Hi all,
I do have a problem with undesirable oscillation noise in output.
The amp is a subwoofer amp for a 18" setup.
It is two Class D power amps connected in bridge.
Two separate oscillators, that is.
Now, even if the preamp is filtered LPF around 100Hz, these ClassD twins have different oscillation frequency!
One is like 102 kHz and the other 98kHz.
That means I get the difference signal in output, i.e. 4kHz.
I have tried my best to figure out what is wrong with thi design, and I should want a synchronizing connections between the two amps...
But is there another solution? Could it be that the Q value of those large LPF's at output has deteriorated? So too much of the carrier is present?
Any help in this matter would be very appreciated!
I would be very grateful for some comments from people with insight in this area.
I do have a problem with undesirable oscillation noise in output.
The amp is a subwoofer amp for a 18" setup.
It is two Class D power amps connected in bridge.
Two separate oscillators, that is.
Now, even if the preamp is filtered LPF around 100Hz, these ClassD twins have different oscillation frequency!
One is like 102 kHz and the other 98kHz.
That means I get the difference signal in output, i.e. 4kHz.
I have tried my best to figure out what is wrong with thi design, and I should want a synchronizing connections between the two amps...
But is there another solution? Could it be that the Q value of those large LPF's at output has deteriorated? So too much of the carrier is present?
Any help in this matter would be very appreciated!
I would be very grateful for some comments from people with insight in this area.
Hi!
The freq. difference alone is not a problem, unless something (inside the amp) demodulates it. Demodulation is possible if two switching signal mixed in the PWM modulator (and other conditions are presented). The "alien" switching signal can come via:
- feedback path (post-filter FB is more sensitive then pre-filter FB)
- power supply (high capacity, LowESR on-board caps, and independent power supply wiring can help)
- ground loop
- capacitive or inductive coupling (proper signal levels and impedances, proper layout).
If no signal comes from other channel, then no interference.
High loop gain, good, balanced comparator, and symmetrical switching times have also good influence.
If you need a specific advice you must tell what have you built.
The freq. difference alone is not a problem, unless something (inside the amp) demodulates it. Demodulation is possible if two switching signal mixed in the PWM modulator (and other conditions are presented). The "alien" switching signal can come via:
- feedback path (post-filter FB is more sensitive then pre-filter FB)
- power supply (high capacity, LowESR on-board caps, and independent power supply wiring can help)
- ground loop
- capacitive or inductive coupling (proper signal levels and impedances, proper layout).
If no signal comes from other channel, then no interference.
High loop gain, good, balanced comparator, and symmetrical switching times have also good influence.
If you need a specific advice you must tell what have you built.
Thanks for your input!
Well to me its rather obvious that it is the differnece frequency.
Problem is that the carriers before th LPF are also different in amplitude, not much, but some.
A working amp (that I compare with) has a bit better Q-factor in the output LPF and are therefore more silent.
I think the problem is that the LPF corner frequency are the same, but the frequencys that are to be cancelled are 1. a bit differnet in frequency (4kHz) and amplitude of carrier is a little different. I guesss the 4,7u polyprops capacitors are too bad...
Or... the frequncys must be adjusted to be the same at silence.
Well to me its rather obvious that it is the differnece frequency.
Problem is that the carriers before th LPF are also different in amplitude, not much, but some.
A working amp (that I compare with) has a bit better Q-factor in the output LPF and are therefore more silent.
I think the problem is that the LPF corner frequency are the same, but the frequencys that are to be cancelled are 1. a bit differnet in frequency (4kHz) and amplitude of carrier is a little different. I guesss the 4,7u polyprops capacitors are too bad...
Or... the frequncys must be adjusted to be the same at silence.
Zoors!
You are searching the problem in a wrong place, but I can't tell for sure where it is. This schematic is almost unusable, and exact method of connection of two sides is very important too.
The problem is not from 1 place. It is a product of a complex process, therefore it can be eliminated by more then one method.
Different amplitude is just a side-effect, it's not a reason, nor a consequence.
I saw a bridged ClassD amp wich had an awful (seen at oscilloscope), very strongly amplitude modulated output because of slightly different switching freq, despite of this there were absolutely no audible interference.
