Fun with Feedback

[Nick]

Yep, its why you want your frequency compensation to reduce the gain to below 1 before the phase shift gets to 180 deg.

[steve s]

Thanks nick, explained perfectly !!

I had it in my mind that positive feedback could occur in the audio frequency band with the wrong cap values .. obviously I had it totally wrong..

I just feel like I've finally got my head around the subject ( in a limited way of course )
So steve, nick and James. thanks

[Cressy Snr]

Thanks nick, explained perfectly !!
...I just feel like I've finally got my head around the subject ( in a limited way of course )
So steve, nick and James. thanks

I'm still learning Steve. There's a lot to learn yet, but I've got the basics and as Nick says once you have a basic understanding of how global feedback works, you become able to use it and take design decisions with it that are more judgement than luck.

[Mike H]

I had it in my mind that positive feedback could occur in the audio frequency band with the wrong cap values .. obviously I had it totally wrong..

No I think I know what that might mean, if it relates to low frequency instability. This can occur where the LF response of the OPT is lower than what the rest of the amplifier circuit can handle (in other words, the LF 'performance' of the OPT is better than the amplifier). It's like the OPT is drifting off zero and then the amplifier is trying to pull it back into line but on the end of a piece of elastic, it takes too long so it overcompensates then it goes too far the other way and so on. One possible reason, cathode bypass capacitor values are not large enough. This is aggravated if open-loop gain is too high.

How do I know all this?

Exactly what happened with my prototype EL84 p-p amp (it's been to Owston couple of times, and Ant made the wood case for it – 2 x EF86 for both input and concertina phase splitter).

It was all designed in LTspice, I got the o/p stage working lovely, on its own – then the front-end, on its own – a rather ambitious push-pull phase-splitter affair (can't remember what the name of the configuration is). I was after lots of open-loop gain, to be tempered by NFB. Seemed like the obvious thing to do.

When I plugged the two sections together, LTspice threw its toys out its pram. Errors everywhere and painfully slloooowww – had to keep aborting the sim while it's only done a few microseconds. I'm blaming the program. But, I thought, let's simplify it to something we know works, i.e., the old fave, one input gain stage and a concertina splitter (cos, can't keep pratting about like this forever). LTspice liked that, and behaved itself. Alrighty then, we'll actually make that!

So I built it, and finally got to the point of firing it up, and one channel had very low frequency oscillation. The speaker cone was visibly pumping in and out by maybe 10mm once per second or so. This was finally cured by reducing the gain of the first EF86 stage, by merely resorting to bunging a resistor in between the cathode and (what was) the bypassed cathode resistor (values modified to retain DC bias). Was all hunky-dory after that and it was lovely.

Much later, I went back to the LTspice sim, wondering why the program won't show this behaviour. Well of course, normally in a sim I rarely go longer that a second, or a few, if it's audio frequencies we're dealing with. But I had an idea, keep increasing the run time ('trans') and see if anything happens. And lo, so it came to pass, at about 15 seconds, off it went into a low frequency oscillation, and yes, about 1 Hz or so. Program vindicated! Even with zero Volts input, same result, which is exactly what the real circuit was doing. And the 'cure' was the same. The crux of it, too much open-loop gain.

[Nick]

The crux of it, too much open-loop gain.

Sounds like classic three stage amplifier Motorboating which in many cases is caused by a hidden feedback path through the power supply rails. Normally worth trying better smoothing in the B+ line to reduce the -3dB point frequency. The R's and C's in the supply between the stages couples the signal in the output stage back to the initial amplifier stage with the RC networks adding the necessary phase shift. With typical amp stages having lowish PSRR, Bob is your proverbial.

[Cressy Snr]

As I have a centre-tapped HT transformer, I thought I may as well try a couple of cheap, 6CL3, TV damper diodes for rectification. I started off with a solid-state rectifier with one 6CL3 as the slow-start element, running off the same heater supply as the rest of the valves. It worked nicely and gave a useful increase in the HT voltage, so that the EL34s were dissipating 20W each, nicely in the sweet spot.
A few days later I thought about removing the SS diodes and having a full wave setup with two dampers, but the heater winding had insufficient capacity to run another tube. I ended up fitting an additional 6V transformer and running the rectifiers off that. It works very well and the whole thing sounds lovely:

Heretik2halfwave.jpeg (121.11 KiB) Viewed 2398 times

HeretikPSU2.png (41.34 KiB) Viewed 2398 times

The 1.2A of heater current each of the dampers draw, isn't too onerous. It would have been a lot heavier with my EY500As or 6D22S.

