I have 2 prototypes of class D audio amplifier: one is selfoscillating, other is carrier based 250khz. When i overdrive with music, at clipping it produces a different distorsion than class AB amplifier, something like cracking sound, very unpleasant.
Inductor at output is T106-2 double stacked, rail +/-90v, 2ohm-8ohm.
I have "seen" the same cracking noise at comercial Behringer PM5000 classD powered mixer.
My question: it is a normal clipping distorsion for class D amp? Something that i cannot remove without a limiter or compressor at input of module?
Inductor at output is T106-2 double stacked, rail +/-90v, 2ohm-8ohm.
I have "seen" the same cracking noise at comercial Behringer PM5000 classD powered mixer.
My question: it is a normal clipping distorsion for class D amp? Something that i cannot remove without a limiter or compressor at input of module?
why dont you use a limiter at input programed to not clip?
I saw in jbl commercial amp a limiter at input and a circuitry around integrator that i think is to reduce the gain at a certain level.In fact the good amplifiers have any sort of circuit to avoid clipping.Google lm13600 northon amplifier as a limiter.There are also some optpcoupler+ldr used in this application.
I saw in jbl commercial amp a limiter at input and a circuitry around integrator that i think is to reduce the gain at a certain level.In fact the good amplifiers have any sort of circuit to avoid clipping.Google lm13600 northon amplifier as a limiter.There are also some optpcoupler+ldr used in this application.
i had a similar problem sometime ago and after some research with scope,i saw an interference at audio input caused by emi only at clipping.the problem was in preamplifier ground picking rf from power stage only at clipping.After grounding the potentiometer of gain and changing the way i did the preamp ground ,cracking at clipping desapeared.
Did you see any strange ringing at integrator at clipping?Another thing that helped was snubber at switching node,maybe because the distance between switching node and filter was very big (20cm).
Did you see any strange ringing at integrator at clipping?Another thing that helped was snubber at switching node,maybe because the distance between switching node and filter was very big (20cm).
Have you got a 'scope so you can look at the output waveform during clipping? I'm very used to 'ucd-style' amplifier designs and they should all exhibit clean clipping. The oscillation frequency drops drastically just befor clipping, which is audible but it shouldn't give you the sound that you describe. Are you sure some other instability isn't happening at the onset of cliiping?
Ouroboros said:Have you got a 'scope so you can look at the output waveform during clipping? I'm very used to 'ucd-style' amplifier designs and they should all exhibit clean clipping. The oscillation frequency drops drastically just befor clipping, which is audible but it shouldn't give you the sound that you describe. Are you sure some other instability isn't happening at the onset of cliiping?
Where'd you read that? If you hear Fs drop you're either part bat or there's refinement to be had somewhere. It can clip without Fs dropping much at all. "Clean clipping" is considered soft clipping which does not manifest as an audible crackle.
UCD enters and leaves hard clipping cleanly, big difference, it crackles. I am aware of no class d amplifier with inherent soft clipping, this must be achieved by additional measures.
Exactly at the onset of clipping a UCD amps oscillation frequency drops way down, but the clipping is 'hard' and clean, with no 'hangover' effects as it comes out of clipping. I'm talking here about ucd designs based on a modified form of the old Philips ucd app-note, of which I've now tested getting on for a hundred units, and not the more sophisticated Hypex versions.
Due to old age and inherently poor hearing, my upper frequency limit is now around 10kHz so all I hear is a normal clipping effect, but my younger work colleagues all claim to be able to hear the effect of the drop in switching frequency (dropping into the 10 to 20kHz region) just before true clipping starts. I did not say that ucd amps have soft clipping, merely that the transition in and out of clipping is little different to that of a conventional linear amp, apart from the (slight) effect of the switching frequency drop.
Due to old age and inherently poor hearing, my upper frequency limit is now around 10kHz so all I hear is a normal clipping effect, but my younger work colleagues all claim to be able to hear the effect of the drop in switching frequency (dropping into the 10 to 20kHz region) just before true clipping starts. I did not say that ucd amps have soft clipping, merely that the transition in and out of clipping is little different to that of a conventional linear amp, apart from the (slight) effect of the switching frequency drop.
There's a philips ucd app note?
It seems our definitions of clipping differ somewhat.
