Primary waveform looks rather strange. Is your chasis AC (or DC) coupled to ground? Also, secondary waveform reveals that you are not waiting for complete resonance before turning on the opposite side.
BTW: Yesterday I toasted a 3KW power factor corrector prototype (80-280V AC input, 415V DC output) because I soldered one end of a jumper to the wrong place. This is prone to fireworks
BTW: Yesterday I toasted a 3KW power factor corrector prototype (80-280V AC input, 415V DC output) because I soldered one end of a jumper to the wrong place. This is prone to fireworks
the chassis is connected directly to ground right at the input terminals via #14 wire. board ground is connected to terminal via two 8ga wires.
tripath module ground is also directly connected to ground.
the first waveform pic is the waveform right at the mosfet gate.
second pic is right at the primary (mosfet drain)
third is after the half bridge of the T amp
luka,
I'm using a headlight bulb with both high and low beam in parallel. the whole thing is connected between car battery + and amp input.
nothing was toasted. the diode seems to be still functional. just a black mark on the heatsink and chassis.
tripath module ground is also directly connected to ground.
the first waveform pic is the waveform right at the mosfet gate.
second pic is right at the primary (mosfet drain)
third is after the half bridge of the T amp
luka,
I'm using a headlight bulb with both high and low beam in parallel. the whole thing is connected between car battery + and amp input.
nothing was toasted. the diode seems to be still functional. just a black mark on the heatsink and chassis.
that's what I'm using here right now.
it's fast recovery. voltage drop in the diode test using DMM is about 300mV.
another note: transformers stay cold.
output inductors get perceptively warm. that's barely warm. residual at speaker outputs is rather high though at 3Vpk (a clean sine wave at the osc freq of the tripath chip) which has always been there even when I first built it.
it's fast recovery. voltage drop in the diode test using DMM is about 300mV.
another note: transformers stay cold.
output inductors get perceptively warm. that's barely warm.
residual at speaker outputs is rather high though at 3Vpk (a clean sine wave at the osc freq of the tripath chip) which has always been there even when I first built it.
I was future-proofing the supply. so when I get a bigger amplifier module, no need to make another supply.
alas, this thing just caught on fire!
well, sort of. I was having weird waveforms at the output when moderately loaded so I adjusted the freq and increased it a bit.
the waveform was a normal square wave but has a huge 40v peak with flat tops during deadtime so I thought "hey, let's increase the freq!" so I did and it worked fine for a couple of minutes then the diodes (1N4937) at the primaries (for voltage doubling for the gate supply) smoked and I have a few delaminated copper foil and four dead diodes. haven't checked for any other damages but it blew the 60A fuse at the power input so something must have also blown.
btw, I have lots of litz wire. it also makes winding the coils easy as they are easier to bend.
alas, this thing just caught on fire!
well, sort of. I was having weird waveforms at the output when moderately loaded so I adjusted the freq and increased it a bit.
the waveform was a normal square wave but has a huge 40v peak with flat tops during deadtime so I thought "hey, let's increase the freq!" so I did and it worked fine for a couple of minutes then the diodes (1N4937) at the primaries (for voltage doubling for the gate supply) smoked and I have a few delaminated copper foil and four dead diodes. haven't checked for any other damages but it blew the 60A fuse at the power input so something must have also blown.
btw, I have lots of litz wire.
it also makes winding the coils easy as they are easier to bend.
Big turn off spikes with moderate load will only happen if transformer is saturating. They will happen in one direction. You have to be careful when rectifying the drains, the circuit has to be able to handle these spikes. I use ferrite beads in series with the diodes to prevent resonances and to skip at least the initial part of the spikes. I also use beads in the auxiliary supply windings for the same reason. Beads also prevent stress to the 1A diodes (and are a loss-less alternative to resistors).
You have to aim for symmetrical turn-off spikes on both MOSFET banks. For a given load, the smaller the amplitude of the spikes the higher the quality of the layout and the transformer. When the two primaries are litz style they should be wound interleaved (ABABAB rather than AAABBB), this reduces the height of the spikes, EMI and MOSFET heating (and causes trouble if you don't use resonant turn-on).
You should add dead time and reduce excess load capacitance on transformer windings (from snubbers, diodes and MOSFETs) until you achieve resonant switching. Reducing frequency will increase magnetizing current resulting in a faster transition, as long as some safety margin to core saturation is left. I mean that you should allow the transformer itself to completely slew the voltage to the opposite state after turning off one bank of MOSFET and before turning on the other. This ensures balanced core operation and gets rid of half of the EMI btw.
The self balancing trick works as follows: When the core is more magnetised in one direction, magnetizing current at turn-off (after excitation in that direction) is higher resulting in a faster transition to the opposite state and a shorter pulse in that direction (and longer in the other).
I optimized my tiny power supply to switch over 60A without any spike or balancing problem (the limiting factor is only heating, which now you know is not significant with class D). This little two-transistor power supply is capable of 800W without any stress other than I^2*R losses.
You have to aim for symmetrical turn-off spikes on both MOSFET banks. For a given load, the smaller the amplitude of the spikes the higher the quality of the layout and the transformer. When the two primaries are litz style they should be wound interleaved (ABABAB rather than AAABBB), this reduces the height of the spikes, EMI and MOSFET heating (and causes trouble if you don't use resonant turn-on).
You should add dead time and reduce excess load capacitance on transformer windings (from snubbers, diodes and MOSFETs) until you achieve resonant switching. Reducing frequency will increase magnetizing current resulting in a faster transition, as long as some safety margin to core saturation is left. I mean that you should allow the transformer itself to completely slew the voltage to the opposite state after turning off one bank of MOSFET and before turning on the other. This ensures balanced core operation and gets rid of half of the EMI btw.
