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.
This one with probe inside the case. Than, must have been the unstable circuit.
 

Attachments

  • 20171127_003628[1].jpg
    20171127_003628[1].jpg
    692.1 KB · Views: 711
Last edited:
If you attach the ground clip of you probe to the probe tip and then look at you scope with same settings the question about your scope generating what you see will be answered.

There is only a straight line in all settings.

If it's RF interference, it will be "talking". Set the time base to audio/video frequency range and see if there is any modulation present. Zoomed in to see the carrier frequency this is often missed.

Once that is ruled out, you need to determine whether it's local or global (loop). The fixes are different.

What is audio/video frequency range in seconds of time base? 10-50nSek?
In fact, it's "talking". It happens when turn the amp on I can hear from a speaker a radio frequency like changing.

Hi guys

Looking at te schematic on the first post, the cdom caps are connected to the bases of the cascode transistors, when I think they should be connected to the top device base.

Cheers

I tried connecting cdoms to drivers in the past as well, but without any behavioural change (without a scope)
 
Last edited:
What is audio/video frequency range in seconds of time base? 10-50nSek?
In fact, it's "talking". It happens when turn the amp on I can hear from a speaker a radio frequency like changing.
)

Couple hundred us to a few milliseconds. Look for an envelope at audio frequency that looks like the waveform of somebody talking. I do RF amps in my day job too - and we're always chasing 100 MHz spurs. If it gets in on the bias lines you can clearly see the modulation envelope on a scope. And a spec-an will demodulate it :). I've had the same happen with audio amps - and sometimes nonlinearity in the input stage will detect it, and you'll hear the radio station you're picking up.

If you're just getting a change in the noise floor you may actually be oscillating. Many times an amp wil break into an evil hiss if it gets unstable. And even the oscillations themselves can be unstable - not really having a fixed amplitude or even frequency. Depends where the amp's poles move with bias and signal level. This kind of behavior is not uncommon, and very difficult to predict even with the best nonlinear models.
 
Member
Joined 2017
Paid Member
I guess the oscilation is not caused by your amplifier but by your scope or external interference like RF noise because the oscillation frequency is relatively high(more than 10MHz from your screenshot) and the amplitude is low(less than 20mV) except #34 which is probably AC hum. I recommend you to improve your measurement circumstance, If you want to find the exact reason. Accurate measurement is necessary for correct judgment irrelevant to hobby or work.

BTW, I designed my amplifier with two-layer PCB several years ago and used it for two or three years. Then I redesigned the next version with four-layer PCB which I use now. When I started my career 35 years ago, I had no chance to use four-layer PCB because it was expensive and almost no requirement to utilize its performance.

But now everything has changed. You can easily design and make multi-layer PCB whose cost is incredibly cheap if I remember my old memory. When I designed my first version, I thought analog amplifier didn't need four-layer PCB. But it was a mistake. As long as you design a power amplifier which handles more than 1A, a ground plane is a shortcut to achieve high performance because a wide ground plane is the best way to keep stable ground potential.

If you have ground plane, you don't bother the position where you connect the ground wire of a scope because it has the same potential wherever you connect. Of course, two-layer PCB works well. But it can't have stable potential which ends up less performance than four-layer. This may be controversial issue because some people who have many careers in hobby audio system dislike multi-layer PCB. I'm sure multi-layer outperform two-layer in almost every case, i.g. SNR, frequency range, oscillation problem.
 
I have a suspicion that the output stage will oscillate. You have a loop of three devices (NPN/PNP/NPN and complement) with transistors all having a high fT. Simulations show that a similar output stage will oscillate when used in a simpler configuration. But the conventional triple (NPN/NPN/NPN) has lower loop gain (<1) and is likely to be more stable.
Things you can do to reduce loop gain in the complementary triple are to add degen resistors in series with the driver and pre-driver emitters. Like Quad did in the 303.
 
AndriyOL,

I followed this schematic for a long time.

