In a well-designed UCD this effect would be insignificant. (Ergo this is not well-designed, the phase lead network in feedback is not efficient enough.)
This is the schematic of what i have on my board: classdlabs :: basicclassdsch10.png picture by TheElectroNoob - Photobucket
The double capacitors on the input is just to accomodate both 2.5mm and 5mm size.
Its also a very unreliable design, third failure. It was running fine with a stone cold heatsink until i turn the oscilloscope on, then the amp makes a few bangs and starts humming, then after a few seconds of humming, the music goes away, then comes the smell of burnt ic package and the heatsink and powet transformer are both too hot to touch.
I guess the scope generated a spike on the mains.
Perhaps i should call off making a stereo version of this piece of **** design. Seem like it was just a waste of time.
The double capacitors on the input is just to accomodate both 2.5mm and 5mm size.
Its also a very unreliable design, third failure. It was running fine with a stone cold heatsink until i turn the oscilloscope on, then the amp makes a few bangs and starts humming, then after a few seconds of humming, the music goes away, then comes the smell of burnt ic package and the heatsink and powet transformer are both too hot to touch.
I guess the scope generated a spike on the mains.
Perhaps i should call off making a stereo version of this piece of **** design. Seem like it was just a waste of time.
Last edited:
- Your PCB layout of differential feedback is wrong
- You are using no EMI filtering on the supply rails, the amplifier can easily disturb itself because new oscillation modes at much higher frequencies may appear due to unwanted feedback
- Your idea of paralleling dissimilar capacitors at the output is likely to result in several oscillation modes, one at the desired frequency and some more at much higher ones, that will easily lead to amplifier failure
Anything can trigger these unwanted oscillation modes if they exist.
A simple frequency detection circuit with 2 capacitors, 2 diodes, one resistor and one transistor (feding a two transistor latch) shuts down my amplifier if oscillation frequency goes too high for more than 2-3 cycles.
Additionally, a minimum pulse width circuit with a quad NAND gate prevents the IR2110 from malfunctioning due to too narrow pulses at the inputs. This is a known failure mode of IR2xxx chips.
Theoretically these circuits are not needed because a good amplifier with a good layout will never exhibit these wrong oscillation modes. However, I'm keeping them in the commercial version because they have already saved many output stages when things go wrong.
- You are using no EMI filtering on the supply rails, the amplifier can easily disturb itself because new oscillation modes at much higher frequencies may appear due to unwanted feedback
- Your idea of paralleling dissimilar capacitors at the output is likely to result in several oscillation modes, one at the desired frequency and some more at much higher ones, that will easily lead to amplifier failure
Anything can trigger these unwanted oscillation modes if they exist.
A simple frequency detection circuit with 2 capacitors, 2 diodes, one resistor and one transistor (feding a two transistor latch) shuts down my amplifier if oscillation frequency goes too high for more than 2-3 cycles.
Additionally, a minimum pulse width circuit with a quad NAND gate prevents the IR2110 from malfunctioning due to too narrow pulses at the inputs. This is a known failure mode of IR2xxx chips.
Theoretically these circuits are not needed because a good amplifier with a good layout will never exhibit these wrong oscillation modes. However, I'm keeping them in the commercial version because they have already saved many output stages when things go wrong.
So a squeeky clean carrier resudial have nothing to do with emi and ringing and such things that can lead to failure ? And a 20-30 year old 25MHz scope is prolly just too slow to see the ringing galore i guess.
My "differential" feedback is taken directly from the valueless schematic at hypex.nl
But yes i know my pcb layout is a very bad one. supply emi filtering goes between the psu and amp or between fets and filter caps ?
My IR2010 have survived two mosfet failures so far. And since i went back to IRF540 from HUF75337 the switching frequency seem to have doubled, meaning its now around 300-400kHz as opposed to 100-200kHz before.
Remember this is a hobby project i do for fun, not to make money. On my way to "school" today i was thinking of adding current sensing to shut down the amp if current rises above a certain value, but i know that this kind of protection is usually too slow to save the fets from say a short on the output.
I also guess that trying to get an AM station picked up with a radio close to the amp is not conclusive either, i cant really perform such test anyways since we dont have any am stations in sweden anymore.
My "differential" feedback is taken directly from the valueless schematic at hypex.nl
But yes i know my pcb layout is a very bad one. supply emi filtering goes between the psu and amp or between fets and filter caps ?
