role of cap in neg f.b. loop?

[farmerjack61]

Hello there, my name is Phil and I am trying to refine my audio amp. (see the attachment). It was built mostly from the parts of dead amps and tellies that were destined for the local refuse site!

The amp is now running fine, but I can't help but think that I can get a teensie bit more grunt out of those transformers (1 x 40-0-40 200VA per side).

I have been testing the jigger for a couple of days using a 1vp2p sine wave at the input. This gives me a nice clean 37.5vp2p at the 8r loaded speaker terminals. The gain is about 40 for this set up.

Now, if i increase the voltage of the 1vp2p input sine wave steadily with the above set-up, I reach a point at which the top and bottom of the output sine wave (remember it is measured across the speakers terminals) begin to collapse the oscilloscope shows that this begins at a level of around 65vp2p whilst under its 6-8r load (ie the 8ohm subbie that is connected to the speaker terminals.

Impulsive logic tells me that I should pull the 33k from the feedback loop and replace it with a 47k. However I know there is a relationship between the feedback loop and the input bandpass filter that I do not fully understand. Also I know that there is some other jiggery pokery on the feedback circuit - that is the 100uF cap. so please could you answer in detail the following question(s)

1. What is the purpose of the 100uF cap shown in the feedback loop which is connected between the 1K and gnd? (below the 33:1k voltage divider).
2. What do I need to know about the relationship between the input bandpass filter and the (neg?) feedback loop?
3. Are you having a good day?

When replying please assume that I know nothing and you may be surprised when you find I have a pretty good basic knowledge of electronics. For instance I have designed and built lots of Test equiptment such as 100uF to 10F capacitor tester, a 300dot matrix scrolling message board and even a couple of other amplifiers and many speaker filters.

Yours sincerely, Phil Elliott

[Mooly]

1. The 100uf cap returns the gain of the whole amplifier to one or unity at DC. This is important for a couple of reasons. You don't normally want an audio amplifier to respond down to DC. More importantly is that if it were removed (shorted out) then the bias current for the two input transistors would be unbalanced unless you reduced the input resistance on the other transistor too. Even then the amp would suffer more with DC offset drift as without the cap the whole amp would have gain at DC and amplify its own offset errors.

2. Good question as the two work together. Normally you make nfb return cap large enough that it has no possible impact on the low frequency response of the amplifier as far as normal audio signals go. It's effect (the -3db roll off point) is 1/2piRC so 100uf and 1K is 1.6 hz. One reason for making the cap value large is to reduce any theoretical possiblity of increased distortion due to small AC voltage developed across the cap (ref... D Self).

The input coupling cap works in the same way and this cap is normally scaled to give the desired low frequency roll off point that would be above the roll off point of the nfb cap.

3. I always do

[MagicBox]

And a small addition which you did not ask for, but is useful anyways. That feedback cap has a little brother next to it with a value of 100nF. Pick an MKT/MKP for that one. It helps high frequencies pass through the filter effortless. The big electrolyt cap is too slow to to respond to high frequency transients. That's where the little cap comes in. Purists may want to add even an 100pF ceramic.

[farmerjack61]

Originally Posted by Mooly

3. I always do

Exxxcellllent (ref...Mr Burns - The Simpsons)

Just to make sure I am on the right track, I will recap the good information you have just given me.

• We are not interested in amplifying DC.
• The roll-off of the front end high pass should normally be (slightly?) > than that of the nfb hp filter?
• In the case of the circuit presented; the front end input hp. filter consists of the 2.2uF cap and the 27k resistor to give a -3db roll-off at about 3Hz which is a whisker more than that of the nfb hpf which has a -3db roll-off at about 2Hz?

Just one more question for this week!

I have assumed (perhaps wrongly) that the 33k & 1k nfb resistors determine the over-all gain of the whole kaboodle. I now know that this 1k resistor and the 100uF cap form a hp filter. The 1k resistor here may be dual purpose, So;

1. Does the 1k resistor in the negative feedback network combine with the 33k to give the over-all gain... or does that 33k and the other 1k at the front end input form the overall gain?
2. Mooools, have you ever hurd the song; Cows with Guns..."When over the horizon came chickens in choppers" & "no-one suspected he was packin' an Uzi"! - It sure cracks me up every time I hear it.
3. Go the bovine.

Cheers, thankyou and bye for now, Phil Elliott

[Mooly]

Thats about it... amplifying DC isn't needed. There is very little audio info below (I guess for most music CD's) about 25 to 30 hz.

There's no hard and fast rule on the actual relationship of input filter and nfb roll off. It's better to use smaller input caps rather than smaller nfb caps to achieve roll off. Large electroylitics have large tolerances for one thing, and as mentioned voltage appearing across large value caps (AC voltage as frequency goes down and the caps impedance comes into play) can cause distortion to rise. That's due to the imperfect nature of components deviating from theoretically perfect ones.

The gain of the amp... normally would be said to be Rf+Rnfb/Rnfb so thats (33K+1K)/1K giving a voltage gain of 34. We assume a mid band frequency (say 1khz) where the effect of the nfb cap is taken virtually nil.
Of course anything at the input will alter the "overall" gain figure slightly. The 1k and 27k on the input form a divider reducing the level slightly. The 470pf cap adds another variable... a low pass filter. But we would still normally refer to the amp has having a voltage gain of 34 and consider the effect of the input filter as "external" to the amp proper.

[Think of the whole amp (without filters or caps) as an opamp. The bases' of the two input transistors are the inverting and non inverting inputs.]

2hz (1.6hz) and 3hz (2.7hz) sound close enough... "percentage" wise they are a long way apart.

[djk]

http://www.diyaudio.com/forums/solid...ent-print.html

[MagicBox]

Hi djk.

That link you posted had some interesting information. I've started thinking though:

How could the powersupply pole actually matter (filter caps + driven load) when a push-pull output configuration inherently has an enormous PSRR? Any LF components introduced on the powerlines are completely negated by the negative feedback?

[djk]

I don't care to speculate, I just have observed that it does.

You mention feedback, think about what is happening when you run out of feedback (near full output).

Even well below full output, staggering the poles as I suggest changes the way an amplifier sounds (think PRAT), even though the changes in amplitude response in the audio passband are trivial.

[MagicBox]

Err? I ask a genuine question. Yes, I mention feedback, under adequate operating conditions. I am not considering when by what ever cause the feedback becomes inadequate.

Edit: Giving your response a little more thought makes me conclude that your DUT somehow doesn't have the slewrate / speed to adequately feed back at near-output levels?

Edit2: I agree completely that both the + and - input poles directly affect each other, but they both sit in the signal path - not so for the PS pole due to aforementioned PSRR. Might the fact that you noticed an influence has different causes?

Edit3: You're right - the PS pole may have an effect when all amp stages are driven from the same power supply. Sections that wouldn't have a good PSRR could be influenced. I designed and built my amps with separate output stage PSUs. All other stages have their own regulated supplies and as such any effects introduced in the output stage PSU are not reflected to any of the earlier stages.