Practical PSE thinking.

[Salas]

I have in mind to put together a PSE. I want to use tubes that most of us have and aren't DHT and expensive. I don't care for heavy super filtered filament supplies etc right now. I don't want to invest in exotic iron also. I don't feel like taking the effort to shoot for the last word in transparency and fluidity. I want to use some moderate NFB (5-9dB) so to get the output Z and the distortion down to really useful, and stabilize the gain for correct stereo. I want to have at least 10W RMS and 20W Peak unclipped. 90dB multiway speakers and 93dB FRs are what I see as existing 'sensitive' stuff around in houses, and they are the minority too. I want it to be a happy build with decent performance and no difficult voltages and parts, just that. So I thought of a frame circuit. It looks like fitting the bill on first glance. What do you think? Is it a good frame to start with? Will that good looking and affordable James 6123HS hold its primary inductance and play any bass with 100mA DC on its shoulders in real life?

[Bas Horneman]

Her is another very practical proposition. Proven design.

http://www.one-electron.com/Randall/Randl_p1.pdf

[Salas]

Yes, the Randall is ok. I just want to have more fun around somewhat different and evolving. You know, the creative bug. Yet not really complex. Hence this general frame I posted and asking if it looks workable enough to other DIY people's eyes.

[Bas Horneman]

Hence this general frame I posted and asking if it looks workable enough to other DIY people's eyes.

It definately looks workable.

[Tubes4e4]

Hi Salas,

for my taste, way too many stages and expensive tubes, also solid state stuff within Here is my entry for that theme:

Circuit:



10kHz squares, zNFB, Y-scale 1V/DIV, X-scale 10µs/DIV:



10kHz squares, 6dB NFB, same scale:



zNFB freq response:



6dB gNFB freq response:



Exact winding instructions by Yves Monmagnon for a suitable transformer are available, or use one of your choice that will do equally well at about 110 mA iddle current down and up to the bandwidth you want

Regards,

Tom

[Salas]

Originally posted by Bas Horneman

It definately looks workable.

It probably won't be a bugger in practice. Only that James I am a bit suspicious about, since the maker publishes it as 90mA and some dealers as 120. But I can relax the DC taxing by increasing the cathode resistors a bit, say to 1.2K each for 86mA total. I don't like to use the KT88 to the colder side of things but since I will experiment with some NFB, it looks like I can keep it in check.

[Salas]

Originally posted by Tubes4e4
Hi Salas,

for my taste, way too many stages and expensive tubes, also solid state stuff within Here is my entry for that theme:

Hi Tom. Very nice amp! NFB is good because it is really needed, your square and FR tests show. But what is the subjective sonic signature when in and out? Your answer will help me to determine the Nfb amount to target for. This defines correct early gain structure decisions.

KT88s, 6072's, 12B4's, 6CG7s or 6SN7s & 6SL7s I already have, that is why I am designing around them. That 12B4 is really strong as a driver and It can lose the DC only CCS, because I will compare with RC if I start making it. The CCS has given a good account for Brian Beck though in the popular 12B4A diy pre. Best practical performance version will show anyway on test. The high gain input 2 stage concept you follow is very attractive, but I am a bit partial to really strong drivers, and since 2 KT88s will be capacitive enough, I crave for 10s of mA drive capacity. That's why the 3 stages. I want it to have persuasive bass not only suave mids, and the KT88s really pushed by an almost buffer like driver, take my fancy as fitting the bill. I may be totally wrong though, and get a poor musical performance with solid state like boom & tizz! But you know, if I get the OPTs and punch some chassis, I can always convert the circuit to 2 stage, and put in use some nice NOS EF86s I have stashed away.

[Tubes4e4]

Hi Salas,

NFB is good because it is really needed, your square and FR tests show.

I don´t think so, about 40kHz at -3dB using zero NFB is okay, I think. Damping factor using zero NFB is already better than many, many pentode PP amps with tons of FB ...

But what is the subjective sonic signature when in and out? Your answer will help me to determine the Nfb amount to target for. This defines correct early gain structure decisions.

I mainly used the 6dB of gNFB to lower the driver distortion a bit (but choosing a decent op point for the pentode driver is of much more influence, of course), but in the end there is not much of a difference to be heared, it is a measuring thing.

If you use gNFB, just keep the amount away from the 12dB mark, because around this value you will get an unfavourable harmonics distribution. You _will_ hear this, especially with human voices. And this effect can be measured, too, of course, see article from P. v. Willenswaard, in "25 Jaar AES Nederland"

For triode(d) SE and PP, if you want to apply some gNFB, just keep it at 6dB most, not higher.

but I am a bit partial to really strong drivers, and since 2 KT88s will be capacitive enough, I crave for 10s of mA drive capacity. That's why the 3 stages.

EF80/6BX6 in pentode mode at about 10mA plate current - as used in my circuit - has no problem whatsoever to drive about 300pF of CMiller at its output side to 40 kHz bandwidth and more. If you are not convinced, just do the math or just measure it on the breadboard

I want it to have persuasive bass not only suave mids,

So in the first place you want a very good OPT, up to the task.

and the KT88s really pushed by an almost buffer like driver, take my fancy as fitting the bill.

I once used two trioded KT88 PSE once, driven by a trioded 6V6 driver in a three stage design. It gave about 10 Watts of Po. It sounded okay, but I felt the result was not really worth all the effort.

Trioded KT88 is neither really efficient, nor particularly linear. Instead of two KT88 trioded in PSE, I would use a triple trioded EL34 PSE into a Ra=2k5 load at Va-k about 400V, which would give better damping factor, better linearity, better efficiency and more Po.

Oh, and by the way, in a PSE design using automatic bias, always give each power tube its own cathode resistor, coupling cap and grid leak resistor, or you are really asking for trouble if even one tube runs out of line...

If you insist on a common cathode resistor for some reason, always give each each power tube its own fuse in the cathode leg with a current rating rather close to the nominal iddle current. Or, even better, use active current sensing and protection.

Regards, Tom

[Salas]

Thanks Tom, for a very complete answer.

As I have put it down, it gets 6dB feedback on sim, maybe less in practice. I have heard the sound stage collapsing over 6dB before, on pre amps mainly. You are totally right. I see you don't get that bad an FR without NFB, but I like what it does to your SQ wave, which means better impulse response, that I value a lot. Anyway, an NFB on/off switch I surely plan.

In calculation 2XKT88s @ 3.5K show me better THD than 2XEL34s. 3XEL34s appear more linear but demand a 5kg 120ma Trans. If James, model 6133. Hmmm purchase & shipping costs start to become considerable. I think, if getting the 6123 which has a 2.5k tap also, I can install 2EL34 and listen VS 2XKT88...

I will use individual Rk & Ck. Thanks for the expert caution. Only reason I put em together is the cost of more high quality Elcos. But since there may be a run away, cost is well allocated.

I will surely test a 2 stage version with a well biased EF86 driver to compare.

Regards, you have been very informative.

[Tubes4e4]

Hi Salas,

I will surely test a 2 stage version with a well biased EF86 driver to compare.

Forget abou that, since EF86 is - by far - not up to that task. Compared to a real driver pentode, it is a current trickler at 1 mA or something and will not be able to drive the CMiller of even only a single trioded EL34. to the required 70-80 Vss without slewing (and distortion).

Whimsy EF86 is only able to drive other pentodes in pentode mode, with accordingly very low (a few pF) CMiller. Trioded EL34 or KT88 will show about 80-100pF of CMiller - per piece to be driven, of course.

Tom