Before it was stable with enough margin in simulation, according to Ian post: One of the Top Solid-State CFA amp design, but not in reality. However I changed some resistors in OP stage.
Did you measure it PM with Tian probe?
Your offer is to use C12, C3, C5 as in post 668? What do you mean by 'any others'?
I tried from 11K to 8.25K now for R5, R6 in front end psu. I don't know why but my dmm showed 1.3V drop on the LEDs.
What I meant was to pick the most natural pole in a circuit and make it ten times greater than any other pole in that circuit.
I have not tested the circuit in post 668 by the Tian method. Your results are too far off what the simulations suggest to be doing that when things are so unsettled.
C12 is in a position where it is in parallel with the impedance of the feedback network so it bypasses any change in phase due to the output stage.
Try 680pF there and if that doesn't do it increase the value to see where this gets you. Re C3 and C5 don't increase the value above 100 pF.
I was referring to your post 778. You wrote there was not any PM. So I'm curious how did you measure it?
This is not my result, Ian suggested the circuit had 60 degree PM.
Last edited:
You would do well to take more control over direction by doing your own Tian simulations. Don't use the simulation method in posts 773 and 775.
I also tested the circuit with a 10kHz square wave into 8R// 1-2u at 15 Volts.
The result was a ragged zig-zag pattern impressed on the square wave.
That could be removed by fitting a 0.22R buffer resistor between the output and the load but there was some loss of shape with increasing signal input.
Attachments
Well i suggested thiese values in an earlier post, then driving 1u...
So as a starting point regarding stab i suggest the same values as a starting point.
Ian's website update
Hi All,
On my updated website (see my Sign links) is a new design that may be of interest to those following this amplifiers development. It's a Class-AB+C amp. I mentioned it here Has anyone seen Rush pairs used this way? For a CFA with true differential inputs
My Class-AB+C amp is not strictly speaking a CFA topology so it isn't a substitute for what AndriyOL is after.
But the TPOIC compensation and/or a new type output stage can be adapted for use in the SSHAP of this thread. In few a days maybe.
Cheers
Hi All,
On my updated website (see my Sign links) is a new design that may be of interest to those following this amplifiers development. It's a Class-AB+C amp. I mentioned it here Has anyone seen Rush pairs used this way? For a CFA with true differential inputs
My Class-AB+C amp is not strictly speaking a CFA topology so it isn't a substitute for what AndriyOL is after.
But the TPOIC compensation and/or a new type output stage can be adapted for use in the SSHAP of this thread. In few a days maybe.
Cheers
Last edited:
OITPC for SSAHP?
Attached is an attempt at OITPC (sorry I got it reversed). I have used the same output stage as the SSAHP.
This compensation gives two pole roll-off from 10kHz. At 20kHz there is now 76dB feedback, an increase of 40dB on the '5C' sim (post 775). The distortion is reduced to 1.7ppm at 20KHz (was 12ppm in post 775 sim). The gain margin is 16dB (was 36dB) and the phase margin is 41 degrees (was 84).
An output inductor is needed for driving more than 1nF but it can be very small (200uH), but in a real amp we can't be certain about values for this.
A base stopper resistor are needed with OITPC for drivers Q18, Q19 to stop burst oscillations at high output swings. It's strange the compensation for post 775 didn't need these stopper and why OITPC needs them.
100uF capacitors have been reintroduced across R14, R15 (VAS emitters) to give an extra 12dB feedback at 20kHz.
Choosing values for OITPC. It seems C3 (for Lag-lead) needs to be kept as small as possible to get return path roll-off closer to 20kHz and just enough to smooth out peaking near 20kHz (see attached Tan plot). The Miller capacitors need to be much larger than C5 (the capacitor on the output) but not too large or peaking occurs near 20kHz in the Tian plot.
C11 (50p) across the feedback resistors reduces the ringing for a squarewave, but too much capacitance causes UHF oscillation.
