L0rdGwyn's DIY Audio

[Tjj226 Angel]

Let me test my mind-reading abilities - elevating the 801A filaments and direct-coupling the mu-output of the pentode CCS to the 801A grid, removing the need for the buffer stage and negative bias supply?

It's a great idea in theory, working it up in LTSpice. Going to be quirky, but if it could be made to work, there could potentially be no capacitors in the signal path...

You are pretty much on the right track. The difference is that I would either use the buffer, or I would find a way to incorporate the buffer into the mu follower such that the preamp tube remains shielded from variations in current.

But yes, DC coupling the amp and getting rid of the negative supply is the way to go. Not only does it get rid of the coupling capacitor, but it also eliminates the negative voltage rail from being in the audio path. If I can incorporate the buffer into the mu follower, I might also be able to eliminate the positive buffer rail from the audio path as well. Im thinking we could use something akin to a current mirror....but idk. There is a solution there I am not quite yet seeing.

Also, I forgot, what driver tube are you using?

 

[Tjj226 Angel]

Well I don't know if what I posted above was going to be your suggestion @Tjj226 Angel , but my idea works and I have a functional LTSpice model. Real world could be quite different, but the amplifier is now a fully direct-coupled capacitorless design with a much-simplified power supply, so thanks for getting the gears turning! There are some practical aspects that will need to be investigated, but the concept passes the LTSpice test. Will likely post my .asc on diyAudio to get feedback (no-pun-intended) from the big tube brains.

Good good.

I can pretty much guess that Euro21 and Audiowize has been helping you along on this project, and I know they will definitely have some meaningful feedback.

 

[L0rdGwyn]

You are pretty much on the right track. The difference is that I would either use the buffer, or I would find a way to incorporate the buffer into the mu follower such that the preamp tube remains shielded from variations in current.

But yes, DC coupling the amp and getting rid of the negative supply is the way to go. Not only does it get rid of the coupling capacitor, but it also eliminates the negative voltage rail from being in the audio path. If I can incorporate the buffer into the mu follower, I might also be able to eliminate the positive buffer rail from the audio path as well. Im thinking we could use something akin to a current mirror....but idk. There is a solution there I am not quite yet seeing.

Also, I forgot, what driver tube are you using?

Exactly, no bias supply and given the 0V grid bias point, the 801A B+ does not need to be insanely high, around 520V with a 200V filament elevation. I'll take a look at buffer options, I thought the same i.e. how the CCS mu-output driving the 801A grid might affect the anode current of the driver tube (EF37A), does not appear to have an ill effect and the impedance appears to be low enough to drive the grid with the NFB applied, so perhaps a separate buffer isn't even necessary, we will see...

If this can be pulled off, this design just got pretty interesting. The transparency should be pretty nuts, it was already excellent with a single coupling cap.

 

[Zachik]

Will likely post my .asc on diyAudio to get feedback (no-pun-intended) from the big tube brains.

I am curious: on which thread is that brainstorming activity taking place?
Always looking for good read... even if 5% would stick - that is a win for me!

 

[Tjj226 Angel]

Exactly, no bias supply and given the 0V grid bias point, the 801A B+ does not need to be insanely high, around 520V with a 200V filament elevation. I'll take a look at buffer options, I thought the same i.e. how the CCS mu-output driving the 801A grid might affect the anode current of the driver tube (EF37A), does not appear to have an ill effect and the impedance appears to be low enough to drive the grid with the NFB applied, so perhaps a separate buffer isn't even necessary, we will see...

If this can be pulled off, this design just got pretty interesting. The transparency should be pretty nuts, it was already excellent with a single coupling cap.

What triode curves are you using for the EF37A? I thought it was similar to a 6j7 which means its similar to a 6sj7. If I look at the 6sj7 curves, I am coming up with a plate voltage of about 100 to driver the 801a to full swing.

 

[L0rdGwyn]

@Zachik I will send you a link.

@Tjj226 Angel EF37A is run as a CCS loaded pentode, the high gain with local NFB is a necessity to get good bandwidth and output impedance out of the 801A, especially with a 5K OPT. It's a tricky tube to design for given its high internal resistance, the parasitics of a high primary impedance OPT make using it in a no NFB design pretty unappealing. You either take the limitations of the high-turns-ratio OPT or run it with NFB, I've had better results with the latter. PSE is another option, but let's be serious, I am not going to match pairs of 801A and use four filament transformers LOL.

 

[Tjj226 Angel]

@Zachik I will send you a link.

@Tjj226 Angel EF37A is run as a CCS loaded pentode, the high gain with local NFB is a necessity to get good bandwidth and output impedance out of the 801A, especially with a 5K OPT. It's a tricky tube to design for given its high internal resistance, the parasitics of a high primary impedance OPT make using it in a no NFB design pretty unappealing. You either take the limitations of the high-turns-ratio OPT or run it with NFB, I've had better results with the latter. PSE is another option, but let's be serious, I am not going to match pairs of 801A and use four filament transformers LOL.

