Hello everyone,
I have started building an active subwoofer using a class-d amplifier. I have built amplifiers of similar power before but never given class-d a try. All other electronics parts are already built except the class-d output stage, and I came to the conclusion that it is best to ask for your comments before I start to burn up FETs and gate drivers
The amplifier is a self-oscillating one using IRS20955 gate driver and IRFB4227, IRFP4227 or IRFB4020 FETs (this is actually one of my questions, which FET would you recommend. I will use a 4 ohm speaker). Will this gate driver be able to drive these FETs and also at which switching frequency. In the simulations I have used about 350kHz.
I have done simulations and it seems to be working. There is one thing in the simulations that is quite concerning. If I use FET models that came with the LTSpice software there is no problem, but if I use FET models that I imported from the spice models found on IRF website you can see huge currents peaks going through them. Dead time is set to max for the IRS20955.
Do you guys think this is something that will be realized when I make the PCB or is it just a glitch in the simulation or in the FET spice models? I have attached a picture to illustrate the situation. There are three pictures with different zoom. Blue graph is the upper FET gate voltage, Red graph is low FET gate voltage and green is the drain current of the upper FET.
Attached is also a schematic used in the simulation model. Over current resistors and dead time resistors are not final. Any comments on this schematic are welcome.
Thanks in advance!
I have started building an active subwoofer using a class-d amplifier. I have built amplifiers of similar power before but never given class-d a try. All other electronics parts are already built except the class-d output stage, and I came to the conclusion that it is best to ask for your comments before I start to burn up FETs and gate drivers
The amplifier is a self-oscillating one using IRS20955 gate driver and IRFB4227, IRFP4227 or IRFB4020 FETs (this is actually one of my questions, which FET would you recommend. I will use a 4 ohm speaker). Will this gate driver be able to drive these FETs and also at which switching frequency. In the simulations I have used about 350kHz.
I have done simulations and it seems to be working. There is one thing in the simulations that is quite concerning. If I use FET models that came with the LTSpice software there is no problem, but if I use FET models that I imported from the spice models found on IRF website you can see huge currents peaks going through them. Dead time is set to max for the IRS20955.
Do you guys think this is something that will be realized when I make the PCB or is it just a glitch in the simulation or in the FET spice models? I have attached a picture to illustrate the situation. There are three pictures with different zoom. Blue graph is the upper FET gate voltage, Red graph is low FET gate voltage and green is the drain current of the upper FET.
Attached is also a schematic used in the simulation model. Over current resistors and dead time resistors are not final. Any comments on this schematic are welcome.
Thanks in advance!
Which FET is the best for class D you will find in other threads.
And the problem with the current peaks is definitely NOT because of the imported model. The problem is in the parameters of selected FET (you can compare simulations with imported and 'built in' models of some FET). The most like reason for these peaks is the slow body diode of the FET. And the max deadtime you selected does makes this problem even worse. To understand, why the body diode is a bad thing, you need to read other threads too.
BTW, I use LTSpice too, and I wonder, how precise the FETs are simulated here. For example, it is possible to see the typical gate charge curve during simulations, the Vds dependable output FET capacitance and so on.
Also, look in other threads about the preferred PCB layout for class D amps (groundplane, FET driver layout, etc). This can be in class D amp even more important than the schematics itself...
Last edited:
Very important parasitics, including MOSFET package inductances and power supply inductances and capacitances, are completely missing from the schematic (and they control di/dt and dv/dt slopes during switching).
So your simulation is completely useless for evaluating switching performance, it's only useful for control loop analysis, and for that purpose the IR chip and the MOSFET can be just replaced by ideal voltage controlled voltage sources and voltage limiter blocks resulting in faster simulation. But for that purpose, output filter parasitics and crosstalk from output inductor back to the comparator and input signal would have to be considered too...
So your simulation is completely useless for evaluating switching performance, it's only useful for control loop analysis, and for that purpose the IR chip and the MOSFET can be just replaced by ideal voltage controlled voltage sources and voltage limiter blocks resulting in faster simulation. But for that purpose, output filter parasitics and crosstalk from output inductor back to the comparator and input signal would have to be considered too...
Last edited:
500w class D
Your schematic has some serious errors.