With 50pF the ringing may be OK. I couldn't get it much better than this at this time. Someone else can try and improve the PM. Maybe if the gain peak in the Bode plot around -180 degrees can be smoother then there should be less ringing on square-waves? Could it be the output transistor emitter resistors need inductance added to them since it affects the 50-100MHz region where -180 degrees is happening with OITPC?
These simulations are obviously too far from reality to make any final decisions about how good OITPC is compared to the previous compensation -- so it needs to be tried in a real amp.
Cheers
Attached is an attempt at OITPC (sorry I got it reversed). I have used the same output stage as the SSAHP.
This compensation gives two pole roll-off from 10kHz. At 20kHz there is now 76dB feedback, an increase of 40dB on the '5C' sim (post 775). The distortion is reduced to 1.7ppm at 20KHz (was 12ppm in post 775 sim). The gain margin is 16dB (was 36dB) and the phase margin is 41 degrees (was 84).
An output inductor is needed for driving more than 1nF but it can be very small (200uH), but in a real amp we can't be certain about values for this.
A base stopper resistor are needed with OITPC for drivers Q18, Q19 to stop burst oscillations at high output swings. It's strange the compensation for post 775 didn't need these stopper and why OITPC needs them.
100uF capacitors have been reintroduced across R14, R15 (VAS emitters) to give an extra 12dB feedback at 20kHz.
Choosing values for OITPC. It seems C3 (for Lag-lead) needs to be kept as small as possible to get return path roll-off closer to 20kHz and just enough to smooth out peaking near 20kHz (see attached Tan plot). The Miller capacitors need to be much larger than C5 (the capacitor on the output) but not too large or peaking occurs near 20kHz in the Tian plot.
C11 (50p) across the feedback resistors reduces the ringing for a squarewave, but too much capacitance causes UHF oscillation.
With 50pF the ringing may be OK. I couldn't get it much better than this at this time. Someone else can try and improve the PM. Maybe if the gain peak in the Bode plot around -180 degrees can be smoother then there should be less ringing on square-waves? Could it be the output transistor emitter resistors need inductance added to them since it affects the 50-100MHz region where -180 degrees is happening with OITPC?
These simulations are obviously too far from reality to make any final decisions about how good OITPC is compared to the previous compensation -- so it needs to be tried in a real amp.
Cheers
Attachments
I'm using the longest 5m 2mm square speaker cables which should be 200-300pF at most. 1-2uF is overkill for home amp.
It is the speaker load that is the main concern.
The phase and return signal generated by the piston movement of the coil and mechanical influences within the box and crossover components vary with speaker design - 8R with some parallel capacitance is a simple way to examine the combination of effects.
Second thoughts. Turns out the THD without these capacitors is slightly better!
Seems the extra gain with the capacitors makes the VAS distort more but the extra 12dB gain from the capacitors does not fully offset this extra distortion. Interesting.I was recently directed to a thread discussing the Tian probe and I saw this:
It seems to me that we are wasting our time trying to get the last dB of GM and last degree of PM by simulations. Even comparing different compensation methods are flawed with our present modeling. From our simulations we can only get a basic idea of what does what and not how good different compensation methods actually are in practice.
Andriyol, I think you are best to get your existing board going using standard compensation like mjona and R Dijk etc advise. Then, if you still have motivation, OITPC appears to offer up to a factor of ten lower THD at 20kHz but in practice there are likely issues hidden from us using our inadequate simulations so it may take a lot of effort to succeed with OITPC. But it would be nice if OITPC works well first try!
You miss the point, it is not total cable capasitance as cable also has resistive and inductice part. This sims where done without output induct or with its paralell resistor.
To Ian Hegglund :
Thanks werry interesting approaches to the compensation issue. I didnt find them werry good in some aspects. Thou i tried sereval "simmulated attack" to improowe i didnt sucseed. even Kind of light mix of your approach and first pole that have shown simmulated stable i didnt succed in my approaches...
Have you had a look at the marantz HDAM input module approach to the cfb ?
Second thoughts. Turns out the THD without these capacitors is slightly better!
