How much feedback for PP penthode?

[Svein_B]

I have rebuilt an EL84 PP amp.
The OPT has no UL taps so I am stuck with straight penthode operation.

I am wondering how much global feedback is generally considered good practice.

Before my rebuild it had a GFB of around 7dB which gave an output impedance of 5 ohm which I think is a little high.
I have increased the GFB to a little over 9dB and get 4.2 ohm output impedance. Does this sound reasonable?

I tried triode operation also. This gave me 5 ohm with no GFB, and 3 ohm with 7dB of GFB - but only around 3W power at 250V.

SveinB.

 

[ray_moth]

It's usually necessary to use about 26dB. For so much NFB, you need lots of open loop gain and the challenge is to maintain a good margin of stability. The necessary feeback ratio does depend on your speakers, though, and you may find you can get good results with less NFB.

 

[planet10]

Are the screens regulated?

It would be interesting to try the Yves/Gingertube scheme for feedback,

That said, our EL84 PP is stuck in triode and sounds so good we don't want to muck with it.

dave

 

[Miles Prower]

I have rebuilt an EL84 PP amp.
The OPT has no UL taps so I am stuck with straight penthode operation.

I am wondering how much global feedback is generally considered good practice.

There isn't any. It all depends on what pentode you're using, how well the open loop is implemented, what speeks you're using, what your preference in sound is.

I did two PP pentode designs. One of these uses 807s. For this particular design, there was 6.95db(v) of local feedback from the 807 plates to the driver, and about 4db(v) of gNFB around the entire amp, including the OPT.

Another design uses the 6BQ6GTB HD PA, also PP pentode since this type, like the 807, has a very low Vsgsg rating that makes either trioding or Ultralinear impractical. 6BQ6s are different, in that there wasn't the same sort of pentode nastiness when running open loop, so I decided to skip the local feedback. For this design, I made the gNFB adjustable, from none at all to 12db(v). For hard driving rock (Ozzy, Nirvana, Foo Fighters) and techno, 6db(v) sounds right. Softer music (Karen Carpenter, Andre Rieu) the full 12db(v) works better, but makes the harder rock sound a bit "subdued".

Too little gNFB makes for some sloppy-sounding bass due to woofer underdamping.

"It's usually necessary to use about 26dB."

No well designed VT amp should ever require anywhere near that much gNFB. That will lead to a very "solid statey" sound

An externally hosted image should be here but it was not working when we last tested it.

If you find that it's necessary, then you need to rethink your open loop design because you did something very wrong. Either that, or your speeks are no good.

Finally, if you're going full pentode mode, good screen regulation is mandatory. I use active voltage regulation to supply screen voltage.

 

[Svein_B]

Thanks for the guidance.

It seems that my increase to 9dB of global feedback is still fairly low according to general wisdom. With existing tubes this is about as much as I can do and still have enough gain. The next re-wiring will be to try a variant of the Yves/Gingertube partial feedback scheme, and different driver tubes.


Please allow me to apologize for my spelling mistake in the thread title.
The correct spelling for the tube discussed should be pentode.
. . unless of course we should happen to use a beam tetrode.

SveinB.

 

[Miles Prower]

Please allow me to apologize for my spelling mistake in the thread title.
The correct spelling for the tube discussed should be pentode.

Not really. Originally, it was spelled "penthode". The 'h' was dropped later.

 

[richwalters]

Re: Re: How much feedback for PP penthode?

There isn't any.

Too little gNFB makes for some sloppy-sounding bass due to woofer underdamping.

"It's usually necessary to use about 26dB."

.

I'd copy what Miles says. Only to add that a well sounding reflex of B4 alignment (traditional) needs no more and will sound quite loose. That's the beauty of tastes. I use 20dB global nfb in UL p-p. Anymore sounds clinical. Despite classical users often saying wanting more nfbfor exactness, I dispute this. As being a player I like to hear some cello and lower end piano resonance. On the other hand, on stage I blow trumpet and use a multiple 807 amp in true p-p and low nfb to get a rasp. Two distinct different usages both with excellent results.

Some MI tube amps often used in HiFi reinforcement often without the listeners aware of it ! and better still prefer the sound !
richj

 

[ray_moth]

No well designed VT amp should ever require anywhere near that much gNFB.

I disagree, 26dB is not a lot for pentode-mode. Many of the 'classic' UL PP amps of the fifties and sixties had that much NFB, if not more, and they shouldn't have needed so much being UL. Don't forget, we're talking about pentode mode here. You have to cope with much higher OP impedance with pentode than with UL or triode. e.g. plate impedance of EL34 is about 1k in triode, 2-3k in UL but 15k in pentode mode!