You cannot drive the comparator chip's output directly into pin 3 of the IRS20955 as the 20955 will latch up. You should not use this chip as it has been discontinued due to noise issues on pin 3. The IRS20957 has this problem solved but you must still use a series resistor between the output of the comparator and the IRS20957. Use about 330 ohm with a capacitor between pins 3 and 4 right at the chip's leads of 100pF.
The IRS chips CANNOT drive the gates of IRFB4227's as they have a higher gate charge than the IRFB4020.
You must use totem drivers with the 4227s and use low value gate resistors (less than 5 ohms) otherwise the 4227s get hot under idle.
1 up 1 down of the 4227s is OK for 500w at 4 ohm using a +/-75v regulated supply.
If you use IRFB4020 then use 3+3 devices and use a totme driver again but you can use 10-15 ohm gate resistors with these.
Make sure that you follow good grounding and Kelvin rules in your layout
I always insert a low value (10 ohm) resistor in series between pin 13 of the IRS20955 and the centee node of the upper and lower mosfets. Then from pin 13 to pin 9 of the IRS20957 insert a catch diode rated at 400v 25-35nsec Trr right at the IC. This will prevent destruction of the 20957 if there are some unwanted switching spikes due to "poor" layout.
The PCB trace from pin 13, through the 10 ohm and to the centre node must use thick traces (3mm) and be as straight and direct as possible.
We use thousands of these chips in a design I did and I work closely with IR in Los Angeles on these issues
Steve Mantz
Zed Audio
Your schematic has some serious errors.
You cannot drive the comparator chip's output directly into pin 3 of the IRS20955 as the 20955 will latch up. You should not use this chip as it has been discontinued due to noise issues on pin 3. The IRS20957 has this problem solved but you must still use a series resistor between the output of the comparator and the IRS20957. Use about 330 ohm with a capacitor between pins 3 and 4 right at the chip's leads of 100pF.
The IRS chips CANNOT drive the gates of IRFB4227's as they have a higher gate charge than the IRFB4020.
You must use totem drivers with the 4227s and use low value gate resistors (less than 5 ohms) otherwise the 4227s get hot under idle.
1 up 1 down of the 4227s is OK for 500w at 4 ohm using a +/-75v regulated supply.
If you use IRFB4020 then use 3+3 devices and use a totme driver again but you can use 10-15 ohm gate resistors with these.
Make sure that you follow good grounding and Kelvin rules in your layout
I always insert a low value (10 ohm) resistor in series between pin 13 of the IRS20955 and the centee node of the upper and lower mosfets. Then from pin 13 to pin 9 of the IRS20957 insert a catch diode rated at 400v 25-35nsec Trr right at the IC. This will prevent destruction of the 20957 if there are some unwanted switching spikes due to "poor" layout.
The PCB trace from pin 13, through the 10 ohm and to the centre node must use thick traces (3mm) and be as straight and direct as possible.
We use thousands of these chips in a design I did and I work closely with IR in Los Angeles on these issues
Steve Mantz
Zed Audio
Why then rating this "1A" mosfet driver? Such driver should be able to drive much higher gate charge. The only drawback I see is heat dissipation limit on irs chip, ?!
1A peak driving current should be enough for up to 1Mhz
6A mosfet driver can drive 20nS 10 000pF load, than how this 1A driver cant driver 1000pF in 20nS
EDIT: Mosfet in first schematics have 4600pF it may be too slow, But I dont think totem driver can help(if we use "bigger" mosfets), I have never seen totem pole 20nS speed. ?
Can you suggest any faster totem pole driver?
I started to build UCD using irs20957 and irfi4020 mosfets: And i have some questions:
1) Why Csd cap is only 100nF in all this diy schematics, datasheet suggest min 5uF: Will bigger cap here, and slower start up of driver prevent self oscillation ?
2) Application note AN-1141 : "precharging with VBS zenner diode" whats the purpose ?: Bootstrap cap will be charged before irs turns on (before CSD cap charge up), after all Bootstrap cap has its oven power supply, do not understand need for this zenner paralell to bootstrap cap here?
3) Should I use schotkey diode parallel to gate resistor for faster discharging ? Official IRF design NEVER use such diode
4) For example LT1106 or TL3016 comparator : why all people use positive eletrode for audio and negative for feedback : why not using positive to GND and negative for sum: feedback+ audio. Such operation show less thd in lt spice .. Is there any drawback for such operation related to comparator itself, what is minimal voltage difference that comparator change output ?
