Has anybody found that the speaker output of most class d amps shows a blurred line, caused by hf remnants getting past the amp's filters. This I found was the main source of listenting fatigue. I was able to reduce this and create a more relaxing sound by more filration after the output, like a low pass filter with a high operating point. This effectively leaves the amp working into high impedance at high frequencies. Switching my scope to see the carrier waveform I found pure sinewave in my prototype Tripath and my little T-Amp but a ragged sawtooth on my current project using IRAUDAMP modules. (zoe tsang variety). Can post filtration like this upset the amp's operation?
You will always see the carrier residual waveform on the output waveform. What is your switching frequency? Any second-order low-pass filter you add to reduce the carrier level by an appreciable amount will have the -3dB point at a frequency low enough to raise concerns about it affecting the frequency response of the amp with non-resistive loads.
No necesarilly. I used 4 pole Butterworth filter, and in a 4" cone tweeter didn´t listen any rare.
Some residual HF allays be there, infinite attenuation doesn´t exist.
Some residual HF allays be there, infinite attenuation doesn´t exist.
Last edited:
Higher order filter means more components and higher series resistance so bass will be sacrificed quality wise...
I,m taking about a two coil and two cap filter. Which is the difference: tenths of milliohms? They don´t care. I can´t see how a .01ohm coil may be worst bass notes, and not middle and treble ones... In the worst case, there must affect more treble because of increased frequency and more eddy and hysteresis currents in core.
I'm not shure that PWM residual actually influenses the sound, I'll try my idea of thinking with you guys here 🙂.
If you feed the speaker with a unfiltered PWM signal, the speakers still produces sound, and to my hearing not different to with a filter. I think that the PWM signal actually controls the "cones" the same way as a pure sine. If you think of the speaker as a motor, instead of control the speed and direction of the motor with DC (alternating direction and voltage), you direct it with pulses (PWM). As long as the switching freq is fast enough, no distorsion would occur in the speaker (the cone can't move fast enough to convert the switching freq to sound)...
What I don't have ideas about, is how the signal is handled by the speakers passiv filters, but it seems like they behave as if their where fed with a sine signal. (I've tested just to see if I could here any difference with or without a filter, and to my ears it's, if any different, more "clean"). I think that the actually only benefit of the filter is to minimize RF, and any "residue" of the PWM isn't harmfull (this will still only be handled by the speaker as the sine signal it's representing).
Any thoughts?
If you feed the speaker with a unfiltered PWM signal, the speakers still produces sound, and to my hearing not different to with a filter. I think that the PWM signal actually controls the "cones" the same way as a pure sine. If you think of the speaker as a motor, instead of control the speed and direction of the motor with DC (alternating direction and voltage), you direct it with pulses (PWM). As long as the switching freq is fast enough, no distorsion would occur in the speaker (the cone can't move fast enough to convert the switching freq to sound)...
What I don't have ideas about, is how the signal is handled by the speakers passiv filters, but it seems like they behave as if their where fed with a sine signal. (I've tested just to see if I could here any difference with or without a filter, and to my ears it's, if any different, more "clean"). I think that the actually only benefit of the filter is to minimize RF, and any "residue" of the PWM isn't harmfull (this will still only be handled by the speaker as the sine signal it's representing).
Any thoughts?
switching frequency reaching speakers -
Thanks everybody for a superb response. I have establshed that there is no operational reason why additional deep filtering should not be added but we seem to be split on the sound quality benefits. I say it improves and the reason for this was on building my first class T amp using a prototypr board I quickly became dissatisfied with the sound. I had a conversation with the supplier about what could be 'wrong'. I described the sound as if some background noise was present but was being drowned out. I then dragged my old scope from the attic and found the hf carrier at the speakers. It was indeed background noise that could not be heard. I used valves for a while then by applying filters to my class T I made it come clear. The difference in opinion may be the type of listening. Mine is classical but for pop and movies class D amps may fare better.
My current project involves 6 channels and active filtration (3 way stereo). The idea was to eliminate the high power crossover, so I now find I need an ..er , crossover.
what irony!
