Feedback fun with K5881

[NealG]

Following on from this thread, I thought I would try AG’s cct on my own 2a3 PP and also K5881MKII to see what effect it had. AG’s cct is clever and designed to work around LF instability by adding more feed back at LF. The amp actually becomes more stable due to LF phase correction, bass is improved as a result. I have to say I agree with this, on both amps I've tried Bass is noticeably improved, as is the midrange for some reason. Both amps seem more tuneful and ‘in control’. K5881 still sounds a little closed in to my ears as do most valve amps with high feedback but the mod has managed to bring out more controlled Bass and midrange detail.

This following should be applicable to any amp which implements loop feedback in the same manner as the K5881, which is based on the Mullard 5-20 cct.

Here’s what you need to do using K5881 MK2 as a reference.

Remove R3, Remove R4, Remove C2, Remove R18

Put R3 in the place of R18.

Put a 1000uf 10v or 16v quality ‘lytic cap in place of R4.

Add a 11R resistor in place of R3 (two 22R in parallel)

Move the feedback wire from the junction of R3/R4 to the junction of R3 and Cathode of V1




For K5881 MK1 the procedure is the same except that R3 becomes 33R not 11R as with K5881 MKII.

--------------

Editted in after following posts in this thread,

<< Also it will be necessary to adjust C10 for best Sqr Wave performance. Change from 220pF to 1.2 ~ 1.5nF. This is best performed wile under test with a 'scope, however, if this cannot be done use a 1.5nF polystyrene cap. For K5881 MK1 use 1.5nF ~ 2nF, this has not been tested. >>

--------------

If you’re thinking of using this in another similar cct work out the PD ratio formed by the existing f/b resistor and the shunt resistor (that is R4 divided by R4 plus R18). When you move the BIAS resistor (R3) to the f/b resistor position you need to recalculate the shunt to get the same ratio as before. If you don’t your f/b level will be different. I aimed for a 10Hz rolloff with the cap value. It can be tweaked a little in value to change Bass response.

[Richard]

Nice one Neal! It looks right except I'd have expected C10 to need to be around 2nF. Did you check it on SW?

Interesting comments about sound quality. I find my 5-20 has noticably better bass than Kit88 - more natural and detailed - probably the best of any amp I've had in fact but I can't say if it's due to the feedback arrangement or not.

I was playing around with bypass cap values at the time Acorn posted which was why I happened to be "tuned in " to all this. You're right that changing the bass affects the midrange - probably the lower harmonics - I found lowering the bass roll off point gave a more natural sound to the mids - less "hifi" - warmth without muddle.

After your previous post about stability I did drag out my LF gen and try to provoke the beast with subsonics but it wasn't playing. It gives 10W at 10Hz and going down to 1Hz upsets it none either. Below this would be heard as motorboating but nothing again. I guess it might be an issue with less stable amps.

Rich

[Richard]

Graphics of the feedback R and compensation C impedance,

Original R 20K with C 220pF,



Changing R to 2K2 and leaving C at 220pF,



Changing R to 2k2 and changing C to 2nF,



The shunt leg of the divider should be linear, or thereabouts at higher f anyway, so I'd expect the feedback leg to need to be a similar curve to before - but - I've not tested this theory

Rich

[Richard]

Should say that this new C value can be worked out as the same ratio as the new R value but plotting the impedance seems to confirm what's happening doesn't it.

Rich

[NealG]

Yes, thanks for the graphs Rich.

I did Sqr Wave test the 2a3 pp and it was fine, with and without f/b, I'll check K5881 when I get a moment.

The LF instability is somthing heard as a 'bloom' or bloat' to the bass so I'm told, not something you're going to see it on the 'scope and not something to do with damping factor. K5881 seems to be rock solid on the 'scope yet it's apparent after listening that the Bass has improved, well, with my 'speakers, room and ears Your Matamps are really punchy and well designed, the PSU is a little different I think, maybe they don't need this sort of 'band-aid'

Next up will be to try a CCS on the splitter, already did this on the 2a3, like to hear how it would sound on the 5881.

ATB

[Richard]

Hi Neal

Perhaps 2a3PP was already rolling off it's top end, it sounded gentle when we heard it but I don't know which version that was?

I'd be surprised if you don't find large overshoot on SW with K5881 using 220pF. That may be audible as a sharpening sound but it would be interesting to hear what you make of listening again after correction.

You're right, 5-20 is known for it's stability and good sound. To be truthful I've come a very long way round to end up back where Mullard left off 50 years ago. Something about Designers and Emperors here I think.

What did you use for CCS on 2A3PP?

Rich

[NealG]

Mornin' Rich!

Ah, 2a3 has moved on since then I tink you heard v0.5! Must be up to...oh, at least v4.0 by now!

For CCS I looked at Gary Pimms simple dual DMOSFET design based on the DN2540. I can't seem to get these in the UK in small quantities so I'm looking for alternatives that can take at least 150v. If you can find a minimum 150v delpletion mode N channel mosfet then we're in business!

[Ianm2]

that's a great post, thanks Neal, I will certainly try that, easy to do, interesting, Richard, after all my playing around with 300bs and alll sorts of esoterica, I find I am returning to el34 types and mullard circuits and wad amps

[acorn]

Hi all

If you changed the 1000uf for a 470 uf would it mean that you would have a rool off at 20 Cs instead of 10 Cs for the 1000 Mfd I have tried this and the Bass seems cleaner, what do you lads think, because I have read if you let to much low frequency response through to your bass speakers, that it will only upset them has they cant handle frequencies much below 30 C/s without getting into real trouble.

Regards Acorn

[NealG]

Yes, you can do that and it will have the effect you describe. AG notes this in his article on the Kel80 mono blocks.