Eva
The bump at around Vgs-th.
I've observed that, but I don't really understand what it indicates: a shoot-through di/dt, a miller charge or a body diode reverse recovery (probably not)?
Shouldn't it be similar bump in turn-off waveform?
The bump at around Vgs-th.
I've observed that, but I don't really understand what it indicates: a shoot-through di/dt, a miller charge or a body diode reverse recovery (probably not)?
Shouldn't it be similar bump in turn-off waveform?
This bump tells that current is starting to flow through the MOSFET. It reflects voltage drop across source lead and bonding wire inductance (plus PCB layout contribution, if any, depending on where the measurement is taken).
For a TO-220, source inductance is around 8nH, so a 2V "bump" tells you that current is rising at 250A/us, this is called "di/dt" (2*1e-6/8e-9=250).
A negative bump tells either that positive current is stopping to flow or that negative current is starting to flow.
When one MOSFET turns off abruptly to reduce switching losses, and the opposite one has to clamp the switching node, you may see spikes as big as 12V on the latter reflectling current slopes in the range of 1500A/us. This may easily damage IR gate drivers on a bad layout.
Everything happening at source inductance appears at the gate too, but this may be actually useful because parasitic resonances may be damped indirectly to some extent with very low dissipation through the gate, and damping the rest directly on the switching node becomes much easier.
BTW: You can make very reliable and EMI-quiet class D amplifiers and SMPS if you learn to analyse and control di/dt (also dv/dt), and to keep di/dt confined in small local loops.
For a TO-220, source inductance is around 8nH, so a 2V "bump" tells you that current is rising at 250A/us, this is called "di/dt" (2*1e-6/8e-9=250).
A negative bump tells either that positive current is stopping to flow or that negative current is starting to flow.
When one MOSFET turns off abruptly to reduce switching losses, and the opposite one has to clamp the switching node, you may see spikes as big as 12V on the latter reflectling current slopes in the range of 1500A/us. This may easily damage IR gate drivers on a bad layout.
Everything happening at source inductance appears at the gate too, but this may be actually useful because parasitic resonances may be damped indirectly to some extent with very low dissipation through the gate, and damping the rest directly on the switching node becomes much easier.
BTW: You can make very reliable and EMI-quiet class D amplifiers and SMPS if you learn to analyse and control di/dt (also dv/dt), and to keep di/dt confined in small local loops.
Hi, EVA,
Could you tell more about this? I saw some designs use quite slow mosfet (like IRFP260N) and put a single fast small dioda accross D-S.
What about "you don't have to avoid body diode conduction completely either"?
Could you tell more about this? I saw some designs use quite slow mosfet (like IRFP260N) and put a single fast small dioda accross D-S.
What about "you don't have to avoid body diode conduction completely either"?
Some designers just copy stuff believing that it is useful but without understanding it or checking that it works. This is quite common on class D (that's why I don't like to release both schematics and PCBs for the same project).
Every time you see these diodes, you can make some funny assumptions about the designer... 😀
You can probably get a few amperes of body diode conduction without too much losses out of these slow mosfet. Note that maximum output current only happens during very brief periods when playing music. Also, stored charge is much lower for low currents (but it increases quicly, non linearly), datasheet rating is usually at maximum drain current.
One of my designs passes 6A to 16A through IRFB4227 body diode (depending on temperature) without any havoc, but turn on losses (and di/dt spikes) increase quickly if you go further (particularly at 80ºC just below amplifier thermal cutoff). Stored charge may be 30% to 50% higher at this temperature w.r.t. 25ºC. But only lower load impedances and signal peaks require body diode conduction, so average losses are ok.
Every time you see these diodes, you can make some funny assumptions about the designer... 😀
You can probably get a few amperes of body diode conduction without too much losses out of these slow mosfet. Note that maximum output current only happens during very brief periods when playing music. Also, stored charge is much lower for low currents (but it increases quicly, non linearly), datasheet rating is usually at maximum drain current.
One of my designs passes 6A to 16A through IRFB4227 body diode (depending on temperature) without any havoc, but turn on losses (and di/dt spikes) increase quickly if you go further (particularly at 80ºC just below amplifier thermal cutoff). Stored charge may be 30% to 50% higher at this temperature w.r.t. 25ºC. But only lower load impedances and signal peaks require body diode conduction, so average losses are ok.
EVA,
Thanks for the explenation. For a not-so-demanding projects, do you think we can use plain cheap mosfets like IRFP260N for classD? I mean plain without any external series-parrarel dioda.
How much max switching frequency for this mosfet? Do you think we can use it for about 400khz?
