One of the Top Solid-State CFA amp design

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I started this thread because I would like all best knowledge, experience and practice on a matter to be gathered in one place: the most common reasons of oscillation in different types of solid state audio amplifiers, as well as the best ways to defeat it.

All are welcomed to share their succesfull experience of oscillation solutions.

Technicaly advanced members are highly expected to participate.

As for the current oscillation problem I'm interested in is a following CFP audio amplifier project.

Current layout without silkscreen errors:
SSA.JPG - Google Drive

Current Schematic:
SSA-HP-BAL.jpg - Google Drive
 

Attachments

  • HP.JPG
    HP.JPG
    767.7 KB · Views: 4,698
  • SSAHP.jpg
    SSAHP.jpg
    411.6 KB · Views: 4,742
  • 20171122_220129.jpg
    20171122_220129.jpg
    738.6 KB · Views: 4,572
Last edited:
I assume we are talking about power amplifiers.

Improper understanding of Niquist and Bode dissertations, mr. Middlebrook and Tian articles.
;)

Study of feedback theory is actually of very little use. Much of the data you need is unknown, and I am thinking here of Class-B output stages. Their frequency response varies with the operating conditions, which change wildly over a cycle. Very little has been published on this important issue, but see Chapter 3 of my book for discussion of the implications:

The Audio Power Amplifier Design Handbook

This is why amplifiers are often seen to be unstable for only part of the output sinewave. Assuming dominant-pole compensation the design process is not much more sophisticated than increasing the compensation until the oscillation stops, then adding a bit more as a safety margin. Lower load impedances (4 Ohm, 2 Ohm) are usually more difficult to stabilise probably because the higher device currents increase local transconductances.

On another level altogether are local oscillations, typically in the output stage, which to a first approximation are not affected by the feedback loop compensation. A simple example is the tendency of an emitter-follower to act as a Colpitts oscillator under the right conditions. This sort of thing is not amenable to calculation nor simulation.
Proper simulating with good gain tuning and sophisticated PCB routing.
I am not sure what gain tuning is supposed to be. I can say that I have never known PCB routing have anything to do with HF stability.
 
"A simple example is the tendency of an emitter-follower to act as a Colpitts oscillator under the right conditions"

I think this needs to be looked at first rather than assuming compensation issues.

Increasing compensation may solves this problem but at the expense of high frequency distortion performance.

Try to identify local oscillation first. Check for RF ingress as well. Source of RF include CFL. Even LED lights replacing fluorescent tubes is suspect. Cable modems, ADSL, WIFI and now Moca networks etc.
 
Maybe ferrite beads in the bases of the TTC5200 and TTA1943 may cure the oscillation tendency ?

Not in my experience. Have never known ferrite beads help anything.

Why not to rise up the emitter resistors of the TTC and TTA from 0.1 ohm to 0.33 ohm ?

On the other hand, why? Do you think that will increase stability? If this was an emitter-follower output stage that Re change would greatly increase the crossover distortion. That may not be true for this triple output stage, but the point needs checking carefully.
 
"A simple example is the tendency of an emitter-follower to act as a Colpitts oscillator under the right conditions"
.

My experience of oscillation is the output changes quicker than the feedback signal can reach the LTP. This causes over shoot and oscillation.
I usually tame it with a CB capacitor at the VAS.

I have also had trouble with emitter follower drivers and these needed a base resistor to slow down the signal.
 
Reasons Of Oscillations In Audio Amplifiers And Best Ways Of Elimination

I assume we are talking about power amplifiers.


Hi, Douglas!
First of all i appreciate your comment and your glory-shine book!
Of course, we’re talking about audio power amplifiers.


Study of feedback theory is actually of very little use. Much of the data you need is unknown, and I am thinking here of Class-B output stages.


Oh, no, why class-B if we could provide some heatsinking headroom?

Assuming dominant-pole compensation the design process is not much more sophisticated than increasing the compensation until the oscillation stops, then adding a bit more as a safety margin.
...

I am not sure what gain tuning is supposed to be.


