In order to convert your clock-based amp to a self-oscillating one, you still need negative feedback, as in a clock based amplifier. You have to remove the oscillator and connect that comparator's input to GND. You have to modify the fb network also.
You have to take the fb signal from the switching stage, go to a comparator (through the fb network), and the overall path until you reach the high side mosfet driver input must be inverting.
If you look at IRF ref. design:
fb goes to integrator's negative input. Then goes to level shifter, which inverts. Then to 3 cascaded NOT gates, and from there to the HS input, so you have a odd number of invertions thus the overall "gain" is inverting.
And you have to provide a mean in which the circuit is able to start. In IRF design, it has a transistor controlled with a RC constant that drops the PWM signal forcing the low side mosfet ON and the HS mosfet OFF for a given time (about 100ms). This way a large error is created and when that PWM signal is released by the transistor, the NFB starts correcting and oscillate.
Needless to say that you will have a big pop at the speaker when powering up!
Hope this helps
You have to take the fb signal from the switching stage, go to a comparator (through the fb network), and the overall path until you reach the high side mosfet driver input must be inverting.
If you look at IRF ref. design:
fb goes to integrator's negative input. Then goes to level shifter, which inverts. Then to 3 cascaded NOT gates, and from there to the HS input, so you have a odd number of invertions thus the overall "gain" is inverting.
And you have to provide a mean in which the circuit is able to start. In IRF design, it has a transistor controlled with a RC constant that drops the PWM signal forcing the low side mosfet ON and the HS mosfet OFF for a given time (about 100ms). This way a large error is created and when that PWM signal is released by the transistor, the NFB starts correcting and oscillate.
Needless to say that you will have a big pop at the speaker when powering up!
Hope this helps
Pierre/Charles,
Thanks for that.
So basically to convert to self oscillating i would need to ground the appropriate input to the comparator and feed the other from the intergrator that is there at the moment.. As for the feedback i divide it down and feed it back into the -VE input of my opamp though the a resistor (R11)..
By the way Pierre and Charles your help is very very much appreciated.
Thanks for that.
So basically to convert to self oscillating i would need to ground the appropriate input to the comparator and feed the other from the intergrator that is there at the moment.. As for the feedback i divide it down and feed it back into the -VE input of my opamp though the a resistor (R11)..
By the way Pierre and Charles your help is very very much appreciated.
Neither UcD nor SODA, I added a simple resistor divider (18k/2k) from the output after filter to the inverting input of the comparator. The lower 2k resistor was paralleled with 240pF to keep the switching frequency below 500kHz (otherwise the driver and the 12V linear regulator heat up too much). This way it acts like a hardwired PD control, I guess. Yes, I can hear you, worse than SODA, but it works surprisingly stable for that application. The bottleneck is the ADC based temperature measurement now. Additionally I added a 10 Ohm resistor in series to the output cap, that seems to help stabilising the stage.
Thank you for your hints!
Regards, Timo
edit: Of course, I "killed" the triangle generator, it's not needed anymore.
Thank you for your hints!
Regards, Timo
edit: Of course, I "killed" the triangle generator, it's not needed anymore.
I have implemented my overvoltage & over current protection.
The over voltage protection works great. However the over current trips, but for varying amounts of time, depending on how much charge can be dumped into the capacitor.
I just had an idea, How would it be if i used a SCR in place of the PNP transistor i am using trip the over current protection. The gate of the scr can be set to fire if the peak current is ever reached, and hence shutting down the amplifier until it is reset / or power cycled.
any comments?? any reasons why this wouldn't work?
thanks
Peter
The over voltage protection works great. However the over current trips, but for varying amounts of time, depending on how much charge can be dumped into the capacitor.
I just had an idea, How would it be if i used a SCR in place of the PNP transistor i am using trip the over current protection. The gate of the scr can be set to fire if the peak current is ever reached, and hence shutting down the amplifier until it is reset / or power cycled.
any comments?? any reasons why this wouldn't work?
thanks
Peter
Forget about my last post, I realise why that would not work.
I ended up using one of the spare XOR gates i have to set up a latch, and it works very well.. I have recently removed all capacitors from the sense circuit in an attempt to speed it up even more.