OHhhhhh thats right. Forgot you went to a lower impedance transformer. That explains the gain requirements.

Hold on, recalculating.

Side note: For what it's worth, if you run the 801a at ~310V at 60ma, ltspice is showing a plate impedance of ~3.6K. I have already seen NFB results in the 800ohm range on my end, however the signal on the 801a grid is screwed up because of the current and impedance shifts.

I still think you are just going to have to use a buffer or at the very least give the 801a its own dedicated current source.

 

[L0rdGwyn]

I have some time off of work and it appears that I have on hand what I need to throw together a single-channel of the direct-coupled 801A A2 amplifier. Looks like I have my activities planned for the week, let's give it a whirl

 

[L0rdGwyn]

Well last night I started breaking down my 801A prototype, making adjustments to the power supplies, etc. As I was moving along and thinking through just how I was going to prototype the DC coupled version, it occurred to me that while the design has two positive effects (removal of the negative bias supply and a coupling capacitor), it also introduces a host of new complexities and design issues, to the point that I am likely to scrap the idea. It's all hunky dory in LTSpice simulations, then you think about how it will actually need to be implemented in the real world, no free lunch in this hobby

 

[Tjj226 Angel]

Well last night I started breaking down my 801A prototype, making adjustments to the power supplies, etc. As I was moving along and thinking through just how I was going to prototype the DC coupled version, it occurred to me that while the design has two positive effects (removal of the negative bias supply and a coupling capacitor), it also introduces a host of new complexities and design issues, to the point that I am likely to scrap the idea. It's all hunky dory in LTSpice simulations, then you think about how it will actually need to be implemented in the real world, no free lunch in this hobby

Can you post your LT spice simulation? I am not seeing good results from using a mu follower without some type of separate current supply for the 801a grid.

 

[L0rdGwyn]

@Tjj226 Angel the dedicated buffer is a necessity I've found, but that isn't the issue. Bias stability of the 801A will be a problem, the grid voltage will be at the mercy of component variances in the pentode driver - changing driver tubes will require an internal manual adjustment. Not only that, but bias stability of the CCS loaded pentode (and thus, if DC coupled, bias stability of the 801A) is already a problem given the very high gain and propensity to drift with small changes in cathode voltage - it's possible a DC bias servo would be needed. Bringing the 801A up to bias safely is also a problem - with the CCS loaded pentode with the soft-start Maida regulator, the pentode plate voltage swings up to ~270V transiently before coming down to bias at 200V (that was with a 350V B+, will be higher with the new higher voltage B+ rail). That means the grid of the 801A will momentarily be at +70V or more relative to the filament, which is likely to blow the plate fuse on startup. There's more! The higher B+ requirement (525V vs. 350V prior) now pushes up against the limit of 600V component tolerances, not to mention any umbilical connectors will have to be appropriately rated. While the negative bias supply is gone, now a separate filament elevation supply is required, so the part count does not go down significantly. Removing the V+/V- supplies also makes the power supply options for the buffer less favorable, which means significantly more excess voltage that will need to be dissipated. All of those problems stated and it does not include the issues elevating the filament is likely to have on the stability of the feedback loop.

So, it's likely to be scrapped, it isn't a simple implementation by any means.

 

[Paladin79]

I have thought through all circuitry in my GS50 amp and did a few cuts for the cabinet but I have too many other projects I need to finish right now. I hope to wrap up the amp in a couple weeks. I also get slowed down by my wife taking up a lot of shop space with pottery but I am having a shed built and it will be delivered in about a month.

This morn I am listening to one of my Incubus amps with a very rare Melz driver tube from 1958, Tung Sol 5998, Gungnir DAC, Focal Utopia headphones. I do not use design simulation, I design for my ears and occasionally others like the sound. It took me a long time to find headphones that opened up the sound on the Incubus. My group uses Senn HD800's with their four amps and they are very good, for my ears, the Utopias allow the highs to progress and reach their potential.