You are searching the problem in a wrong place, but I can't tell for sure where it is. This schematic is almost unusable, and exact method of connection of two sides is very important too.
The problem is not from 1 place. It is a product of a complex process, therefore it can be eliminated by more then one method.
Different amplitude is just a side-effect, it's not a reason, nor a consequence.
I saw a bridged ClassD amp wich had an awful (seen at oscilloscope), very strongly amplitude modulated output because of slightly different switching freq, despite of this there were absolutely no audible interference.
Sometimes the whistling noise produced by the preamp. There is no general reason for interference. It depends on actual realisation.
Of course! I didn't tell other! But what you hear is not directly the difference! If you have only sum of 2 signals, 98 kHz and 102 kHz, they are still inaudible! The only way to make them audible is connecting both of them to a nonlinear network, wich makes many other frequency, for example the difference freq. This is called demodulation. If you can disable demodulation in some way (eliminate nonlinearity, or eliminate some of the input signals), you won't hear interference.
Of course! I didn't tell other! But what you hear is not directly the difference! If you have only sum of 2 signals, 98 kHz and 102 kHz, they are still inaudible! The only way to make them audible is connecting both of them to a nonlinear network, wich makes many other frequency, for example the difference freq. This is called demodulation. If you can disable demodulation in some way (eliminate nonlinearity, or eliminate some of the input signals), you won't hear interference.
Well my theory is this:
The output LPF (20mh in series, 4.7uF to ground) is not working on one side as well as on the other.
Remember this is a bridge setup, therefor we have one 18" loudspeaker driven by tw identical poweramps with two of these, suppose to be identical, LP filters.
The LP filters are leaking HF into the speaker coil, where they collide and a ringmodulation effect occurs. Resutl frequency is not the sum, but the difference of the input HF power PWM scuarwaves when no audio are input (silence). The power squarewave frequencys syuppose to be 100kHz each here, but they are not. They have a 4 kHz difference, and that signal is very clearly heard when amp is silent.
Solution =
1. better Q value of LPF filter. Matching of caps, hi quality hi stabilty caps etc. With a resonance frquency that matches the carrier power HF as exact as possible.
2. better synchronicity between the two bridged amps feedback loops (something I learned here, thanks!). This could be done by have some sort of link from one amp output (master) to the others feedvback loop (slave)
(#1) will minimise the ringmodulator effect probability, and (#2) will make the amp totally silent in this cas, as the difference of two absolutely identical input frequencys will be Zero. (silence)
This is my theory and solution proposal, and I would love to have it challanged!
The output LPF (20mh in series, 4.7uF to ground) is not working on one side as well as on the other.
Remember this is a bridge setup, therefor we have one 18" loudspeaker driven by tw identical poweramps with two of these, suppose to be identical, LP filters.
The LP filters are leaking HF into the speaker coil, where they collide and a ringmodulation effect occurs. Resutl frequency is not the sum, but the difference of the input HF power PWM scuarwaves when no audio are input (silence). The power squarewave frequencys syuppose to be 100kHz each here, but they are not. They have a 4 kHz difference, and that signal is very clearly heard when amp is silent.
Solution =
1. better Q value of LPF filter. Matching of caps, hi quality hi stabilty caps etc. With a resonance frquency that matches the carrier power HF as exact as possible.
2. better synchronicity between the two bridged amps feedback loops (something I learned here, thanks!). This could be done by have some sort of link from one amp output (master) to the others feedvback loop (slave)
(#1) will minimise the ringmodulator effect probability, and (#2) will make the amp totally silent in this cas, as the difference of two absolutely identical input frequencys will be Zero. (silence)
This is my theory and solution proposal, and I would love to have it challanged!
??? Millihenry? It would be extraordinary huge! If you think "mh" means microhenry, then: 20 uH is a little small for this low fsw.
Do you know anything about ring modulators? Do you see any ring modulator on the output? If yes, then cut it out!
Do you know what is Q of an LC filter? I don't know what do you think.