However I am trying to set the amplifier up so that the output is short circuit proof...... (if possible) At the moment if i pump too much power into a very low impedance load it instantly trips and stays off until reset, which is exactly what i want.. btw, each rail is fused with a 10A fast blow fuse. There is only 1000uF of capacitance after the fuse...(ie if the fuse blows there is 1000uF on each rail, as opposed to about 17,000uF)
However if i short the outputs it trips but always manages to kill my fets in the process...
1) is it because I am only sensing on the +V rail into the h-bridge... I would have thought that the lower FET could stand being shorted for a couple of uS......
2) Should i add a little bit of impedance inside the amplifier to limit the short circuit current to something a bit more reasonable? Ie 0.1Ohm , I though that the 100uH output inductor would have been able to hold back one cycle cycle at least.
The problem with this is i will generate a fair bit of heat with the resistor,,Or should i just give up on the idea of protecting the amplifier in this way...
Regards
Peter.
I ended up using one of the spare XOR gates i have to set up a latch, and it works very well.. I have recently removed all capacitors from the sense circuit in an attempt to speed it up even more.
However I am trying to set the amplifier up so that the output is short circuit proof...... (if possible) At the moment if i pump too much power into a very low impedance load it instantly trips and stays off until reset, which is exactly what i want.. btw, each rail is fused with a 10A fast blow fuse. There is only 1000uF of capacitance after the fuse...(ie if the fuse blows there is 1000uF on each rail, as opposed to about 17,000uF)
However if i short the outputs it trips but always manages to kill my fets in the process...
1) is it because I am only sensing on the +V rail into the h-bridge... I would have thought that the lower FET could stand being shorted for a couple of uS......
2) Should i add a little bit of impedance inside the amplifier to limit the short circuit current to something a bit more reasonable? Ie 0.1Ohm , I though that the 100uH output inductor would have been able to hold back one cycle cycle at least.
The problem with this is i will generate a fair bit of heat with the resistor,,Or should i just give up on the idea of protecting the amplifier in this way...
Regards
Peter.
Hello Peter, the experiments I have done indicated that two IRF840's can safely survive passing the power of only about two 50 uF caps with 160 volts on each. For your amp, 1000uF and 50v could be too much, depending on how the die size of your MOSFETs compares to the 840. For a rough idea, I'd say compare the two energy-related products, 8000 vs. 50000. I don't think the dies inside your MOSFETs are 5 times bigger, especially since the trend is to try to make them smaller to improve switching, and I suspect the trade-off is less robustness.
You might try reducing your capacitors after the fuses to 470 or 220uF. I'd try to avoid adding the resistor. If the MOSFET Rds-on is .1, the efficiency could be cut in half without accounting for switching losses. Best regards.
You might try reducing your capacitors after the fuses to 470 or 220uF. I'd try to avoid adding the resistor. If the MOSFET Rds-on is .1, the efficiency could be cut in half without accounting for switching losses. Best regards.
Hi Subwo1,
thanks for your reply.
I am actually running +/- 80V rails with a 1000uF on each rail... so the energy related product is much larger than 50,000.... I am just wondering little capacitance i can get away with ( post fuses) maybe 10uF on each rail?? or less.. Actually i might try these 2.2uF pollyprop caps i have.... I guess they really only need to supply power until the big caps can take over.
What do you think??
Regards
Peter
thanks for your reply.
I am actually running +/- 80V rails with a 1000uF on each rail... so the energy related product is much larger than 50,000.... I am just wondering little capacitance i can get away with ( post fuses) maybe 10uF on each rail?? or less.. Actually i might try these 2.2uF pollyprop caps i have.... I guess they really only need to supply power until the big caps can take over.
What do you think??
Regards
Peter
Pete, you are on the right track. The main function that they may have to do is the absorption of transients.
I would actually recommend something like relays which drop out between the big capacitors and the small ones. I have found that fuses tend to be lacking in terms of protecting semiconductors from overload unless the semiconductors are generously over-sized. You may even consider very large MOSFETs in place of the fuses since the main concern is on-state channel resistance. The control circuit could be charge pump or photovoltaic and simple resistor turn-on for the upper and lower rails respectively. Then, the shut-down could be by separate optocouplers pulling the protection MOSFET gates low. Only reasonably fast shut-off is needed and is easily done in ten or twenty microseconds by a common 4N35 opto. You could even pull the gates low with an NPN transistor for the top rail and a PNP/NPN combination for the lower rail. The charge pump for the upper rail could operated by the power transformer.