 

[Tjj226 Angel]

@Tjj226 Angel the dedicated buffer is a necessity I've found, but that isn't the issue. Bias stability of the 801A will be a problem, the grid voltage will be at the mercy of component variances in the pentode driver - changing driver tubes will require an internal manual adjustment. Not only that, but bias stability of the CCS loaded pentode (and thus, if DC coupled, bias stability of the 801A) is already a problem given the very high gain and propensity to drift with small changes in cathode voltage - it's possible a DC bias servo would be needed. Bringing the 801A up to bias safely is also a problem - with the CCS loaded pentode with the soft-start Maida regulator, the pentode plate voltage swings up to ~270V transiently before coming down to bias at 200V. That means the grid of the 801A will momentarily be at +70V relative to the filament, which is likely to blow the plate fuse on startup. There's more! The higher B+ requirement (525V vs. 350V prior) now pushes up against the limit of 600V component tolerances, not to mention any umbilical connectors will have to be appropriately rated. While the negative bias supply is gone, now a separate filament elevation supply is required, so the part count does not go down significantly. Removing the V+/V- supplies also makes the power supply options for the buffer less favorable, which means significantly more excess voltage that will need to be dissipated. All of those problems stated and it does not include the issues elevating the filament is likely to have on the stability of the feedback loop.

So, it's likely to be scrapped, it isn't a simple implementation by any means.

Yeup, I already know. Don't worry. I got you fam. Just sit tight, and wait for my signal. : P

And if I finish my schematic and you just hate it, that's fine. I will post it anyways just in case it gives you or someone else ideas later on in life.

I will however leave you with an idea to ponder.

If you take the 801a bias from -10v to 0v and you raise your bias current to 60ma, your plate impedance goes down. And not by a trivial amount either. You should really define how much NFB in db you need in order to get the 801a impedance down to 2K, 1K, and ~500ohms. 2K being the maximum for a 5K load, and 500 ohms being the minimum for a 3K load.

By knowing how much feedback in DB you need in order to achieve a particular outcome, you can better define your gain requirements. I HAVE A FUNNY FEELING THAT IF YOU DO THIS, YOU MIGHT COME UP WITH SOME NEW IDEAS!!!

NUDGE NUDGE WINK WINK

 

[L0rdGwyn]

I have thought through all circuitry in my GS50 amp and did a few cuts for the cabinet but I have too many other projects I need to finish right now. I hope to wrap up the amp in a couple weeks. I also get slowed down by my wife taking up a lot of shop space with pottery but I am having a shed built and it will be delivered in about a month.

This morn I am listening to one of my Incubus amps with a very rare Melz driver tube from 1958, Tung Sol 5998, Gungnir DAC, Focal Utopia headphones. I do not use design simulation, I design for my ears and occasionally others like the sound. It took me a long time to find headphones that opened up the sound on the Incubus. My group uses Senn HD800's with their four amps and they are very good, for my ears, the Utopias allow the highs to progress and reach their potential.

Nice, sounds like a lovely combination, look forward to seeing your LS50 / GU50 amplifier. I use a combination of simulation and tuning by ear, LTSpice is a powerful tool IMO when the design warrants it. This amplifier is quite complex, at least by my standards, so being able to do a test run in sim is very useful!

 

[L0rdGwyn]

Yeup, I already know. Don't worry. I got you fam. Just sit tight, and wait for my signal. : P

And if I finish my schematic and you just hate it, that's fine. I will post it anyways just in case it gives you or someone else ideas later on in life.

I will however leave you with an idea to ponder.

If you take the 801a bias from -10v to 0v and you raise your bias current to 60ma, your plate impedance goes down. And not by a trivial amount either. You should really define how much NFB in db you need in order to get the 801a impedance down to 2K, 1K, and ~500ohms. 2K being the maximum for a 5K load, and 500 ohms being the minimum for a 3K load.

By knowing how much feedback in DB you need in order to achieve a particular outcome, you can better define your gain requirements. I HAVE A FUNNY FEELING THAT IF YOU DO THIS, YOU MIGHT COME UP WITH SOME NEW IDEAS!!!

NUDGE NUDGE WINK WINK

Well thanks for putting in the time! You may have missed it, but I am using a 0V 60mA bias point, and yes the plate resistance does fall significantly at A2 bias points. I have trialed it using the 6.5K 50mA LL9202 (just because they are what I have on hand the moment), but the transformers coming from Sowter are 5K 60mA for that reason, that is where we are headed. The CCS loaded pentode is highly adjustable from a gain standpoint, so the feedback can be finely dialed in by adjusting the plate load or the feedback fraction via the feedback resistors. Leaving the pentode cathode unbypassed significantly reduces the gain due to cathode degeneration, but increases linearity and of course the sound is MUCH better sans electrolytic. I am only adding enough NFB for a suitable output impedance anyway. In the now broken down prototype, I have trimmers on the cathodes such that I can adjust the feedback fraction in real time and see the adjustment on scope, maximizing the degree of feedback while providing just the right amount of gain to drive the 801A to full output in A2, I've been able to get output impedances around 2ohm with a 470K resistor on the EF37A plate. Obviously adjustments will need to be made when the 5K transformers arrive, more feedback will be necessary, but that is the beauty if the CCS loaded pentode, you can basically make the gain, and thus the feedback, whatever you want! It's pretty nifty