Last edited:
what exact buffers do you use, and can I see screenshot of your output stage? Also how buffers you use increase timings ?
Thanks , very nice
what resistor do you use from totem driver to mosfet ?, I presume from totem driver to mosfet shotkey diode is not needed
Also what about bootstrap : now both irs chip and this driver have to have external bootstrap cap, right? or we can tie together totem driver Vcc pin and irs Vb pin ?
for gate resistor I used 10r and 1n4448 you don't need more
For bootstrap I use 47uF (upper one) for IRS, and something like that for buffer, which is SMD as is this capacitor, so yes, both have external cap. you could use only one from IRS, but this is like 0.5mm from the buffer, so that is the best I could do
(everything is charged from lower side)
For bootstrap I use 47uF (upper one) for IRS, and something like that for buffer, which is SMD as is this capacitor, so yes, both have external cap. you could use only one from IRS, but this is like 0.5mm from the buffer, so that is the best I could do
(everything is charged from lower side)
ok
What is optiaml totem driver to mosfet resistor ? now maybe something like 1-2 ohm would be optimal, cause it will increase speed of large gate charge mosfets
I wonder why people usually use High-low side mosfet driver(ir2110) for "big mosfets"(read high current big gate charge ones), instead this driver + totem pole ?, This seems much better, irs20957 has nice protection logic
Usually all newer power mosfets have less than 10nF gate
This way there is practicly no limit to gate charge when we have 9A mosfet driver, right?!
Last edited:
why? coz they know it, and/or can only get them... but even 2110 is getting old now
and also fets are, if you get them, 10000x better then what most people use, those big irfp250's or such
well if you look at when you get 9A, you will see that won't happen, unless you will put 1A into driver itself... while you can get over 2A with just 10ma!
and also fets are, if you get them, 10000x better then what most people use, those big irfp250's or such
well if you look at when you get 9A, you will see that won't happen, unless you will put 1A into driver itself... while you can get over 2A with just 10ma!
Yes, you are right
But you get 5A with 100mA imput, graph is exponential (in datasheet)
2A with 10mA
9A with 1A
i had to order irs chip from "farnell" and wait them for 1 month to come from USA
I ordered 6 pcd, I hope i Will not burn them all before I make it work, soon Ill post schematics related to irs20957 driver ad mosfets for inspection (just to be safe)
One thing that I am concerned most will comparator work if positve electrode is to gnd and negative is audio + feedback (never seen practical design with suck setup) , but spice show less thd..
I use the irs2092 with a single pair of 4227's and have no problems.
The 2092 gets a little warm but nothing to worry about.
INTERNATIONAL RECTIFIER|IRFB4020PBF|MOSFET, N, 200V, TO-220AB | Farnell United Kingdom
What about irfb4020, double the irfi4020 current rating and gate charge is the same, also power dissipation is much higher?
I wonder what is maximum power rating with half bridge with 18A mosfet ?
lets say we go on safe side, i think this should work up to 500W 4 ohm ?
My goal is to make 4-8 ohm ampplifier, cause ucd has excellent filter control, and exact load impedance does not matter:
+/-70V and irfb4020 vs irfi4020 , latter one may not work above 250W 4 ohm cause too low current rating (official irf paper)
so far irfb4020 without totem pole seems the simplest solution .
What about irfb4020, double the irfi4020 current rating and gate charge is the same, also power dissipation is much higher?
I wonder what is maximum power rating with half bridge with 18A mosfet ?
lets say we go on safe side, i think this should work up to 500W 4 ohm ?
My goal is to make 4-8 ohm ampplifier, cause ucd has excellent filter control, and exact load impedance does not matter:
+/-70V and irfb4020 vs irfi4020 , latter one may not work above 250W 4 ohm cause too low current rating (official irf paper)
so far irfb4020 without totem pole seems the simplest solution .
Last edited:
If I am correct, , 500W around 15A, and 25W dissipation caused by Rds (on),
could work with good cooler, but, what about paralleling two mosfets in half bridge ? can that work? half bridge with 4 mosfets than only 5W per mosfet. Usually mosfets can be paralleled cause positive temperature coefficient without any resistors
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.