Your simulation file picture have some strange data as the component value looks to be in wrong position in relation to component number.
I tried to make my own simulation file for this circuit and it do not have a good behavior in simulation. I saw other simple version of this topology in some place here and this circuit do not simulate well too.

When I find oscillations in simulator or prototype, the first thing I do is try find what is creating the oscillation. I mean, some times you have problem in output stage and can cure it adding capacitors to slow down this stage. In other situation you have polarization error and it is easy to find and solve in simulator. Others you have improper compensation and you see it simulation frequency response or even in prototype when you have good behavior in simulation. Other situation it depends on the load impedance like a very capacitive load and in this specific case an little inductor in output can cure oscillation. All this can solve oscillation or instability problems.

This circuit (at least your version) is near to impossible to find a stable operating point. I really believe this topology is not a good choice for amplifier. I also do not know about any commercial product that uses this circuit arrangement.

Ronaldo
 
In fact, it's "talking". It happens when turn the amp on I can hear from a speaker a radio frequency like changing.

This sounds like straightforward intrusion of broadcast RF signals. It can also happen that a stage oscillates at a frequency too high to be seen on a scope, and heterodynes in a passing RF signal. I can do this repeatably with some opamps but have not seen it in a power amp. (yet)

If this is happening it is probably in the output triples. Output stages like this are very prone to parasitic oscillation and it can be very hard to stabilise them. I have avoided output triples in all my production designs so far.

Are there any stability issues, say at full output into 8 Ohm?

One other thing. Either single or double-sided PCBs are fine for power amplifiers, though obviously layout is easier with double. There is no technical need for multilayer.

The great problem in power amp PCB layout is avoiding inductive distortion, ie stopping those big half-wave currents getting into the feedback path etc.
 
Last edited:
I guess the oscilation is not caused by your amplifier but by your scope or external interference like RF noise.
I'm sure multi-layer outperform two-layer in almost every case, i.g. SNR, frequency range, oscillation problem.
I'm sure this is not a scope, because the amp works the same with turned off scope. The only possible ingress is wideband RF.
As I know, ground plane is not good for audio amps in respect of THD+N performance, it's widely deliberated in manufacturers' appnotes. The best practice here is star grounding.
I've never seen any 4-layer PCB amp layout and don't know how to route it. Could you please share a routing technic?
If you can hear radio frequencies then connect an output coil on speaker out. It will filter them out. Also be careful of LED light drivers, they generate a lot of rubbish.
Out coil if for capacitive loading, I don't have one. Also it will kill HF responce.
I don't have LED drivers in the schematic and won't use them in the future. Thanks!
I have a suspicion that the output stage will oscillate. You have a loop of three devices (NPN/PNP/NPN and complement) with transistors all having a high fT. Simulations show that a similar output stage will oscillate when used in a simpler configuration. But the conventional triple (NPN/NPN/NPN) has lower loop gain (<1) and is likely to be more stable.
Things you can do to reduce loop gain in the complementary triple are to add degen resistors in series with the driver and pre-driver emitters. Like Quad did in the 303.
I don't use NPN/PNP/NPN and complement configuration, but conventional triple NPN/NPN/NPN and complement.
What values should I add to drivers and predrivers emitters?
Your simulation file picture have some strange data as the component value looks to be in wrong position in relation to component number.
This circuit (at least your version) is near to impossible to find a stable operating point. I really believe this topology is not a good choice for amplifier. I also do not know about any commercial product that uses this circuit arrangement.
There was a mistake in silkscreen components placement before, I'm aware of it or what do you mean by this "component value looks to be in wrong position in relation to component number"?
This is not my version, LC made it, I only changed some transistors with the same or better specs.
There is a working model of such triple output configuration, designed and built by Roender in his FC100 amp.
 

