In my SOCD design (stopped due to no time) I used MAX961, so I would automatically get a inverted drive signal.
See http://www.diyaudio.com/forums/clas...illating-post-filter-feedback-topology-6.html
Maybe there is something in the design you can use, like analogue input filter, supply circuits, blocking diodes,....
See http://www.diyaudio.com/forums/clas...illating-post-filter-feedback-topology-6.html
Maybe there is something in the design you can use, like analogue input filter, supply circuits, blocking diodes,....
Hm... and you skipped the level shifter between comparator and IRS and simply accepted that the speaker terminals are at +78V DC vs GND.
Also possible, but not my preference.
The blocking diodes is definitely a key question and one of the main risks in my approach. From the view of a proper high power design they would be indicated, no doubt - yes. But then I have the same amount of power devices like paralleling 3x IRFB4227. ...circuit is already now to complex. On the other hand - just 4 devices more, not to bad.
Hard switching with the body diodes of the IRFP4668 is the reason why I reworked my design from 400kHz to 320kHz.
I remember earlier real life measurements with the IRFP4668 where I was not comfortable with the body diodes of that beast for a clocked 400kHz design, unfortunately I messed the documentation....
For a self oscillating type which reduces the switching frequency at high levels I think (or pray?) it should be possible to get it going. I guess most critical will be the situation of a short, which results in high currents and high switching frequency until the protection trips.
Do you have real life comparisons on hand at 300kHz hard switching 50A at 170V of the IRFP4668 with and w/o blocking diodes?
P.S. To bad that you stopped your SOCD.
Also possible, but not my preference.
The blocking diodes is definitely a key question and one of the main risks in my approach. From the view of a proper high power design they would be indicated, no doubt - yes. But then I have the same amount of power devices like paralleling 3x IRFB4227. ...circuit is already now to complex. On the other hand - just 4 devices more, not to bad.
Hard switching with the body diodes of the IRFP4668 is the reason why I reworked my design from 400kHz to 320kHz.
I remember earlier real life measurements with the IRFP4668 where I was not comfortable with the body diodes of that beast for a clocked 400kHz design, unfortunately I messed the documentation....
For a self oscillating type which reduces the switching frequency at high levels I think (or pray?) it should be possible to get it going. I guess most critical will be the situation of a short, which results in high currents and high switching frequency until the protection trips.
Do you have real life comparisons on hand at 300kHz hard switching 50A at 170V of the IRFP4668 with and w/o blocking diodes?
P.S. To bad that you stopped your SOCD.
My 2 cents.
Every thing is fine when the ambient temperature is close to 25-35C, you have RDS to shunt the body diode and it works great, but in reality in high power amplifier, the temperature goes up and then the RDS shunting no longer helps you anymore[RDS wents up], more the body diode conducts more the Qrr charge/discharge takes place, more the Trr it takes, thus the reverse recovery current increases much more than expected, more EMI, less reliability. PRONE to failure in field.
The only proven and bruteforce solution which gives you peace of mind is series schottky with anti parallel high speed but soft recovery diode which must be rated for 125C for its specs, not to get fooled by some datasheets showing specs with 25C, 20V, 30A etc like deceptions. Look for 125C, 400V, 30A values and associated di/dt and Trr results along with it.
Yes the dissipation rises because of series schottky but not much, but in reality what you get is a reliable product. Free from ill-effects of temperature rise.
Case history:
Crest audio launched LT series class-D which had cross-conduction/low reverse recovery current design using FREDs and series 300nH inductors, but the reliability went down considerably.Then they switched back to series schottky in CD series of amps. PL380 from QSC is another example of this.
I have seen people making claims that parallel low RDS mosfets and you can get away with body diode conduction[RDS shunting Body diode] but what about high power-high temperature operation where RDS starts increasing and it no longer shunts the body diode effectively[You don't have air condition fitted inside the amp, though the dissipation in class-D is much less but it happens when output power is at kilowatt levels.]??. Also by paralleling mosfets you again need a hefty gate drive for them, what about that?
Think about that.............rest, the choice is yours.
Every thing is fine when the ambient temperature is close to 25-35C, you have RDS to shunt the body diode and it works great, but in reality in high power amplifier, the temperature goes up and then the RDS shunting no longer helps you anymore[RDS wents up], more the body diode conducts more the Qrr charge/discharge takes place, more the Trr it takes, thus the reverse recovery current increases much more than expected, more EMI, less reliability. PRONE to failure in field.