Thanks for the explenation. For a not-so-demanding projects, do you think we can use plain cheap mosfets like IRFP260N for classD? I mean plain without any external series-parrarel dioda.
How much max switching frequency for this mosfet? Do you think we can use it for about 400khz?
According to the datasheet, if they are kept cool (like 50ºC max.) and almost no dead time (like 10ns max.) is used, body diode conduction is avoided up to 15A (for 1A or so in body diode). You can probably go slightly above that without much problem, like 3A..5A more.
When body diode conduction is allowed, the maximum dv/dt on Vd-s has to be kept below 5V/ns. For example, this implies turn-on crossover times limited to 30ns or higher for +/-75V. The MOSFET may latch-up in an ON state (regardless of gate drive) if it's subject to higher Vd-s slopes in the next 100s of ns just after body diode recovery.
The question is: Is it worth the effort? Look for better for sources of more suitable stuff.
btw: It's funny to see other people making questions whose replies are useful to the initial poster, who should be the one asking but he doesn't...
When body diode conduction is allowed, the maximum dv/dt on Vd-s has to be kept below 5V/ns. For example, this implies turn-on crossover times limited to 30ns or higher for +/-75V. The MOSFET may latch-up in an ON state (regardless of gate drive) if it's subject to higher Vd-s slopes in the next 100s of ns just after body diode recovery.
The question is: Is it worth the effort? Look for better for sources of more suitable stuff.
btw: It's funny to see other people making questions whose replies are useful to the initial poster, who should be the one asking but he doesn't...
Back
I'm back from vacancy,also the output mos-fets have arrived too!
ARE these GOOD?
I'm back from vacancy,also the output mos-fets have arrived too!
ARE these GOOD?
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
Yes, these are among the best currently available, four pairs of them in a full bridge will do 3kw on 4 ohm and 4kw to 6kw (or even more) into 2 ohm (I'm still trying to find the limit), but only if you use them properly, otherwise they will do nice fireworks 😉
Suppose 200V isn't enough. Are the FDP51N25's decent for switching losses? They *look*like a better deal than putting a half dozen or more older 15A units in parallel.
wg_ski said:
Your tranistor looks OK at first seen (Gate Charge 55nC) but it have a Reverse Recovery Charge (Qrr) of 4uC, wich is too big.
Am I right?
luka said:
So till now, we must understand the BEST Class D mos-fets are those IRFB4227?
Or is anything else better for class D application?
well not THE BEST, but yea, for high power amps, this ones seem to be best yes... if you would use lower voltage, up to +/-45v you have a lot better fets, made for ClassD amps.. IR has them..
FDP51N25 datasheet says 4uC body diode and 5V/ns recovery... That's not particularly good. See IRFB4229.
High gate charge is not a problem nowadays, we have good buffer transistors like ZXTP23015CFH and ZXTN23015CFH. These can do 10A peak with a hFE of 100, in SOT-23 😀 I prefer local buffering rather than very high current driver ICs far from the gates.
High gate charge is not a problem nowadays, we have good buffer transistors like ZXTP23015CFH and ZXTN23015CFH. These can do 10A peak with a hFE of 100, in SOT-23 😀 I prefer local buffering rather than very high current driver ICs far from the gates.
I agree with this. Sometime PCB track distance can be big problem, like if you build full bridge classD with HIP4080, the IC is full bridge, but how can we put 4 quadrant mosfets (which should be attached at the edge of the PCB to the heatsink) very near to the IC pins?
lumanauw said:
You can't unless the packaging is small enough to allow it. Which limits how big an amp you can build and still maintain quality. I think that's the whole point behind some of those new flip-chip packages. But for hand assembling PCB's I'm not sure that's too viable.
With TO-220's, especially multiples and 50+ amp peak currents, it doesn't seem to make sense to even try to use single-chip drivers anymore. 6N137-->HCMOS driver-->NPN/PNP pairs (separate for each paralleled FET, right on the gate) on each quadrant looks like it could produce a workable layout for something pretty darn high powered.
Offer
I'm offering 5 pairs (10 pieces) of IRFB4227 in trade for some Micrometals Cores for output coils.
I'm intending around 30-40A saturation, even more if is possible.
Waiting for offers.
Picrures of cores are welcomed!
Also Pafi, I'm back,now we can do trade,that we have spoken.
I'm offering 5 pairs (10 pieces) of IRFB4227 in trade for some Micrometals Cores for output coils.
I'm intending around 30-40A saturation, even more if is possible.
Waiting for offers.
Picrures of cores are welcomed!
Also Pafi, I'm back,now we can do trade,that we have spoken.
- Status
- Not open for further replies.