Simple Symetrical Amplifier

[Esperado]

my star ground plan is slightly different

Please, take your output ground at the star point.
Are all resistances non inductive ?
If you do not have an oscillo to see what happen exactly:
You can try to add an input filter at 200khz, can only have benefits.
If it does not solve the problem, try to add 20pf in parallel with the feedback resistances.

 

[guitar89]

Please, take your output ground at the star point.
Are all resistances non inductive ?
If you do not have an oscillo to see what happen exactly:
You can try to add an input filter at 200khz, can only have benefits.
If it does not solve the problem, try to add 20pf in parallel with the feedback resistances.

So where should my output ground takes from ? according to my layout ?
What do you mean by inductive resistance ?

I don't quite like input filter since they tends to degrade the input (more parts integrated to modify the input) , but I will give it a try. So it is -3db at 200khz ?

Ohhh yeah, never though of that idea. Anyway, does adding that cap parallel to feedback resistor have any drawbacks ? Will try it tmrw !

 

[Lazy Cat]

Ohhh yeah, never though of that idea. Anyway, does adding that cap parallel to feedback resistor have any drawbacks ? Will try it tmrw !

I would not try this, since CFB resistors paralleled with cap usually cause immediate oscillations. More you'll gain with correct Cdom. 😉

 

[guitar89]

I would not try this, since CFB resistors paralleled with cap usually cause immediate oscillations. More you'll gain with correct Cdom. 😉

Ok, get it. So VFB won't be affected by this ? Still couldn't understand why CFB resistor would trigger oscillation, maybe its why I couldn't precisely understand how CFB works😉

 

[Esperado]

I would not try this, since CFB resistors paralleled with cap usually cause immediate oscillations. More you'll gain with correct Cdom. 😉

On my amp, at those values (10/20pf), it just dump the little overshoot on square waves. Don't you had some pf on you own boards, that you adjust with a variable cap ?

guitar89: inductive resistances are wired-wound power ones (to be avoided).
Your output grounds has to be taken on each IC board, where the power ground goes-in.

 

[Esperado]

I couldn't precisely understand how CFB works

VFB returns the output signal to an inverting device (the differential -input) to be compared with the original signal.
CFB return the output signal to the same input device used for signal. The signal is amplified in an common emitter cofiguration, while the feedback signal is entered in a common base configuration. So, the same input device is used to compare both.
As common base is low impedance, that's why it is called 'Current" feedback.

 

[Lazy Cat]

Don't you had some pf on you own boards, that you adjust with a variable cap ?

Always Cdom only, never have had any cap on current feedback resistors. Can try it by VSSA at squares loaded with capacitor, to compensate ringing.

 

[homemodder]

I would not try this, since CFB resistors paralleled with cap usually cause immediate oscillations. More you'll gain with correct Cdom. 😉

Cdom is the worst compensation one can use on a CFB circuit, severly limits bandwith and the vas linearity (also causes imbalanced currents to flow through the two vas transistors and limits PSRR ) its much better using shunt compensation in combination with a little compensation accross the feedback resistor.

 

[MiiB]

how to implement shunt compensation..???

 

[Lazy Cat]

Cdom is the worst compensation one can use on a CFB circuit, severly limits bandwith and the vas linearity (also causes imbalanced currents to flow through the two vas transistors and limits PSRR ) its much better using shunt compensation in combination with a little compensation accross the feedback resistor.

It is unconditionaly stable with Cdom=12 pF only, squares response is perfect. If you know more, you have to support your claimes with explanations. Criticism is easy in any form. 🙄

 

[Esperado]

Cdom is the worst compensation one can use on a CFB circuit, severly limits bandwith and the vas linearity (also causes imbalanced currents to flow through the two vas transistors and limits PSRR ) its much better using shunt compensation in combination with a little compensation accross the feedback resistor.

I must agree with this with big nuances on the 'severly'..
In fact, i use little // C with feedback to linearize the bandwidth (often some ringing at HF limit). Then adjust low pass filter in the input, to smooth the square wave.
Working with output Zobel (as well as output inductance protection with very capacitive loudspeakers cables) can save too some oscillation problems.

LC, Cdom limit the slew rate, better to avoid-it as much as possible.
It is just about compensating parasitic capacitances in the feedback loop where their occurs: in the input transistor junction and printed board around.
Increasing this // C, you will see first a flatten of the response curve, then ringing appears again, but at a lower frequency, more and more, until the amp oscillate: We are playing with phase at ~10Mhz, here, not hf voltages.

 

[guitar89]

On my amp, at those values (10/20pf), it just dump the little overshoot on square waves. Don't you had some pf on you own boards, that you adjust with a variable cap ?

guitar89: inductive resistances are wired-wound power ones (to be avoided).
Your output grounds has to be taken on each IC board, where the power ground goes-in.

I don't have any variable cap, but I will try them out.
Aww.... I'm not sure does all my resistor are non inductive, most of them are carbon resistor, but my output emitter resistor are 😱 How lethal they are ?😱

According to your ground explanation, I don't quite get it, it seems kinda fit into connection I used. (My output ground is taken from the point where 2 secondary ground meets, which is my star ground. It then branch of to OP, IPS)

 

[Lazy Cat]

Esperado, please, I am not beginer at this. I know what Cdom causes, that's why I use trimmer caps on this position to trim to exact conditions and than re-trimm with squares on capacitive load. What makes you feel that nice responses on squares are coming from, mid air?