The only proven and bruteforce solution which gives you peace of mind is series schottky with anti parallel high speed but soft recovery diode which must be rated for 125C for its specs, not to get fooled by some datasheets showing specs with 25C, 20V, 30A etc like deceptions. Look for 125C, 400V, 30A values and associated di/dt and Trr results along with it.
Yes the dissipation rises because of series schottky but not much, but in reality what you get is a reliable product. Free from ill-effects of temperature rise.
Case history:
Crest audio launched LT series class-D which had cross-conduction/low reverse recovery current design using FREDs and series 300nH inductors, but the reliability went down considerably.Then they switched back to series schottky in CD series of amps. PL380 from QSC is another example of this.
I have seen people making claims that parallel low RDS mosfets and you can get away with body diode conduction[RDS shunting Body diode] but what about high power-high temperature operation where RDS starts increasing and it no longer shunts the body diode effectively[You don't have air condition fitted inside the amp, though the dissipation in class-D is much less but it happens when output power is at kilowatt levels.]??. Also by paralleling mosfets you again need a hefty gate drive for them, what about that?
Think about that.............rest, the choice is yours.
Last edited:
I have found that you don't need the series Schottky + Inverse parallel fast diode. Simply using a suitably rated Schottky connected in inverse parallel between drain and source of each MOSFET does just as well, with much lower losses. A high-speed hall-effect current probe feeding a 'scope shows that this works effectively to bypass the current away from the internal MOSFET diode.
The reliability of my VA/PA amplifier designs, when operating under short-circuit conditions, improved enormously when I added the inverse Schottky. (It's a fact of life that 70/100V output commercial amplifiers used for Public Address and Voice Alarm installations all too often find themselves working into short circuits!)
The reliability of my VA/PA amplifier designs, when operating under short-circuit conditions, improved enormously when I added the inverse Schottky. (It's a fact of life that 70/100V output commercial amplifiers used for Public Address and Voice Alarm installations all too often find themselves working into short circuits!)
Only possible if the Schottky Vf is less than Vf of body diode at a given freewheeling current and temperature. No matter Schottky's are available upto 250V as single diodes, one such is available from OnSemi but the VF at rated current is already close to 0.6V which is similar to VF of body diode in similar case mosfet.
Last edited:
Very true, but on my boards both the MOSFET and the Schottky are surface-mount parts only a few mm from each other. The Schottky heats up as the MOSFET heats up so the forward voltage of the Schottky drops as the temperature increases.
It's true that as my amps are designed to feed 100V rms directly from the amplifier at only 3 or 4 amps maximum, the freewheeling current is not very high.
It's true that as my amps are designed to feed 100V rms directly from the amplifier at only 3 or 4 amps maximum, the freewheeling current is not very high.
How unpleasant. In case of failure, I cannot blame anybody than me!
I fully agree that shunting the diodes by the N-Chanel will not work (especially at high temperatures), except you use unsaint low Rdson in relation to the concerned currents.
Furtheron I decided to double check simulation results of switching losses with manual calculation of switching losses (instead of praying...).
Result: Simulation contains Qrr, but it seems like the increase of Qrr with di/dt is modeled in a way that shows systematical lower values than the data sheet. Also sim temperature seems to have almost no influence below 100C, but extremely catastrophic impact already 125C - not realistic.
Manual calculation for 170V/50A/320kHz/125C in hardswitching situation tells me turn ON losses of 462W and turn OFF losses around 82W. Total 544W.
Sim would indicate total losses of 390W at 100C.
SeeingThat's the typical situation of 'it can work' and 'completely unexplainable defects'
Not my cup of tea. Especially when considering a realistic case temp of the MosFet of 75C (instead 25C) the allowed losses of the IRFP4668 are not 520W anymore, but 345W.
@Ouroboros:
Yes, I also think that a shottky can help in case of lower voltages, on the other hand for lower voltages in the mean time there are MosFets with pretty good body diodes up to 100V and acceptable body diodes up to 150V.
As you and workhorse already indicated, at 200V and current levels around 50A todays shottkys won't save my world
To bad. Time schedule killed. Originally I intended to set up the layout in a rush and have a PCB in hands during winter holidays. Forget that.
Synergies between the power stages of SystemD_MD and SystemD_2kW will be low, some essential parts of the 2kW power stage have to be designed from scratch.
Not sure if I should be sad or happy.
Most likely happy. Designing from scratch is more hobby fun than doing a layout in a rush.
fun ahead Many thanks to all of you for your comments and special thanks to Tom for critically asking if I really want to rely on the body diodes!