He-he...
Dominant pole correction is out of interest.
Best OPS for ~100-300 Wt output is good-cooked EF-triple. It’s best THD is near 0,2-0,5 % range.
So wishing no more 1 ppm distortion we need not less than ~70 dB feedback at 20 kHz.
Keeping in mind fT of the 3281/1302 we can’t reach that depth with one-pole correction.

So you’re welcome to the wild west of two-three-four pole compensation.
Gain tuning in my mind is precise math around six-order equation with real-complex parts. Being carefully solved they allow us to tune feedback loop for audiophile-liked aperiodic responce.
For example:
IMG_6534.JPG

On another level altogether are local oscillations, typically in the output stage, which to a first approximation are not affected by the feedback loop compensation. A simple example is the tendency of an emitter-follower to act as a Colpitts oscillator under the right conditions. This sort of thing is not amenable to calculation nor simulation.


Yes, common EF-triple is unstable.
There are some different ways to stabilize it, we could discuss them here or i can create corresponding theme.
First decision - is to increase base-stoppers at third devices up to 10-15 Ohms. This is very bad case, usually tends to drop in Class-B or even Class-C at higher loads.
Second decision is to have clearly capacitive source impedance at Triple input. I.e. no VAS load resistors. Again bad. After first follower we have ~active character of impedance, after second we have inductive, after third - inductive^2. So instability even at truly-active load.
Third decision - create capacitive source character at second stage input, so picking inductive output from third stage.
This is best practice and must be used widely.



Could you please be more specific about "sophisticated PCB routing" solution?



I can say that I have never known PCB routing have anything to do with HF stability.


Oh, no, Douglas, you break my heart!
What about this king of absolutely improper layout:
IMG_5638.JPG
Green and magenta shows load current path at positive and negative halves.
IMG_5611.JPG

This OPS must be routed like this:
IMG_5639.JPG

Even improper placing of decoupling caps could drop us in the instability, especially with high-GBW devices.
 
Last edited:
I haven't tried that. Maybe higher value of base limiter resistors?

Mr. Douglas, Davada would you please advise, what could a sophisticated solution in my case?

I have Wi-Fi adapter 1m close


For RF ingress. Attach a short wire to your scope probe. Crank the sensitivity of the scope amplifier up 1mV/div or better. Turn you amp off and see if you see any oscillations. If not then it's likely not an issue.

RF can trigger small oscillatory condition in an amplifier as well. Still it's amp at fault.

With the scope at the same setting turn the amp back on and move the prob about the circuit but not touching anything. As you move closer to the source of a local oscillation it may appear larger on the scope. This doesn't always work but you can give a try. Then try to figure out what the cause is.
 
You are right, I have oscillation shown even when amp is off. This is without any wire, just the probe on the output of the amplifier.
This one with input signal connected.
 

Attachments

  • 20171122_220109.jpg
    20171122_220109.jpg
    703.6 KB · Views: 1,551
  • 20171120_012541[1].jpg
    20171120_012541[1].jpg
    646.7 KB · Views: 1,509
Last edited:
I'm sorry, I don't understand what do you mean. 50pF isn't appropriate compensation value?
These are the probe with the piece of wire attached. The highest sensitivity of my scope only 5mV. Seems like my ambient is very polluted
 

Attachments

  • 20171126_213458.jpg
    20171126_213458.jpg
    749 KB · Views: 1,204
  • 20171126_213429.jpg
    20171126_213429.jpg
    467.9 KB · Views: 1,158
  • 20171126_213514.jpg
    20171126_213514.jpg
    832.6 KB · Views: 1,138
  • 20171126_213909.jpg
    20171126_213909.jpg
    533 KB · Views: 150
  • 20171126_213947.jpg
    20171126_213947.jpg
    741.2 KB · Views: 157
Last edited:
Right. So what your seeing at the output of your amplifier might not be the amplifier. You have to isolate from what you see on the scope.

Is your amplifier in a case or just trapped to a heat sink sitting on your bench.

Try to find the source of the RF and shut it down.

I know of one DIYer who spent two days trying to find an oscillation in his amplifier only to find it was ingress. His amp is actually well behaved.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.