My IR2010 have survived two mosfet failures so far. And since i went back to IRF540 from HUF75337 the switching frequency seem to have doubled, meaning its now around 300-400kHz as opposed to 100-200kHz before.
Remember this is a hobby project i do for fun, not to make money. On my way to "school" today i was thinking of adding current sensing to shut down the amp if current rises above a certain value, but i know that this kind of protection is usually too slow to save the fets from say a short on the output.
I also guess that trying to get an AM station picked up with a radio close to the amp is not conclusive either, i cant really perform such test anyways since we dont have any am stations in sweden anymore.
Earlier today i made a single ended to balanced converter just out of curiosity, turns out i get twice the volume with it, compared to single ended input with the other input grounded.
Heres a vid of the experiment with a 7 meter cable:
YouTube - Balanced ausio signal transmission
Heres a vid of the experiment with a 7 meter cable:
YouTube - Balanced ausio signal transmission
Im now trying out sa passive pole on my prototype, this is what i've wired up:
Except theres 2 of these, one on each input wheres one have "spk out" connected to ground since it doesent have an output to go to since its not a BTL amp.
Switching frequency according to my old scope is around 6.5µS which i guess is around 150-200kHz.
Im not sure if i need larger or smaller cap in phase lead network to increase frequency, but it does get higher with a resistive load and drops a bit with an inductive load due to an inefficient phase lead network.
More experimenting to come.
Except theres 2 of these, one on each input wheres one have "spk out" connected to ground since it doesent have an output to go to since its not a BTL amp.
Switching frequency according to my old scope is around 6.5µS which i guess is around 150-200kHz.
Im not sure if i need larger or smaller cap in phase lead network to increase frequency, but it does get higher with a resistive load and drops a bit with an inductive load due to an inefficient phase lead network.
More experimenting to come.
This looks like a nice project, and I'm tempted to try it myself. One thing about the single and double sided PCB's, is somehow possible to make double sided PCB's at home using the toner transfer method by simply sandwiching together 2 single sided PCB's back to back? I guess you'd have to make sure everything was soldered on both sides, but this may allow a more noise free design if it works? Or is there some reason why this would not work?
First, you have to look at carrier residual with several amperes flowing from and to the output of the amplifier, most ringing is hidden when the amplifier is idle because it's switching in resonant mode.
Second, a 25Mhz oscilloscope may not be able to display the ringing frequencies with its proper amplitude (serious attenuation and phase shift above 25Mhz). I can get ringing as high as 110Mhz in my IRFB4227 output stage if I remove the snubbers (high ringing frequencies in hard switching circuits are a proof of good layout).
Third, what you tell is not enough criteria to ensure that the self oscillating loop is properly compensated.
Paralleling dissimilar output capacitors may result in peaks and dips in output impedance that will disturb carrier residual only for certain pulse widths (it looks fine with no signal but may misbehave as the amplifier approaches clipping).
Turning your oscilloscope on and off may produce brief EMI spikes, usually picked up by and heard as harmless clicks in audio equipment, faint in good designs, loud in bad designs. One of these clicks can easily clip an amplifier for a small fraction of a second. If the amplifier has some hidden design flaw, it may show up...
For example, yesterday one of my prototipes got into protect mode as my mobile phone started to ring (no damage). It was unusually close. These things happen...
Im NOT parallelling dissimilar output capacitors, just provinding the ability to use more than one size, like 2.5mm and 5mm.
I tried using an am radio close to the amp as well as in another room and nothing, with the antenna just millimeters away i was abe to get a barely hearable signal around 1MHz with the radios volume on max. Compared to my previous ones which were clearly hearable all over the am band.
Btw eva, you do this stuff for a living so you always try to look down at others designs in fear of competition and such just to remain the queen of class d on here.
The only way to check the carrier resudial with several A of current is to use DC on the input and a dummy load, otherwise the resudial will just be a blur on the screen.
Turning the scope on/off does nothing, not even a click in the speaker.
I tried using an am radio close to the amp as well as in another room and nothing, with the antenna just millimeters away i was abe to get a barely hearable signal around 1MHz with the radios volume on max. Compared to my previous ones which were clearly hearable all over the am band.
Btw eva, you do this stuff for a living so you always try to look down at others designs in fear of competition and such just to remain the queen of class d on here.
The only way to check the carrier resudial with several A of current is to use DC on the input and a dummy load, otherwise the resudial will just be a blur on the screen.
Turning the scope on/off does nothing, not even a click in the speaker.