You need to regulate the screen and to apply a healthy dose of NFB, with pentodes, to give decent damping with almost any speaker and to tame the higher-order odd harmonics distortion that pentodes tend to generate. Of course, it goes withopuit saying that NFB should be as low as possible but no lower.

 

[Miles Prower]

I disagree, 26dB is not a lot for pentode-mode. Many of the 'classic' UL PP amps of the fifties and sixties had that much NFB, if not more, and they shouldn't have needed so much being UL.

Yes, and all those "classic" designs sounded just like what they were: typical "Big Box" designs.

An externally hosted image should be here but it was not working when we last tested it.

Using s-loads of gNFB to force impressive THD numbers gives the marketing department something to brag about, but what good does it do to say you have an amp with a THD of 0.00000000000000000000000000000001% if it sounds horrible? When I was doing the design for the 807 amp project, I originally had it set up for some 30db(v) of gNFB, just like conventional "wisdom" says. That lasted all of 30 seconds as it sounded even worse than the solid state amp I was using at the time. And, yeah, even worse than running open loop with no NFB connected at all. I'd take the sloppy bass and even the pentode nastiness over that any day of the week. Next, I dropped that to 12 db(v) and it wasn't much better.

Would you throw a heavy wool comforter over your speeks and listen through that? Nah, me neither, but too much gNFB is the electronic equivalent: the highs were completely gone, bass sounded like monotonic thumping, vocals faded into a bland background, and lyrics became noticeably harder to understand, and fine details were lost. Just like that SS amp that got such good reviews at the time, and which I "thought" sounded pretty good. It turned out that some 4.0db(v) of gNFB was about right for that design. Sounds just fine: bass has plenty of "authority" without sloppiness, and the pentode nasties are gone. Playing familiar material through it was quite a revelation as there was so much detail that I literally never heard before.

The current project uses PP 6BQ6s running as pentodes (since the low Vsgsg rating pretty much eliminates the trioding and UL options) with adjustable gNFB from none at all to 12db(v). The full 12db(v) is definitely tending towards a "solid statey" sound, but does work well with some material. For metal and techno, 6db(v) is better.

Don't forget, we're talking about pentode mode here. You have to cope with much higher OP impedance with pentode than with UL or triode. e.g. plate impedance of EL34 is about 1k in triode, 2-3k in UL but 15k in pentode mode!

Tell me something I don't already know.

You need to regulate the screen...

Dunnit

...and to apply a healthy dose of NFB, with pentodes, to give decent damping with almost any speaker and to tame the higher-order odd harmonics distortion that pentodes tend to generate. Of course, it goes without saying that NFB should be as low as possible but no lower.

Yes, gNFB should be as low as possible, and I still say 26db(v) is too damned much. In the final analysis, hearing is believing.

 

[ray_moth]

Hi Miles,

Yes, and all those "classic" designs sounded just like what they were: typical "Big Box" designs.

Actually, not all of those old designs sounded bad. The Cit II used a lot (30dB?) of NFB but didn't sound too bad. Maybe today's speakers call for something else, though?

Tell me something I don't already know.

This wasn't intended for your benefit especially (since it's plain from your earlier posts that you already know it) but was aimed at the Forum at large.

Yes, gNFB should be as low as possible, and I still say 26db(v) is too damned much. In the final analysis, hearing is believing.

You make a compelling case for low NFB. No one would be happier than I to achieve acceptable performance from pentodes with, say, 12dB or even less. If the speakers are up to it, with reasonably linear pentodes (or beam tetrodes, of course), then why not?

I think the thing is do design for stability at 26dB of NFB, then if it turns out that this much is unnecessary all well and good - reduce NFB to suit your needs. It's better to have it available and not need it than the other way around.

 

[Tubes4e4]

Hi Miles,,

Yes, gNFB should be as low as possible, and I still say 26db(v) is too damned much. In the final analysis, hearing is believing.

This argument only proves you never had the chance to audition a V69a, which employs well in excess of 30dB gNFB and runs the BPT finals in pure class A to boot

In general, I second your opinion, though.

Tom

 

[Miles Prower]

Hi Miles,


Actually, not all of those old designs sounded bad. The Cit II used a lot (30dB?) of NFB but didn't sound too bad. Maybe today's speakers call for something else, though?

Of course, the speeks figure into it. I can only use what I got.

I think the thing is do design for stability at 26dB of NFB, then if it turns out that this much is unnecessary all well and good - reduce NFB to suit your needs. It's better to have it available and not need it than the other way around.