In my view especially for hard clipping it's any sheared signal, and with switching amps that equates to any non modulated portions. It switches in order to modulate signals, and when it runs dry of gain, or as with "a ucd" as Bruno might say the feedback "bails out", modulation falls dead, "switching" stops, and it clips hard to either rail, for the duration of the clipped portion of the signal for which it lacked the gain to modulate. Obviously, this falls into the audioband, but I think you might agree it is not "switching" any longer, but hard clipping.
You might look at it as switching simply because for UCD the clipping recovery is robust and graceful, and the very onset of the highest frequencies would be rapid, but it is most assuredly not switching in the audio band! Lesser amps might hit the rail a single time and stick to it like a bug on flypaper.
Further, in terms of it occuring with "a UCD", the same exact thing would happen for a clocked amp that also lacked clipping countermeasures. That's just how a switching amp rolls, and yes it begins at higher frequencies first because it's those that are the first to be clipped. So it really has nothing at all to do with a non fixed frequency amp as any self oscillating is, switching into the audioband.
What does hard clipping in class d sound like?
If you have an amp which is suffering from penny pinched implementation or any other issue making it less that perfect, the clean crackles of the sheared signal portions being snapped off will first be masked by noise and high distortion, so you'll never see the first signs comming. At some point, the sound will stretch thin, grow coloured, become really loud and offensive, you know it's clipping hard already by now, but you just can't hear it crackling because it's not up to that kind of accuracy at all. Push it a little harder and it will crackle suddenly, so hard you'll ball up in the fetal position and suck your thumb.
So with some conditioning you end up hearing your warning that it's time to turn it down.
With a very clean and accurate amp it's another matter altogether. The sound will remain pure and uncoloured right up to the razor's edge. At this point speaker distortion, given the likely absurd power level, and room interractions and noise can easily mask the first signs of a beginning "fuzz" over of the fine detail. It takes very little more for it to be clearly crackling beyond that edge, so not much warning at all to be had.
Conversely, if you've a very clean and revealing amp as a high grade type of UCD, and played highly compressed music that already offers sheared transients en masse in the very recording, with the amp of course well away from ever clipping itself..the sheared recorded transients will be just as piercing and painful as though the recording were clean and dynamic but with the amp driven hard into clipping, even while at semi low power levels! While that's somewhat self protecting as you can't endure such junk on that kind of system, you can see how that too could mask the onset of real clipping entirely.
Suffice it to say yes crackling is the obvious marker that you've simply gone way too far and it's normal for any switching amp that doesn't employ countermeasures against hard clipping.
A single clipping event however should never produce a series of crackles, that would be estable and you need clean recovery, or make very certain that under no circumstances could it ever clip at all.
Whether or not such an event could couple noise produced on the carrier back into the input more than puts the cart before the horse.
Oh and if you should ever find a link to this philips paper I'd like to read it.
It seems our definitions of clipping differ somewhat.
In my view especially for hard clipping it's any sheared signal, and with switching amps that equates to any non modulated portions. It switches in order to modulate signals, and when it runs dry of gain, or as with "a ucd" as Bruno might say the feedback "bails out", modulation falls dead, "switching" stops, and it clips hard to either rail, for the duration of the clipped portion of the signal for which it lacked the gain to modulate. Obviously, this falls into the audioband, but I think you might agree it is not "switching" any longer, but hard clipping.
You might look at it as switching simply because for UCD the clipping recovery is robust and graceful, and the very onset of the highest frequencies would be rapid, but it is most assuredly not switching in the audio band! Lesser amps might hit the rail a single time and stick to it like a bug on flypaper.
Further, in terms of it occuring with "a UCD", the same exact thing would happen for a clocked amp that also lacked clipping countermeasures. That's just how a switching amp rolls, and yes it begins at higher frequencies first because it's those that are the first to be clipped. So it really has nothing at all to do with a non fixed frequency amp as any self oscillating is, switching into the audioband.
What does hard clipping in class d sound like?
If you have an amp which is suffering from penny pinched implementation or any other issue making it less that perfect, the clean crackles of the sheared signal portions being snapped off will first be masked by noise and high distortion, so you'll never see the first signs comming. At some point, the sound will stretch thin, grow coloured, become really loud and offensive, you know it's clipping hard already by now, but you just can't hear it crackling because it's not up to that kind of accuracy at all. Push it a little harder and it will crackle suddenly, so hard you'll ball up in the fetal position and suck your thumb.
So with some conditioning you end up hearing your warning that it's time to turn it down.