Photovoltaic MOSFET gate enhancement:
http://www.electronicstalk.com/news/mat/mat146.html
I would actually recommend something like relays which drop out between the big capacitors and the small ones. I have found that fuses tend to be lacking in terms of protecting semiconductors from overload unless the semiconductors are generously over-sized. You may even consider very large MOSFETs in place of the fuses since the main concern is on-state channel resistance. The control circuit could be charge pump or photovoltaic and simple resistor turn-on for the upper and lower rails respectively. Then, the shut-down could be by separate optocouplers pulling the protection MOSFET gates low. Only reasonably fast shut-off is needed and is easily done in ten or twenty microseconds by a common 4N35 opto. You could even pull the gates low with an NPN transistor for the top rail and a PNP/NPN combination for the lower rail. The charge pump for the upper rail could operated by the power transformer.
Photovoltaic MOSFET gate enhancement:
http://www.electronicstalk.com/news/mat/mat146.html
Pete, the shutdown circuit would also need a separate power supply derived before the protection MOSFETs. Your idea about triggering an SCR could work. Using a logic gate could be even better since you could use regeneration to make it latch itself on, plus its response is loads faster than an SCR. It would keep the protection MOSFETs off until the main power is cycled off and on to reset the protection circuit. The short can then be found and remedied while the power is off.
If you want a really modern feel to the protection circuit operation, include a greatly reduced power supply that is constantly waiting behind blocking diodes. You would then not include the latch. But then the sensitivity of the protection circuit must simultaneously increase so that it can remain engaged under the reduced power.
If you want a really modern feel to the protection circuit operation, include a greatly reduced power supply that is constantly waiting behind blocking diodes. You would then not include the latch. But then the sensitivity of the protection circuit must simultaneously increase so that it can remain engaged under the reduced power.
If you add active protection like the MOSFETs, you might like to try smaller capacitors there again just to see how much the difference is again. In fact, if you were to try it, I suggest the IRFBA90N20D. With an Rds-on of .023 Ohms, I'd say they wouldn't need much of a heat sink if any. A small clip-on type could suffice. Peak power damping should be way less than with a fuse because of the much greater thermal mass over a small wire filament.
Can people please have a look at my schematic / pcb files for suggestions.
I'll look into that, but i think for my first board i will leave off the active protection FETS, and just stick with fuses. It should never really be shorted anyway ( and even then its only really a problem at a high power output)......
I have attached the board / schematic i am planning to get manufactured in a couple of days, could anyone thats interested please have a look at my schematic / pcb. If anyone has any suggestions etc I would really appreciate them... I am yet to add some sort of triangle wave generator i am thinking i will just just a fast dual op amp, use one as a schmitt trigger, the output squarewave will feed the input to intergrator (ie the other opamp).... and just feedback the triangle wave to the schmitt trigger......
the pdf file can be found at http://tesla.reidconsulting.com.au/Sch-PCB.pdf
Thank you.
Regards
Peter.
I'll look into that, but i think for my first board i will leave off the active protection FETS, and just stick with fuses. It should never really be shorted anyway ( and even then its only really a problem at a high power output)......
I have attached the board / schematic i am planning to get manufactured in a couple of days, could anyone thats interested please have a look at my schematic / pcb. If anyone has any suggestions etc I would really appreciate them... I am yet to add some sort of triangle wave generator i am thinking i will just just a fast dual op amp, use one as a schmitt trigger, the output squarewave will feed the input to intergrator (ie the other opamp).... and just feedback the triangle wave to the schmitt trigger......
the pdf file can be found at http://tesla.reidconsulting.com.au/Sch-PCB.pdf
Thank you.
Regards
Peter.
I am glad to see that you will finally stick to a triangle based amplifier, I like them more 
I have a few questions for you:
-Did the overcurrent protection work as it is now? (based on the IRF design but with a latch) When you start increasing amplitude on a heavy load, does it shutdown until you reset the pushbutton?