The resonance freq. must not match to carrier. It have to be much lower. LC filter must have only small influence on switching freq, especially in an amp with pre-filter feedback.
This can be a solution indeed, but you asked another solution in your first post.
And synchronisation can turn interference into distortion if you don't do it properly.
Eva said:
Eva, I agree with you.
Zoors, you need:
- the same carrier frequency for the two amps
OR
- a difference between carrier frequencies greater than 20Khz: for example, 85 KHz and 115 KHz will give you the beating frequencies at : 115+85=200kHz (who cares?
) and 115-85=30Khz (out of audio band)Another "raw" solution: a post- lowpass filter (on the speaker) at 200 Hz will give a lot of attenuation at 4Khz...
Pafi:
"Do you know anything about ring modulators?"
Yes I do indeed.
However, I did not say there was a ringmodulator.
I did say there was a ringmodulator "effect", intension was to describe it so knowhow people would understand what I was rambelin about.
A ringmodulator is a devise that produces a sum and a difference signal from two input signals.
In this case the sum of 102kHz and 96 kHz is out of question.
You wont hear it.
But the difference 102-98 is significative.
I would agree on the idea that a diff > 20KHz would make it, BUT... it will make the whole setup unbalanced.
Ive been thinking of a post LPF, but I feel it must be unnecessary, ther must be an simpler solution.
I will try to find out what component network will make the final working frequency for the carrier, and try to adjust that.
Btw... The Q factor of a LC resnonance frequency is just the very resistance when the two components agree on letting the frequency pass =) In a LC LPF it will be about pass it to ground.
And, yes youre right, Pafi, it is 20uH the inductor is.
I have been dealing with old american schematics too long, where a mmfd is a pF. And a mF is a uF. =)
In this case the 100V 100kHz pwm square is feed thru a 20uH coil, a capacitor 4,7uF to ground, then on to the speaker coil, and the samt to the other end of the speaker coil, but with a second self contained pwm oscillator system.
Only shared system except the speaker, is the PSU.
And in idle theres no pumping what so ever.
Thanks for your thoughts about it, folks!
"Do you know anything about ring modulators?"
Yes I do indeed.
However, I did not say there was a ringmodulator.
I did say there was a ringmodulator "effect", intension was to describe it so knowhow people would understand what I was rambelin about.
A ringmodulator is a devise that produces a sum and a difference signal from two input signals.
In this case the sum of 102kHz and 96 kHz is out of question.
You wont hear it.
But the difference 102-98 is significative.
I would agree on the idea that a diff > 20KHz would make it, BUT... it will make the whole setup unbalanced.
Ive been thinking of a post LPF, but I feel it must be unnecessary, ther must be an simpler solution.
I will try to find out what component network will make the final working frequency for the carrier, and try to adjust that.
Btw... The Q factor of a LC resnonance frequency is just the very resistance when the two components agree on letting the frequency pass =) In a LC LPF it will be about pass it to ground.
And, yes youre right, Pafi, it is 20uH the inductor is.
I have been dealing with old american schematics too long, where a mmfd is a pF. And a mF is a uF. =)
In this case the 100V 100kHz pwm square is feed thru a 20uH coil, a capacitor 4,7uF to ground, then on to the speaker coil, and the samt to the other end of the speaker coil, but with a second self contained pwm oscillator system.
Only shared system except the speaker, is the PSU.
And in idle theres no pumping what so ever.
Thanks for your thoughts about it, folks!
Ring modulator is a special kind of modulator, it contains 4 diodes in a "ring" topology. For some odd reason some people call "ring modulator" every amplitude modulator, however they don't have any ring. But at the output there is no any kind of modulator either! A modulator needs strong nonlinearity, but filter and speaker are almost perfectly linear at this level and frequency. They surely won't demodulate carriers.
Nonlinear elements are in the amplifiers, so demodulation can be done there. This is not just a guess, it is proven by experience. (However it's strange for me that somebody doubt about it.)
Sorry, I don't know what is "LC resnonance frequency", "very resistance" and "agree on letting the frequency pass". Maybe it's because of I'm not a native english speaker.
I know only resistance, impedance, reactance, etc...