Attachments

  • SSA TT HP.jpg
    SSA TT HP.jpg
    641.3 KB · Views: 692
Last edited:
This sounds like straightforward intrusion of broadcast RF signals. It can also happen that a stage oscillates at a frequency too high to be seen on a scope, and heterodynes in a passing RF signal. I can do this repeatably with some opamps but have not seen it in a power amp. (yet)

If this is happening it is probably in the output triples. Output stages like this are very prone to parasitic oscillation and it can be very hard to stabilise them. I have avoided output triples in all my production designs so far.

Are there any stability issues, say at full output into 8 Ohm?

One other thing. Either single or double-sided PCBs are fine for power amplifiers, though obviously layout is easier with double. There is no technical need for multilayer.

The great problem in power amp PCB layout is avoiding inductive distortion, ie stopping those big half-wave currents getting into the feedback path etc.
It's possible to hear broadcast RF signals when setting input bias current to 4mA. Also I think there won't be any oscillations with 2mA bias current.

There is perfectly working triples, FC100 for example.
As I remeber before when instability was much higher (without 50pF caps) TO126 transistors were relatively hot than they are now and it was required 500R current source resistors (now 1K) to get a stable 15VDC potential on zeners (TL431 on the PCB).

I haven't tried it so extensively yet, with oscillation present it may overheat and burn I guees.

Do you mean inductive coupling between closely routed traces? There is an input devices too close to feedback routing in this layout. What can you say about second edition layout idea (not finished yet of course), won't there be a layout issues?
 
Last edited:
AndriyOL
In the each part there is 2 information. One is component part number like 2SC2240 (or NPN when you choose component), the other one is component number like Q21. In you simulation file you can find 2SC2240 and NPN. The NPN should be 2SC2240 and there is missing the component number.

The schematic you show in my answer I tried to simulate without success. There are other simple version that simulate but with very low performance.

Could you please share your real schematic to simulate and try to find something really helpful?

Ronaldo
 
AndriyOL
In the each part there is 2 information. One is component part number like 2SC2240 (or NPN when you choose component), the other one is component number like Q21. In you simulation file you can find 2SC2240 and NPN. The NPN should be 2SC2240 and there is missing the component number.

The schematic you show in my answer I tried to simulate without success. There are other simple version that simulate but with very low performance.

Could you please share your real schematic to simulate and try to find something really helpful?

Ronaldo

Real schematic with measured voltages in stand by. But I didn't simulate it, I use this as a schematic only.

Maybe simulation software isn't familiar with such type of circuit)

Btw, I was advised for making the amp stable to calibrate HF compensation in such a way that the phase margin will be sufficient not to colide with OLG (positive feedback). Could you please check in simulation if phase marging is sufficient?
 

Attachments

  • SSAHP1.asc
    23.4 KB · Views: 114
Last edited:
This output stage it is rock solid. I use it in 4 different power amplifier schematics designed by me. One of them it is here on diyaudio, SARA-2016 topic. I built dozens of amplifiers and none oscillates. There are 2 amplifiers used by local dealer of the Martin Logan louspeakers, with all models and everything is perfect.

I test them with step signal and 1uF load (with L//R output circuit of course) and again, everything it's ok, just normal resonance between L and C.

This type of output stage is very performant and very stable.
 
I did some adittonal tests.
On first photo the amp has 2.92mA input bias current, on second 2.2mA. The difference in oscillation level significant.
On third photo the output bias current is raised when light bulb tester starts to shine. It's approximately from 50mA to 170mA. I think there is very high frequency oscillations.
On forth photo shows when I touched by probe tip to one of the emitter resistor of input transistors.
Could the input transistors be a source of oscillations? Before I thought the reason could be because I used smd type (metal element) output emitter resistors.
 

Attachments

  • 20171127_225732.jpg
    20171127_225732.jpg
    585.8 KB · Views: 625
  • 20171127_224827.jpg
    20171127_224827.jpg
    597.9 KB · Views: 556
  • 20171127_231006.jpg
    20171127_231006.jpg
    609.4 KB · Views: 529
  • 20171127_225852.jpg
    20171127_225852.jpg
    571.3 KB · Views: 498
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.