The self balancing trick works as follows: When the core is more magnetised in one direction, magnetizing current at turn-off (after excitation in that direction) is higher resulting in a faster transition to the opposite state and a shorter pulse in that direction (and longer in the other).
I optimized my tiny power supply to switch over 60A without any spike or balancing problem (the limiting factor is only heating, which now you know is not significant with class D). This little two-transistor power supply is capable of 800W without any stress other than I^2*R losses.
I know the core is not saturating. as during initial testing, I can drop the number of turns to 5 turns per primary before drawing lots of current. I'm using 6.
when loaded at freq's around 60kHz I get lots of ringing but the spikes aren't that big and the supply works fine.
when running at 40kHz, there is a near perfect square wave at idle/low loads and has spikes at turn on and turn off of the fets. but it is only one large spike with flat tops per zero crossing (so the square wave appears as "_H_H_" rather than "_n_n_" and the supply output voltage drops even when the input remains the same.
I think what blew the diodes were as you said, the resonances and ringing that have rather high freq and the diodes could not switch that fast.
I'm thinking another aux winding to use for the gate supply rather than rectify the drains.
now I'll look for ferrite beads and add them on the aux windings.
regarding the windings, I just bundled up 30 wires and wound everything litz style into the core. then split them into two to make two primaries. each trafo has only one secondary.
I doubt that the core is the wrong type as I have another SMPS using the exact same core (but this one is lower power and uses one core) and is running without problems.
I'm suspecting I may have a rather "heavy" diode. I'm using four RHRG75120 diodes which might be messing with the resonance of the trafo?
when loaded at freq's around 60kHz I get lots of ringing but the spikes aren't that big and the supply works fine.
when running at 40kHz, there is a near perfect square wave at idle/low loads and has spikes at turn on and turn off of the fets. but it is only one large spike with flat tops per zero crossing (so the square wave appears as "_H_H_" rather than "_n_n_" and the supply output voltage drops even when the input remains the same.
I think what blew the diodes were as you said, the resonances and ringing that have rather high freq and the diodes could not switch that fast.
I'm thinking another aux winding to use for the gate supply rather than rectify the drains.
now I'll look for ferrite beads and add them on the aux windings.
regarding the windings, I just bundled up 30 wires and wound everything litz style into the core. then split them into two to make two primaries. each trafo has only one secondary.
I doubt that the core is the wrong type as I have another SMPS using the exact same core (but this one is lower power and uses one core) and is running without problems.
I'm suspecting I may have a rather "heavy" diode. I'm using four RHRG75120 diodes which might be messing with the resonance of the trafo?
well, that's what the DMM diode test said. hehehe datasheet volt drop must be at full load.
I'm at a loss as to what to do right now. might be just tired. I'm gonna think of a solution in my dreams.
BTW, I have snubbers at pri. 10R + 5n6 at each primary and 10k + 5n6 at the secondary. just so other might have an idea what I did wrong.
I have tried trafo freq from 18kHz to 70kHz and none works best. with high freq around 60kHz, we get lots of spikes and ringing but the supply works and output voltage is stiff.
below about 40kHz, we get a nice square wave with single spikes at each zero crossing and output voltage drops and spike voltage rises when loaded.
I'm at a loss as to what to do right now. might be just tired. I'm gonna think of a solution in my dreams.
BTW, I have snubbers at pri. 10R + 5n6 at each primary and 10k + 5n6 at the secondary. just so other might have an idea what I did wrong.
I have tried trafo freq from 18kHz to 70kHz and none works best. with high freq around 60kHz, we get lots of spikes and ringing but the supply works and output voltage is stiff.
below about 40kHz, we get a nice square wave with single spikes at each zero crossing and output voltage drops and spike voltage rises when loaded.
Hi
Most amps run 40kHz and lower, so try to stay there if you can. DMM won't tell you nothing usefull but if diode is ok or not, use datasheets for info... and you must look for voltage at some current, not 0 or idle/minimum. Those diodes will produce probably most of the heat, if you don't include amp.
What you could do first is test smps with gates on input voltage. You are worried about fets getting out of saturation or get lower saturation, but your car voltage should be pritty high all the time, since you have this powerfull amp inside, which should be enough to drive all fets properly. Hope you got some good battery, that doesn't sag a lot...
Most amps run 40kHz and lower, so try to stay there if you can. DMM won't tell you nothing usefull but if diode is ok or not, use datasheets for info... and you must look for voltage at some current, not 0 or idle/minimum. Those diodes will produce probably most of the heat, if you don't include amp.
What you could do first is test smps with gates on input voltage. You are worried about fets getting out of saturation or get lower saturation, but your car voltage should be pritty high all the time, since you have this powerfull amp inside, which should be enough to drive all fets properly. Hope you got some good battery, that doesn't sag a lot...
everything is running fine on the SMPS. the only thing I'm having issues with is the trafo.
tested with various freq's and none gave the best performance. so I'm guessing there's a problem somewhere but I'm too tired to think about it. I've been working on this these past few days.
I got a bitmap file of the PCB layout if you want to see it but it is almost 4MB.
tested with various freq's and none gave the best performance. so I'm guessing there's a problem somewhere but I'm too tired to think about it. I've been working on this these past few days.
I got a bitmap file of the PCB layout if you want to see it but it is almost 4MB.
I'll try to post it full size first....wait a few minutes.
edit:
PCB layout
just note that the high current traces are not filled in because I did this board manually using felt tip pen and filled it by hand as I drew the layout on blank FRP.
edit:
PCB layout
just note that the high current traces are not filled in because I did this board manually using felt tip pen and filled it by hand as I drew the layout on blank FRP.