Yes, intentionally it is. But unintentionally everything is connected to every other thing. Sometimes this coupling is too strong, and makes funny thing. But let this alone! Let's assume really the PSU is the only "shared system"! Does it mean that absolutely no coupling between amps? Not of course! PSU have impedance, wires have impadance, and speaker is a coupling too!
Of course. So...?
The PWM comparators of each amplifier are the real demodulating element, and the leakage magnetic flux from output inductors is usually the offending signal that is being picked up in the PWM modulator of the opposite channel.
Input signal wiring may be also picking up the carrier of the other channel and resulting in intermodulation products if the signal is not properly filtered just before the modulator. Remember ground loops too.
If there is a problem with output filters, it's just the coil producing a too strong leakage flux.
Input signal wiring may be also picking up the carrier of the other channel and resulting in intermodulation products if the signal is not properly filtered just before the modulator. Remember ground loops too.
If there is a problem with output filters, it's just the coil producing a too strong leakage flux.
"Ring modulators combine (or heterodyne) two waveforms, and output the sum and difference of the frequencies present in each waveform. "
Pafi - I dont really understand why you put down so much energy to prove something that is not at all relevant for this subject. It wont help the solution to to the problem at all.
LC Resonance
In the Class D amp output to speaker is a L (inductor coil) in series with the speaker, and in between them a large polypropylene capacitor (in this case 4.7uF) to ground.
In a balanced setup this LC pair is doubled, one on each pole of the speaker. As you understand there is some calculation to get this corner frequency to match, A. the feedback frquency of the carrier, B. the rolloff frequency in order to cancel the carrier at speaker coil.
Wjat I can see on my outputs, is that those carriers are of different frequencym 4kHZ diff, and, they are also on differént amplitudes. Menaing the LPFs (LC circuit) are not damping the carier with the same efficiency. Therefore you can say that they have differnet quality factors. Q value of a LC circuit is tha actual resistance at resonance. (ideally should be Q=zero)
Pafi - I dont really understand why you put down so much energy to prove something that is not at all relevant for this subject. It wont help the solution to to the problem at all.
LC Resonance
In the Class D amp output to speaker is a L (inductor coil) in series with the speaker, and in between them a large polypropylene capacitor (in this case 4.7uF) to ground.
In a balanced setup this LC pair is doubled, one on each pole of the speaker. As you understand there is some calculation to get this corner frequency to match, A. the feedback frquency of the carrier, B. the rolloff frequency in order to cancel the carrier at speaker coil.
Wjat I can see on my outputs, is that those carriers are of different frequencym 4kHZ diff, and, they are also on differént amplitudes. Menaing the LPFs (LC circuit) are not damping the carier with the same efficiency. Therefore you can say that they have differnet quality factors. Q value of a LC circuit is tha actual resistance at resonance. (ideally should be Q=zero)
megajocke said:
Yes... one of the totally scrapped burnouts I have in a pile =)
You can see the large polyprop caps in dark orange between the relays and the coils. The smaller coild up north is actually transformers, the output current sense.
An externally hosted image should be here but it was not working when we last tested it.
Zoors said:
I don't know why do you talk about irrelevant things! All I wanted to convince you about is this (ring modulator) is irrelevant (because no ring modulator there). So thank you, finally we agree!
I've already told you in my first post what can make interference. I can't do more while you don't want to talk about any relevant thing.
If understanding don't help...
No. Quality factor can be measured or calculated at resonance freq, while switching freq is much higher. At that freq Q is irrelevant.
Absolutely not. You've completely misunderstood. (Q is infinite in an ideal resonating circuit. But in a ClassD amp resonance is not a desired thing, ideal Q is 0.7...1. )
1.: There is serial and parallel LC circuit. Both of them has only 2 poles. The filter has 4 poles, so it is actually not an LC circuit. It's a network, it shows a parallel resonance for the output, but serial for the power stage.
2.: In a basic series L(R)C circuit the resistance is in series with the other elements. In a parallel L(R)C R is parallel.