Look, any amp can be stabilized in many ways, but please 12 pF as Cdom looks like too much for you, or degrades slew-rate too much? C'mon, homemodder is known provocator, giving nothing concrete in return, but very good at criticism.

 

[Esperado]

Esperado, please, I am not beginer at this.

It seems you are in some bad mood, those last times, not accepting or considering any other's inputs (witch do not reduce your and was not criticism). don't you read: "with big nuances on the 'severly'" witch was just provocative, indeed, and reason too for my "I must admit' ?
You are the one witch wrote:"I would not try this, since CFB resistors paralleled with cap usually cause immediate oscillations". Witch is not true in all CFB amps and not friendly while i just wrote the contrary.
Well, i will not bore you any more on this thread and your others, my two cents were in order to help and experienced, with a friendly intention.
I hope-you the best for your future designs.

but my output emitter resistor are How lethal they are ?

Using wire-wound resistances at the source of my power FETs (not the same amp than your) brings oscillations:
I was obliged to use paralleled metal film resistances instead, to reach the same power rating without the inductance problem. Problem ended. Too, and it is just a personal opinion, film resistors brings a more 'fluent' sound, i find carbon resistances to sound a little granular. And they are noisy.
I take care too of a big margin in the power and quality of the feedback resistances, as the resistance distortion is here the most important and increase with temperature.

 

[Lazy Cat]

its much better using shunt compensation in combination with a little compensation accross the feedback resistor.

Did that many times, first attempt like 20 years ago, much worse effect in practice than a little linearization of VAS Cbc.

It works strictly locally, as local freq comp feedback on VAS. No other compensation needed whatsoever, only low pass input filter. 😉

 

[shaan]

AFAIK, Cdom has an influence on the HF performance of typical VFB amps where fast charge/discharge of this cap is difficult because of inherent current limitation at HF. That's why it is always suggested that this cap be of high quality. I use 100pF 2KV ceramic ones in my P3A and other vfb amps(same as input RF suppressor). But only 5pF in my SSA, as a minimal bandlimiter. To be frank, I don't know if it's doing its job, but no ill behavior so far.

Now, many literature on CFB say that placing a cap parallel to the feedback resistor for bandlimiting is a no-no. I guess this is because it will decrease the AC impedance of the parallel RC network to a very low value, much lower than the safe minimum for the amp(a characteristic of nearly all CFB amps). So, for superfast amps like the SSA, I am afraid to install any cap at the global feedback node. Cdom works at the VAS, providing local feedback, in a much safer way with no "minimum safe" limitation; speed control done in an earlier stage of the signal chain. Seems more preferable than parallel RC global feedback.

 

[Lazy Cat]

It seems you are in some bad mood, those last times, not accepting or considering any other's inputs (witch do not reduce your and was not criticism). don't you read: "with big nuances on the 'severly'" witch was just provocative, indeed, and reason too for my "I must admit' ?
You are the one witch wrote:"I would not try this, since CFB resistors paralleled with cap usually cause immediate oscillations". Witch is not true in all CFB amps and not friendly while i just wrote the contrary.
Well, i will not bore you any more on this thread and your others, my two cents were in order to help and experienced, with a friendly intention.
I hope-you the best for your future designs.

Dear Esperado, lately we are in some small misunderstandings without any particulary strong reason. It is just the situation with a lot of questions from your side, since at the moment I am really not in a position to test and answer all your requests. On the other hand I dislike the situation when open questions remains unanswered from my side, so at the end it looks I ignore you, but that is far from truth, simply I don't have enough spare time to make it happen. So please Esperado, you would be most helpful not to question so much but rather test and present the results of your interest. I'm not that capable to please all your requests, besides I have to run my company as priority, to remain in the business.

Regarding CFB resistors compensations and shunt compensations on both sides of VAS, I tested all of them, but these measures are not closely effective and benign as simple and tiny Cdom. I usually not publish anything rejected since I would cause more and more debate I couldn't participate and again it would look unpleasant from my side, as I am ignoring.

Please be patient with me and please try stay present here with your great knowledge, to remain helpful to DIY community. 🙂

 

[Jay]

Look, any amp can be stabilized in many ways, but please 12 pF as Cdom looks like too much for you, or degrades slew-rate too much? C'mon, homemodder is known provocator, giving nothing concrete in return, but very good at criticism.

Let's combine "numbers" with perceived sound quality.

In general Cdom has negative effect on sound quality. The question is:
(1) How big is too big? (2) Can it be done with other solution?

Different Cdom value has different effect on each circuit, but as a general rule I will not accept 47pF and higher except for not so serious design, or if the other parameters are so good and I cannot find better solution. 12pF is not an issue for me.

Cdom or leadcomp is usually not options. Technically Cdom solves issue better than leadcomp. When in simulation we can get away with XpF Cdom, in reality we usually need more than that.

Leadcomp OTOH needs to be precise (in combination with input filter). This is then difficult to implement. But if it work, I would prefer this.

 

[Esperado]

Example from a CFA i'm working on.
First image:
1 -no cap. the amp is not stable.
2 -7pF paralleled with feedback resistance of 5.9 k: Amp stable.
3- 1Mhz square wave with 1K/50pf of low pass filter.
(Yellow is the feedback )

Second image: idem, Cdom: 0.6pF (for no peak in bandwidth) low pass: 1K/55pF

 

[Jay]

0.6pF??