I was recently directed to a thread discussing the Tian probe and I saw this:
It seems to me that we are wasting our time trying to get the last dB of GM and last degree of PM by simulations. Even comparing different compensation methods are flawed with our present modeling. From our simulations we can only get a basic idea of what does what and not how good different compensation methods actually are in practice.
Andriyol, I think you are best to get your existing board going using standard compensation like mjona and R Dijk etc advise. Then, if you still have motivation, OITPC appears to offer up to a factor of ten lower THD at 20kHz but in practice there are likely issues hidden from us using our inadequate simulations so it may take a lot of effort to succeed with OITPC. But it would be nice if OITPC works well first try!
I simulated and built both type CFA and VFA with OITPC. The built amps had very small changes compared to the simulated versions and were very stable.
OITPC gives very generous PM and GM, and by that very possible good result when built.
Is it the same if I put in series 3-4uF capacitor for tweeter?
I tried with values from post 775. Them amp from the start have little instability which fades in a minute. Scope pics added.
Also without C10, C11 the amp isn't stable. It must be with those 1uF caps.
I noticed the main contributors of higher THD are R39, R36, output capacitance C21 and the value of R22, R23 where higher value produce better THD, but little less stability margin. Also without C19 THD is worse, sim says, at least 100uF must be present.
For testing I'm using a bare sp driver with short cable, thus no any output capacitance should be expected. However I noticed the sound became a bit distorted or lower fidelity, I suspect because of R36, R39. DC offset almost hasn't changed.
Attachments
Last edited:
I don't mind to try it out, but I've counted 10 new components, except 100uF. Are they all OITPC compenation net? Frequency responce will be limited compared to current schematic?
As I read, in CFA amps it's not allowed to use a capacitor accross feedback resistors which can cause severe osc unlike in VFA amps.
Last edited:
The answer is no however I mentioned the need for a buffer resistor in series with the load if a 1uF capacitor is used for testing. The value was 0.22R.
You still need a zobel network at the output. How does the amplifier perform with this but without the 1uF cap in parallel with your test speaker.
Than if cable capacitance and in series capasitor aren't a problem, than only parallel capacitor in crossover net can cause such a load on the amp. I doubt spk coil with internal resistance can have any capacitance, at least dmm shows nothing.
I haven't tried to load the amp with 1uF capacitor. Do you think the amp is stable now to test with such a load or should I increase vas compensation over 44p?
I haven't tried to load the amp with 1uF capacitor. Do you think the amp is stable now to test with such a load or should I increase vas compensation over 44p?
Hi AndriyOL,
I have tried to optimize the OITPC. No 100uF. No cap across the feedback R's. Instead, added 330pF across R14,R15 to improve PM, now 47 degrees. Version '5D' attached.
THD 1.5ppm at 20kHz (vs '5C' 12ppm). BW with OITPC 1.5MHz about same as '5C'.
OITPC uses 6 C's plus 3 R (2 for driver bases). Sim '5C' uses 4 C's plus 2 R's.
The output inductor is optional. Not required when driving capacitance's below 5nF -- assuming sims are realistic. Notice I have added 50nH in series with emitter Rs for more realism around where -180 degrees happens.
Cheers
Is this correct PM GM measurement procedure?
About Zobel at the output, someone wrote here the amp should be stable in the sim without Zobel. I have it on PCB, btw.
If a 1uF capacitor is used for testing with the speaker, the proper cable itself can have 0.22R series resistance
About Zobel at the output, someone wrote here the amp should be stable in the sim without Zobel. I have it on PCB, btw.
If a 1uF capacitor is used for testing with the speaker, the proper cable itself can have 0.22R series resistance
Attachments
Last edited:
It is not clear if you are powering the amplifier with a direct connection to the mains. Whatever the answer is don't use a 1uF capacitor as a load.
The values of compensation capacitors in simulations don't allow for stray capacitance in the layout and it comes down to experimenting with values increasing these in steps and monitoring the results as you go.
You could try 100pF as your next step.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.