-What happens if you short the output and then turn on the amplifier? The mosfets shouldn't die. I think the only possibility of mosfet failure is shorting the output when the amp is delivering high power, right?
-I did the +15/-15V supplies exactly as you! But I don't have a resistor before the transistor. Did you tried without them? (I mean R2 and R4). BTW: It is possible to do the 12V supply for the driver exactly the same way, taking the input to the transistor from GND, as it is a positive voltage with respect to -HT. However, the average current consumption is a bit higher, say 75mA. What is this current in your circuit?
-What are R5 and R6 for?
-I see you have paralleled 2 coilcraft coils, this way you have halft the inductance and twice the current capability. I would make sure that they have exactly the same impedance in order to ensure proper current sharing.
-What ultrafast diodes are you using for the output stage bootstrap, clamping and gate drive? I use MUR120 with success.
-Have you done measurements of THD and noise?
I will make more comments as soon as I have time to look at the sch carefully.
Best regards,
Pierre
I have a few questions for you:
-Did the overcurrent protection work as it is now? (based on the IRF design but with a latch) When you start increasing amplitude on a heavy load, does it shutdown until you reset the pushbutton?
-What happens if you short the output and then turn on the amplifier? The mosfets shouldn't die. I think the only possibility of mosfet failure is shorting the output when the amp is delivering high power, right?
-I did the +15/-15V supplies exactly as you! But I don't have a resistor before the transistor. Did you tried without them? (I mean R2 and R4). BTW: It is possible to do the 12V supply for the driver exactly the same way, taking the input to the transistor from GND, as it is a positive voltage with respect to -HT. However, the average current consumption is a bit higher, say 75mA. What is this current in your circuit?
-What are R5 and R6 for?
-I see you have paralleled 2 coilcraft coils, this way you have halft the inductance and twice the current capability. I would make sure that they have exactly the same impedance in order to ensure proper current sharing.
-What ultrafast diodes are you using for the output stage bootstrap, clamping and gate drive? I use MUR120 with success.
-Have you done measurements of THD and noise?
I will make more comments as soon as I have time to look at the sch carefully.
Best regards,
Pierre
PCB:
- do not place vias into pads (possibility of short circuits, clearance)
- try to avoid antennas (unconnected tracks), delete them or connect them to GND, using vias
- try to get shorter tracks on the power side (beginning from the gate driver upt to the connections between output filter elements
- I would also try to use only one inductor
I used AD8034 in the triangle generator, because they have low noise, 1pA input current (not needed really) and 80MHz bandwith combined with +/-12V capability and about 3mA quiescent current.
Regards and good luck,
Timo
- do not place vias into pads (possibility of short circuits, clearance)
- try to avoid antennas (unconnected tracks), delete them or connect them to GND, using vias
- try to get shorter tracks on the power side (beginning from the gate driver upt to the connections between output filter elements
- I would also try to use only one inductor
I used AD8034 in the triangle generator, because they have low noise, 1pA input current (not needed really) and 80MHz bandwith combined with +/-12V capability and about 3mA quiescent current.
Regards and good luck,
Timo
Pierre,
Yes, The overcurrent protection works as it is now. When then the threshold is met, set by R20,R21 & The capacitor sets the time constant. It trips until reset by the N/C switch. This also latches on over voltage ie supply pumping.
If i have the output shorted when the amplifier turns on - well it depends if there is any input signal, if there isn't then the over current doesn't trip. I do know that it trips if there is a decent signal coming out of the amplifier on start up if it is shorted. But last time i checked, shorting the output leads at very high power levels has always killed the fets.
With regards to the +/- 15v supplys, I had figured that i didn't really need R2 & R4, I guess i will leave them off the PCB to make it simpler. I realise i could to the +12v supply the same way, My driver circuit takes about 70mA, which equates too 5.1 watts of heat generated. Hence i have decided to go to a small transformer. I didn't want to set up a switching converter in this design.
R5 & R6 are shown in the schematic as resistors, but they are actually a snubber circuit, ie Resistor & capacitor in series.
Yes i have paralleled two coils, but i am guessing i should actually use one as tiki suggested, I think i will leave provision for a hand wound toroid one on there.
I have done THD & noise tests I posted them up a while ago, However i suspect that a decent circuit board will help them a lot.
