuit with choke loaded anodes, but using a CCS on each anode and then the output transformer between each anode with a blocking cap. If you meed any depletion mosfets to make up the CCS's just let me know, I have lots, so happy to put a few in the post for you. I prefer the IXYS ones to the DN2540 ones as they seem to be less sensitive to static when assembling. But I have some of those as well and the IXYS current regulators.

The blocking cap becomes essential if you have nickel in the transformers. I would use it anyway to isolate the DC and AC current paths.

[izzy wizzy]

You might already have seen this https://www.diyaudio.com/forums/tubes-v ... -pull.html
I found it focusing and reckon it's not far off where I'm headed in terms of topology.

[Max N]

Just to be clear, The best result I got was not using the original Raven circuit with choke loaded anodes, but using a CCS on each anode and then the output transformer between each anode with a blocking cap. If you meed any depletion mosfets to make up the CCS's just let me know, I have lots, so happy to put a few in the post for you. I prefer the IXYS ones to the DN2540 ones as they seem to be less sensitive to static when assembling. But I have some of those as well and the IXYS current regulators.

The blocking cap becomes essential if you have nickel in the transformers. I would use it anyway to isolate the DC and AC current paths.

Hi Nick, just trying to picture the circuit. So with the CCS’s on the anodes, I can see that the AC can only flow through the primary of the transformer. And if the two triodes are not perfectly matched, some DC will also flow through the transformer, which would be a definite no-no for the Sowters. But where is the blocking cap? Are there two caps, one from each anode to each end of the primary? Or are there two primary windings, with the cap between the two?

Yes please, I would like to try the IXYS depletion mosfets, I’ve only tried the DN2540’s. I’ll pm you.

[Max N]

You might already have seen this https://www.diyaudio.com/forums/tubes-v ... -pull.html
I found it focusing and reckon it's not far off where I'm headed in terms of topology.

Thanks for the link, I’ve had a quick look, but there’s a lot of stuff in there so I’ll need to spend time digesting

[Nick]

If there are two separate primaries then cap in the middle. Otherwise one on one side.

[Max N]

Got it now, thanks

[Max N]

I haven't had much time to experiment with parafeed. All my breadboards have got really messy and I need to have a good sort out. But before I started that, I did do a few tests. Wasn't able to listen, but did take some measurements. This is with the large bifilars, which should be OK with some DC. I also have some Sowters which definitely don't support DC, and I will try with those as well. I think Nick's use case was with transformers that didn't support much DC. This test is just a low frequency sine wave input.
Direct feed

Parafeed

What surprised me is that even with these transformers, there is a very measurable change, especially in the higher order even harmonics. The direct case is with some well-balanced tubes, so not much DC present at all. The other benefit of parafeed is you can size the cap to prevent core saturation at the maximum signal level you expect to put through the transformers.
I expect there will be a bigger effect with the Sowters.

[Max N]

Time for a long overdue update on this project!
The output stage is unchanged

• PP 6B4G

• Auto bias from Tent bias module

• Separate choke input PSU for each output channel (just because the power tx I'm using suits that)

• Lehane VCCS heater supplies

Except I invested in some better output transformers.

The front end has changed quite a lot:

[Max N]

The front end is still PP but is not currently IT-coupled.
It has cascade CCS loads (thank you Nick, I still owe you beers for those IXTP08N100D2). Cap-coupled to the bias voltage and buffered by mosfet followers. The mosfet follower idea is as per Tubelab powerdrive, MJ Bulwer-Lytton, etc. I'm using Pete Millett's A2 driver boards.
Just the front end takes up about 4 square feet on the bench. I'm trying to shrink it down by getting some PCBs made. Originally the PCBs were going to include the followers, but Pete's boards are small and neat so I'll stick with those for the followers.
Output 1 & 2 drive the grids of the 6B4Gs