With a very clean and accurate amp it's another matter altogether. The sound will remain pure and uncoloured right up to the razor's edge. At this point speaker distortion, given the likely absurd power level, and room interractions and noise can easily mask the first signs of a beginning "fuzz" over of the fine detail. It takes very little more for it to be clearly crackling beyond that edge, so not much warning at all to be had.
Conversely, if you've a very clean and revealing amp as a high grade type of UCD, and played highly compressed music that already offers sheared transients en masse in the very recording, with the amp of course well away from ever clipping itself..the sheared recorded transients will be just as piercing and painful as though the recording were clean and dynamic but with the amp driven hard into clipping, even while at semi low power levels! While that's somewhat self protecting as you can't endure such junk on that kind of system, you can see how that too could mask the onset of real clipping entirely.
Suffice it to say yes crackling is the obvious marker that you've simply gone way too far and it's normal for any switching amp that doesn't employ countermeasures against hard clipping.
A single clipping event however should never produce a series of crackles, that would be estable and you need clean recovery, or make very certain that under no circumstances could it ever clip at all.
Whether or not such an event could couple noise produced on the carrier back into the input more than puts the cart before the horse.
Oh and if you should ever find a link to this philips paper I'd like to read it.
The normal ucd app-note is at:
http://www.nxp.com/acrobat_download/usermanuals/UM10155_1.pdf
There is another document on the NXP website as well which shows their recomended MOSFETs to use with ucd amplifiers, and very importantly, states the conditions under which license-free commercial use of the ucd design is allowed.
http://www.nxp.com/acrobat_download/usermanuals/UM10155_1.pdf
There is another document on the NXP website as well which shows their recomended MOSFETs to use with ucd amplifiers, and very importantly, states the conditions under which license-free commercial use of the ucd design is allowed.
Yes, this shows the most basic ucd circuit that has been known here for a long time. The Hypex ones are a great deal more sophisticated, but of course, their circuitry is not in the public domain!
People wanting to base their own commercial designs on the basic ucd circuit need to note the requirements for license-free use of the patented circuit as shown on the second note though.
(NXP simply want to use it to sell their own semiconductors, which seems fair enough to me).
People wanting to base their own commercial designs on the basic ucd circuit need to note the requirements for license-free use of the patented circuit as shown on the second note though.
(NXP simply want to use it to sell their own semiconductors, which seems fair enough to me).
The question was asked:
"Is it possible to show us the scope image just before and at clipping of output line after filter,before filter,gates,supply+-,drivers input?"
I could, although I'm in the middle of EMC susceptibility testing of the equipment using a full-bridge amplifier, based on the Philips ucd app-note circuit, at the moment.
Early next week I'll capture some 'scope traces of a simple, bridged ucd amplifier, just at the point of clipping, to show the drop in the self-oscillating frequency. (Although those of you with a SPICE program can easily see the effect by doing a simple transient run).
"Is it possible to show us the scope image just before and at clipping of output line after filter,before filter,gates,supply+-,drivers input?"
I could, although I'm in the middle of EMC susceptibility testing of the equipment using a full-bridge amplifier, based on the Philips ucd app-note circuit, at the moment.
Early next week I'll capture some 'scope traces of a simple, bridged ucd amplifier, just at the point of clipping, to show the drop in the self-oscillating frequency. (Although those of you with a SPICE program can easily see the effect by doing a simple transient run).
I experienced hearable frequency drop too, but I used a different implementation of UcD, with the saidly slow IR2110, and there is a small hysteresis. Also the idealized simulation of UcD shows significant freq drop. Hypex UcD claimed to be free of this kind of audible problems (due to it's semi-linear behaviour near comparing point, I guess), but I suspect that with pure bass it can be heard also.
Attached an output waveform of a dedicated bass amplifier (at 50 Hz).
Hysteresis-based self oscillating amps suffers much more on this problem.
Attached an output waveform of a dedicated bass amplifier (at 50 Hz).
Hysteresis-based self oscillating amps suffers much more on this problem.
Attachments
Hi!
My 5 cents for cracling problem
During hard clipping, the power capacitor for high side driver does not becomes charge, so it can fully discharge and high side mosfet will close. Next cycle the capacitor becomes charge, but amplifier is still in clipping and the process repeats... So, try to increase high side power capacitor saying to 470 mf. In UCD prototype it is in position C5.
Regards,
Alex
My 5 cents for cracling problem
Regards,
Alex
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
- Home
- Amplifiers
- Class D
- Cracking noise at clipping