Thank you for you input so far.
Regards
Peter
Yes, The overcurrent protection works as it is now. When then the threshold is met, set by R20,R21 & The capacitor sets the time constant. It trips until reset by the N/C switch. This also latches on over voltage ie supply pumping.
If i have the output shorted when the amplifier turns on - well it depends if there is any input signal, if there isn't then the over current doesn't trip. I do know that it trips if there is a decent signal coming out of the amplifier on start up if it is shorted. But last time i checked, shorting the output leads at very high power levels has always killed the fets.
With regards to the +/- 15v supplys, I had figured that i didn't really need R2 & R4, I guess i will leave them off the PCB to make it simpler. I realise i could to the +12v supply the same way, My driver circuit takes about 70mA, which equates too 5.1 watts of heat generated. Hence i have decided to go to a small transformer. I didn't want to set up a switching converter in this design.
R5 & R6 are shown in the schematic as resistors, but they are actually a snubber circuit, ie Resistor & capacitor in series.
Yes i have paralleled two coils, but i am guessing i should actually use one as tiki suggested, I think i will leave provision for a hand wound toroid one on there.
I have done THD & noise tests I posted them up a while ago, However i suspect that a decent circuit board will help them a lot.
Thank you for you input so far.
Regards
Peter
tiki,
I have taken into account what you have said, I have corrected those via's though the IC pads, also i have dramatically shortened the leads to the gates etc of the Mosfets.
I am yet to tighten up the Power side of things, that will be determinted by the footprint i need to leave to the single inductor... The low voltage side is not much tighter as well.. I have posted new schematics / pcb files. The triangle wave generator has also been added. Some intput offset adjustment
resistor pads have been added for removing any dc at the output.
the updated pcb / schematic is here
http://tesla.reidconsulting.com.au/Fixed.pdf
Regards
Peter
I have taken into account what you have said, I have corrected those via's though the IC pads, also i have dramatically shortened the leads to the gates etc of the Mosfets.
I am yet to tighten up the Power side of things, that will be determinted by the footprint i need to leave to the single inductor... The low voltage side is not much tighter as well.. I have posted new schematics / pcb files. The triangle wave generator has also been added. Some intput offset adjustment
resistor pads have been added for removing any dc at the output.
the updated pcb / schematic is here
http://tesla.reidconsulting.com.au/Fixed.pdf
Regards
Peter
Peter, I have noticed you have made 2 separate ground planes in the PCB. That's the way to go, I think. But how do you join them? I usually join them with a small trace at a position that is the nearest to the GND faston (as possible). Other people use small chokes to separate signal and power grounds. What have you done with this issue?
The board looks quite good to me.
Best regards,
Pierre
The board looks quite good to me.
Best regards,
Pierre
Relying on a post from Bruno, I think, one issue is to let the signal go back to the incoming edge of the board, NOT to flow through. This avoids ground noise sometimes. The UcD layout is done in this way - all connectors at one edge.
On my board I found a layout (conceptual) mistake this way, the RS232-driver with charge pumps included, disturbes the GND and Vcc-lines heavily. It is located at the "end" of the Vcc-line, the µcontroller with integrated ADC is between it and the supply IC. It needed additional 100µF Oscon to decouple it fairly.
I for myself love fat GND-planes, the more, the better. I used 4 copper planes in sum (2 of them are GND), connected all floating islands to GND and put a lot of vias between the planes. Look at RF PCBs like mobile phones.
Regards, Timo
For clarity: I did not do the layout by myself (too lazy
), but I gave a lot of hints to the (already well experienced) person.
On my board I found a layout (conceptual) mistake this way, the RS232-driver with charge pumps included, disturbes the GND and Vcc-lines heavily. It is located at the "end" of the Vcc-line, the µcontroller with integrated ADC is between it and the supply IC. It needed additional 100µF Oscon to decouple it fairly.
I for myself love fat GND-planes, the more, the better. I used 4 copper planes in sum (2 of them are GND), connected all floating islands to GND and put a lot of vias between the planes. Look at RF PCBs like mobile phones.
Regards, Timo
For clarity: I did not do the layout by myself (too lazy
), but I gave a lot of hints to the (already well experienced) person.- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.