Thanks everybody for a superb response. I have establshed that there is no operational reason why additional deep filtering should not be added but we seem to be split on the sound quality benefits. I say it improves and the reason for this was on building my first class T amp using a prototypr board I quickly became dissatisfied with the sound. I had a conversation with the supplier about what could be 'wrong'. I described the sound as if some background noise was present but was being drowned out. I then dragged my old scope from the attic and found the hf carrier at the speakers. It was indeed background noise that could not be heard. I used valves for a while then by applying filters to my class T I made it come clear. The difference in opinion may be the type of listening. Mine is classical but for pop and movies class D amps may fare better.
My current project involves 6 channels and active filtration (3 way stereo). The idea was to eliminate the high power crossover, so I now find I need an ..er , crossover.
what irony!
I didn't measure with a microphone. I probably wouldn't pick anything up at 400kHz anyway. I believe that the problem is not at 400kHz but at lower frequencies. If you superimpose 40kHx over the 400kHz switching tone you will get nice harmonic undertones as the two frequecies 'knock' together. If you do the same with two tones not mathematically so related, say 43kHz then you will get a rough unharmonic sound with odd harmonics. I think this is why digital amps seem to suit films and grungy pop but not natural instruments with purer tones.
Jus for info, I'm including photos of my scope screen (vintage Heathkit 1960 still going strong). top trace (no 1)
Jus for info, I'm including photos of my scope screen (vintage Heathkit 1960 still going strong). top trace (no 1)
An externally hosted image should be here but it was not working when we last tested it.
shows the carrier by itself. trace (no 2) shows the carrier waveform ( same as 1 but expanded scale, showing distorted sawtooth waveform, finally trace (no 3) shows an audio signal playing comfortably load over the carrier. Here the carrier is 'louder' than the audio signal.This happens just in case of a nonlinear link in the system.
If you superimpose the two tones in a linear system, the fourier analysis will show nothing but the two tones.
But speakers definitely are sufficiently non linear to act as demodulators and can make f1-f2 and multiples audible. On the other hand.... 400kHz minus anything in the audible band should still be far above the audible band.
I am not sure that your blurring effect can be explained with this simple mechanism or if there is a more complex non linear effect.
HF tone interference test: SHOCK RESULT
I've just carried out an experiment on thid and can't understand the result.
I used my old cool edit software to simulate the effect of a continuous hf tome
TEST 1 - 11khz tone mixed with a second sweeping tone starting and ending 100 hz either side. RESULT: a rough gobbly sound as the two frequencies crossed.
TEST 2 - 11kHz(+-100) sweep mixed with 22kHz (inaudible to me)'carrier' RESULT:- No audible effect on the 11kHz tone despite the wave map clearly showing measurable interference.
TEST 3 - two inaudible frequencies beating together (22kHz) RESULT: NO SOUND despite the wave map (appended 10 second sample screenshot) showing all sorts at audible frquencies I could hear absolutely nothing.
so;
As long as the carrier frequency is inaudible it has no effect whatsoever on audible tone. I can,t believe what I'm saying!
I've just carried out an experiment on thid and can't understand the result.
I used my old cool edit software to simulate the effect of a continuous hf tome
TEST 1 - 11khz tone mixed with a second sweeping tone starting and ending 100 hz either side. RESULT: a rough gobbly sound as the two frequencies crossed.
TEST 2 - 11kHz(+-100) sweep mixed with 22kHz (inaudible to me)'carrier' RESULT:- No audible effect on the 11kHz tone despite the wave map clearly showing measurable interference.
TEST 3 - two inaudible frequencies beating together (22kHz) RESULT: NO SOUND despite the wave map (appended 10 second sample screenshot) showing all sorts at audible frquencies I could hear absolutely nothing.
so;
As long as the carrier frequency is inaudible it has no effect whatsoever on audible tone. I can,t believe what I'm saying!
Attachments
This is the first time you've realized that something you could measure was inaudible? 🙄
(with a stern defensive glare on my face....) .... but the undulations shown on the wave map ARE at audible frequencies !
Also played with these effects recently.
If you are not already aware of my old thread, you may find further infos there.
http://www.diyaudio.com/forums/class-d/201738-clock-not-clock.html
If you are not already aware of my old thread, you may find further infos there.
http://www.diyaudio.com/forums/class-d/201738-clock-not-clock.html
- Status
- Not open for further replies.