3.: Q=R/(omega*L)=R*omega*C for parallel resonance, and Q=omega*L/R=1/(R*omega*C) for series resonance. At resonance freq both circuit have an impedance equal to R (not Q!), but for the filter there are 2 completely different R values depending on what do you analyze (impedance from output or impedance from power stage).
But as I said, the whole thing is irrelevant, because you don't have to do anything with resonance.
For building a good ClassD amp, having the basic knowledge of electrotechnics is not neccessary, but advisable IMHO. But you won't be able to learn it from forums and Wiki.
"Ringmodulator effect"
Pafi, Please stop confusing things by making them more complex than what they are. If I say "ring modulator effect" I mean a description of something that reminds us of a ringmodulator, mnamely the substraction of two frequencys, in my case 104000 Hz and 98000 Hz which gives me a measured, de facto, proven and recorded alien TONE of 4 kHZ in the output.
What does a differnece tone of 4000 Hz in output reminds YOU of, if you deal with two carriers that "collide" wher one is 104000 HZ and the other is 98000 Hz? Please give me a very short answer. What sort of effect does it remind you of?
You can probably propose 100's of different effects that you gert reminded of EXCEPT for the ringmodulator effect, but the sad thing is that I choosed just the ringmodulator effect, and I just did not know that you where allergic to ringmodulatoers, Pafi.
In the case of the subject howevever - I was right. It DOES remind us of a ring modulator effect. In deed.
Im sorry to say this, but if I change one of the carriers to 96000 Hz, the alien tone goes to 6oooHz... Just as a ringmodulator would do...
"Just like a ringmodulator effect".... Can it be more clear?
Please do not develop this thread on this sidestep.
Oh yes, you are so right, I am very flattered by having You as my main advisor. I am really impressed by your deep knowledge in this area.
Q factor will affect the damping steepness. It will affect the carrier amplitude (damping, db/octave) if it is damped by for example a tuned LC network.
I guess that is too simple for your highness understand.
Instead of helping me solving the problem, you just annoy me with all sorts of elitistic behaviour. What is your problem really?
Do you have to take everything apart and claim everybody around you is ignorant peasants and you are the light of the universer, or what? You just bugs up my time with a lot of complex irrelvant ramblings of totally simple facts. Please find another thread to infect.
Pafi, Please stop confusing things by making them more complex than what they are. If I say "ring modulator effect" I mean a description of something that reminds us of a ringmodulator, mnamely the substraction of two frequencys, in my case 104000 Hz and 98000 Hz which gives me a measured, de facto, proven and recorded alien TONE of 4 kHZ in the output.
What does a differnece tone of 4000 Hz in output reminds YOU of, if you deal with two carriers that "collide" wher one is 104000 HZ and the other is 98000 Hz? Please give me a very short answer. What sort of effect does it remind you of?
You can probably propose 100's of different effects that you gert reminded of EXCEPT for the ringmodulator effect, but the sad thing is that I choosed just the ringmodulator effect, and I just did not know that you where allergic to ringmodulatoers, Pafi.
In the case of the subject howevever - I was right. It DOES remind us of a ring modulator effect. In deed.
Im sorry to say this, but if I change one of the carriers to 96000 Hz, the alien tone goes to 6oooHz... Just as a ringmodulator would do...
"Just like a ringmodulator effect".... Can it be more clear?
Please do not develop this thread on this sidestep.
´
Oh yes, you are so right, I am very flattered by having You as my main advisor. I am really impressed by your deep knowledge in this area.
Q factor will affect the damping steepness. It will affect the carrier amplitude (damping, db/octave) if it is damped by for example a tuned LC network.
I guess that is too simple for your highness understand.
Instead of helping me solving the problem, you just annoy me with all sorts of elitistic behaviour. What is your problem really?
Do you have to take everything apart and claim everybody around you is ignorant peasants and you are the light of the universer, or what? You just bugs up my time with a lot of complex irrelvant ramblings of totally simple facts. Please find another thread to infect.
If you add or subtract two different tones you do not get a difference freqency component. However, if a nonlinearity is introduced then there will be a difference freqency. It can be seen as the amplitude envelope getting detected like in an AM-receiver. But without nonlinearity there will be no difference freqency component.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.