I already do that. It isn't good enough just to have the amp not make like a QRP SW rig. You could also have a rising amplitude characteristic at the high frequencies that'll make for shrillness if your phase margin is borderline. It's definitely something to be avoided, and if you can get stability at high gNFB, then, in all likelyhood, you won't be seeing this at lower gNFB's. There's no reason that I couldn't pour on huge amounts of gNFB. It's that I don't so it because I don't like how it sounds.

 

[Shoog]

It seems that my increase to 9dB of global feedback is still fairly low according to general wisdom. With existing tubes this is about as much as I can do and still have enough gain. The next re-wiring will be to try a variant of the Yves/Gingertube partial feedback scheme, and different driver tubes.

This is an excellent idea. I have had great success with this approach, and have heard a Rogers Cadet III which was converted to the Gingertubes schematic - excellent !!!
The great thing with partial feedback and pentodes is that in theory at least they perform better than the same pentode with triode strapping.

This approach seems to bring all of the benefits of gNFB without most of the penalties. Care needs to be taken to make certain that the proportion of feeback doesn't swamp the current through the driver, and there can be issues with high frequency roll off if overdone.

I messed about with gNFB and found all the classic penalties that are often attributed to it were manifested. Of course I only messed about with it so i wouldn't claim to have got the best from global feedback. My current main amp is PP triode class A with zero feedback.
Have fun.

 

[pchw]

Very educational discussion. Does anyone have a link to the Yves/Gingertube scheme handy?
I have an Eico ST70 integrated amp taken apart for a rebuild. Since the OPT are pentode only, everything discussed here are applicable to it. Why bother to restore per original schematic, let's make it better per diy spirit

 

[Yvesm]

The original one

http://www.dissident-audio.com/PP_ECL86/Page.html

Yves.

 

[richwalters]

Remember in most p-p /tetrode /pentode amps using global nfb greater than 30dB, instability trouble often starts at the HF end not low LF with long LS cables. The capacitive layout of the drive circuit plays a big part and when designing a push pull amp, both phase sides should be as symmetrical as possible. Any layout should have the grid (g1) components well away from the H.V anode and screen.

There is another issue. Perhaps more controversial. The case of the Harmon K Citation2, uses inverse feedback within the driver / o/p tubes circuitry and may account to 12dB, the rest 18dB may be global nfb. I'm only quoting an example. The global nfb has the op tranny included to the input tube...a very big difference and may have more effect on sound quality than the inverse nfb. The designer of the Citation 2 might have had 20dB global nfb in mind as the compromise to best sounding with the o/p tranny in the circuit.
Of all the decades I've been on the tube amp ball, the deciding component has the quality of the wretched o/p tranny.

Question;
Does inverse feedback within the tube stage circuitry i.e as per Citation 2 really effect the sound quality ? THD okay less but ->o.m.o... I'm not convinced. Global nfb is far more influencial.
Views welcome.

richj

 

[Miles Prower]

Question;
Does inverse feedback within the tube stage circuitry i.e as per Citation 2 really effect the sound quality ? THD okay less but ->o.m.o... I'm not convinced. Global nfb is far more influencial.
Views welcome.

richj

Yes it does. Adding local NFB around the finals reduces the effective r(p) which improves woofer damping, and also serves to make the inductive reactance of the OPT primary a bigger percentage of r(p) which also helps the low end. The improved linearity helps to clear up pentode nastiness. It doesn't get the whole job done, of course, since there is still no error correction for OPT nonlinearities, and nonlinearities in the front end. It does make the job of gNFB error correction easier.

 

[richwalters]

It doesn't get the whole job done, of course, since there is still no error correction for OPT nonlinearities, and nonlinearities in the front end. It does make the job of gNFB error correction easier.

The way I see it, by using a triode or two in cascade in the front is the best solution regards linearity. THD is higher but that can be dealt with.
Having sorted this out and TRYING to get to the best configured tube amps, we are bogged down by the o/p tranny. Unfortunately what I'm finding by using tube to tube inverse feedback in p-p parallel output pairs there isn't much to be gained as the r(p) which you mentioned is already low enough. There I shall have to leave it as it is.

thanks for reply richj

 

[Wavebourn]

For my tastes, the more the better (or none!) in terms of clearness of sound when the amp is driven below clipping. Of course, more feedback means more OL gain, so more noises is a compromise. Also, stability is an issue.

I am happy with 50W / channel amps with deep feedbacks driving very sensitive line arrays; they never approach clipping.

Usually amps with 6BQ5 output are not powerful enough so used with soft clipping to compress dynamic range, in such case a negative feedback causes harshness on clipping, and especially on recovery from clipping.

Conclusion: weak amps for loud sounds should contain no global feedback at all.
Powerful amps should contain as more as you can afford (stability and noises).