I agree the simplistic approach would be to brave with respect to todays components.
"P.S. To bad that you stopped your SOCD." -> Very little time and nobody there to help with layout software...
BTW.: From this project, I still have some unused 4668 at home. Choco, my contribution to your project could be that I could send them over to you for free if you want. Also, I think I can help you getting some schottkies at no cost. The IRS2011S would also be found in my desk's drawer...
BTW.: From this project, I still have some unused 4668 at home. Choco, my contribution to your project could be that I could send them over to you for free if you want. Also, I think I can help you getting some schottkies at no cost. The IRS2011S would also be found in my desk's drawer...
You mean the cascode which is often also called emitter switched BJT?
I remember ST has a strong portfolio where they have integrated both in one device.
Could be an option, but still you will have to provide the large base current consumption of the BJT - possible, but ugly on the high side.
And live with a higher overall voltage drop - not the typical solution for 200V.
IMHO more typical for >500V.
Nevertheless I will have a look to that, thanks for reminding me.
Did not consider these devices for classD up to now.
I remember ST has a strong portfolio where they have integrated both in one device.
Could be an option, but still you will have to provide the large base current consumption of the BJT - possible, but ugly on the high side.
And live with a higher overall voltage drop - not the typical solution for 200V.
IMHO more typical for >500V.
Nevertheless I will have a look to that, thanks for reminding me.
Did not consider these devices for classD up to now.
" ....simply accepted that the speaker terminals are at +78V DC vs GND."
My intended PSU is single +178VDC (actually 3 PSUs of 59VDC secondary in series). The secondary side of the PSUs is isolated against PE (500V isolation), so I could just build a high-ohmic voltage divider to get a center voltage of +88V5 and connect this center point to PE -> speaker terminals free of DC. Not really an issue.
My intended PSU is single +178VDC (actually 3 PSUs of 59VDC secondary in series). The secondary side of the PSUs is isolated against PE (500V isolation), so I could just build a high-ohmic voltage divider to get a center voltage of +88V5 and connect this center point to PE -> speaker terminals free of DC. Not really an issue.
@Tom
That's a great offer!
The schottky's I could definitely need. Also freewheeling diodes.
IRFP4668: I have 14pcs, hoping that I will not fry more, but in case the project gets popular and enthusiast who want to build, then there will for sure happen a shortage in this forum...
IRS2011: I think I will stick with the IRS20957, accepted this chip as a friend during designing SystemD_MD and still have 16pcs available.
Let's check for further details by PM.
That's a great offer!
The schottky's I could definitely need. Also freewheeling diodes.
IRFP4668: I have 14pcs, hoping that I will not fry more, but in case the project gets popular and enthusiast who want to build, then there will for sure happen a shortage in this forum...
IRS2011: I think I will stick with the IRS20957, accepted this chip as a friend during designing SystemD_MD and still have 16pcs available.
Let's check for further details by PM.
I have not digged for best freewheeling diodes so far.
V30200C is giving me sorrows regarding dv/dt rating. 10kV/us.
Already in SystemD_MD there are dv/dt slightly above that under hard switching close to current protection.
And this typically goes up with increasing currents....
The fat series shottkys are cool. 60CTTN015.
IMHO, the massive capacitance will not be a problem in the planned application.
Any real life experience on both topics above?
V30200C is giving me sorrows regarding dv/dt rating. 10kV/us.
Already in SystemD_MD there are dv/dt slightly above that under hard switching close to current protection.
And this typically goes up with increasing currents....
The fat series shottkys are cool. 60CTTN015.
IMHO, the massive capacitance will not be a problem in the planned application.
Any real life experience on both topics above?
We use the V30200C as a rectification diode for a forward converter. I do not think dv/dt in a diode is critical, they do not tend to fail due to excessive dv/dt - there is simply no mechanism that causes a diode to fail of dv/dt. Diodes fail of thermal runaway or too high cathode voltage 99% of the time. Most times it is the MOSFETs failing due to to high dv/dt (parasitic BJT starts to conduct or gate gets charged via miller cap). Still, I agree, that a reliable design should not exceed specified max rating - true.
To me for freewheeling this one is also looking promising,
but max. dv/dt is not specified at all...
HFB35HB20
https://ec.irf.com/v6/en/US/adirect/ir?cmd=catProductDetailFrame&productID=HFB35HB20
but max. dv/dt is not specified at all...
HFB35HB20
https://ec.irf.com/v6/en/US/adirect/ir?cmd=catProductDetailFrame&productID=HFB35HB20