[Cressy Snr]

Anyway, back to the solid state amp. The Heretik was a great engineering exercise, but I’ll be keeping it for occasional use due to the power consumption. I listen for at least six hours a day, often more, so the solid state amp is a no-brainer.

[pre65]

Anyway, back to the solid state amp. The Heretik was a great engineering exercise, but I’ll be keeping it for occasional use due to the power consumption. I listen for at least six hours a day, often more, so the solid state amp is a no-brainer.

Just out of interest, what are the relevant costs for (say) eight hours listening per day over (for example) a 30 day month ?

[Cressy Snr]

Anyway, back to the solid state amp. The Heretik was a great engineering exercise, but I’ll be keeping it for occasional use due to the power consumption. I listen for at least six hours a day, often more, so the solid state amp is a no-brainer.

Just out of interest, what are the relevant costs for (say) eight hours listening per day over (for example) a 30 day month ?

Valve class A single ended.
£8.00
£24 per quarter.
£96 per year.

Solid state class B amp - £20 per year.

Every little helps.

[Cressy Snr]

OTOH
I could listen less, read more, write more and go for more walks and keep the valve amp in the system.

[pre65]

Anyway, back to the solid state amp. The Heretik was a great engineering exercise, but I’ll be keeping it for occasional use due to the power consumption. I listen for at least six hours a day, often more, so the solid state amp is a no-brainer.

Just out of interest, what are the relevant costs for (say) eight hours listening per day over (for example) a 30 day month ?

Valve class A single ended.
£8.00
£24 per quarter.
£96 per year.

When you think what a packet of fags cost these days (£12 - £15 ?) £24 for 3 months audio nirvana seems good value.

[pre65]

OR, £96 - £20 = £76

76/365 = 20.8p

So using the valve amp over SS cost an extra 21p per day.

[Cressy Snr]

Anyway, the pair of damper diodes were a tweak too far.
The extra heater transformer's electric field was inducing a slight hum into the input stage and also it was causing the bottom plate to utter a loud clang on the initial magnetisation at switch-on, because it was inside the case. The choice was either to return to a GZ34 or go back to solid-state with a damper diode as the slow-start element. I went back to the solid-state plus slow-start because I couldn't be arsed to change the Novar socket for an octal, and because I got more voltage out of the arrangement and more voltage is always a bonus, all within reason of course.
That's it as far as any more tweakery goes. I'm well chuffed with the results and it would be a shame to mess it all up, especially after all that brain pain involved in learning about then working out and testing out the feedback and frequency compensation.

Would I build an IDH single ended amp without global feedback in future? Not on your nelly!

Final schematics below:

HeretikV2.0.png (66.71 KiB) Viewed 2237 times

Heretikslowstart.png (34.11 KiB) Viewed 2237 times

[Paul Barker]

OTOH
I could listen less, read more, write more and go for more walks and keep the valve amp in the system.

Im listening much less, and still working full time for survival, and subscribe to full gym membership. So I graft hard all day then go to gym for a workout then sauna;ice room after. Im now addicted to the endorphins and the comrade-ship (couldnt make spell checker happy with ery or ary so went ship, ehh aaaw, another fail, so put a hyphen in. apple spell checker is no help whatsoever ! save a fortune in eleccy trying to write messages without spelling wrong instead of listening to valve amp!) of other guys going. Most are good company. Im becoming fitter, instead of sitting around listening to music for prolonged periods.

Aswell as that the fight for life to get into worthwhile work wasnt going well at all in this age. Got a local job but its too hard for a almost 65 year old. You cant make youre thumbs support youre fingers in placing screws in holes, not dropping them and with youre eyesight thats makes you see small screw heads as well as Marty Feldman, jobs Im sent to take me 4 times as long as before deteriaration. So it’s tough right up to the last.

I dont do any music reproduction upgrades, that includes the opamp project. Non of it matters when youre fighting to feed the family.

[Cressy Snr]

Music is my first love and it will be my last. Music of the future and music of the past. To live without my music would be impossible to do. In this world of Johnson, Sunak, Truss, rip off energy companies, climate change and complete utter bastards, my music pulls me through.