6000W By IRS2092

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
There is no way in which this circuit can produce 6000W, at least not without bridging two channels with +/-90V rails on 2 ohm (sagging to +/-80V under load, which definitely requires a regulated SMPS or far more than 3000uF per rail).

This involves 80A peak and switching such a high current reliably is not for beginners, not to mention that a schematic without a working PCB layout (or the experience required to do it) is useless.

I have reasons to think that this schematic is not based in any real working thing, and I say this from the point of view of someone that has a few working prototypes capable of 5000W bursts on 2 ohms on her workbench.

Why supply capacitors are 200V? They should be 100V. 4.7 ohm is too low as a turn on gate resistor for that MOSFET and too high as an efficient turn off resistor. Big thru-hole films on supply rails are useless, they result in resonances. Etc...
 
Last edited:
There is no way in which this circuit can produce 6000W, at least not without bridging two channels with +/-90V rails on 2 ohm (sagging to +/-80V under load, which definitely requires a regulated SMPS or far more than 3000uF per rail).

This involves 80A peak and switching such a high current reliably is not for beginners, not to mention that a schematic without a working PCB layout (or the experience required to do it) is useless.

I have reasons to think that this schematic is not based in any real working thing, and I say this from the point of view of someone that has a few working prototypes capable of 5000W bursts on 2 ohms on her workbench.

Why supply capacitors are 200V? They should be 100V. 4.7 ohm is too low as a turn on gate resistor for that MOSFET and too high as an efficient turn off resistor. Big thru-hole films on supply rails are useless, they result in resonances. Etc...

I wouldent really worry about this thread, the original poster hasn't even posted after the first post.! GooD info tho :)
 
I wouldent really worry about this thread, the original poster hasn't even posted after the first post.! GooD info tho :)

Good info is exactly right - even if the OP decides it's not worth the trouble. Some of us 'outsiders' are still interested in doing this sort of thing. Out of the confusion comes enough information to learn what topologies work, what the switch timing requirements are and how to model everything. And I'm not in a big hurry to throw something together without adequate understanding (and watch it explode).

PCB layout is the least of my worries - I already design multi-GHz high efficiency (including switch mode) MMIC power amplifiers for a living so I know a thing or two about distributed networks.
 
To start converting this layout into something capable of switching some current fast reliably I suggest:

- Add a continous ground plane that may be connected to the negative or the positive rails rather than ground. Single sided PCB won't be up to the job.
- Remove the 2.2k gate resistors, they are not required at all and make the layout more complex by taking very useful space.
- Group the MOSFET in close lo/hi side pairs.
- Add 100n SMD capacitors from low side MOSFET sources and high side drains to ground plane (or from low side sources to hi side drains if the plane is a rail).
- Add more 470u 100V capacitors. Capacitors tend to exhibit 1nH parasitic inductance per mm of lead spacing. Paralleling many divides inductance by the amount of them.
- Add LC filters to supply rail inputs to prevent the HF ripple due to high current switching from reaching wirings.
- IRFP260 can't go very far once the body diodes start conducting. Suitable MOSFET are the much newer IRFB4227 and the improved IRFB4127.
- TO-247 MOSFET are good when power dissipation is high (class A or AB), but class D is not about that, it's about low parasitics and fast switching. TO-220 does that much better. See lead inductance data on most IR datasheets.
- Consider replacing the two driver ICs by NPN and PNP SMD buffers near the gates, the inductance of the long gate track results in all MOSFET not switching at the same time. There are good high current low voltage bipolars in SOT-23. 3-4A peak are required for fast turn off.
- Add diodes from IR chip outputs to their rails to prevent latchup when the outputs are forced to exceed the rails.
- Add a diode from VS to COM to prevent the IR chip from failing due to excess negative spikes on VS. These diodes should be SMD to keep things small and low inductance.
- The parasitic inductance of the RC snubbers from the switching node to the rails matters a lot, the optimum value makes the snubber more efficient since it becomes a RLC resonator that can be tuned at the frequency where MOSFET capacitance and package and supply parasitic inductances resonate. Too high inductance and the snubbers will do nothing. Too low inductance makes them less efficient but it's not that bad.

With a highly optimized layout and supply capacitors arrangement I'm getting 36A peak per IRFB4227 reliably. Your layout would be nice as a class AB amplifier, but the usual visual symmetry is worth nothing for high speed switching. Anyway, I think you are going in the right direction by daring making a layout an publishing it to get some feedback (that's why I'm giving you these hints ;) )
 
Last edited:
But this is a commercial product :)
(Why so many fans? Is there really so much heat to get rid of? :) )
+1000
it seems that these guys from Thailand (Detex and now this Digitclass thing) have no idea how make switching efficient. And look at these SMPSs, so big so expensive for the claimed output power, where smarter/more efficient/cheaper design can do it... And I don't talk about the air flow, with fans pulsing air 90 degrees perpendicular of each other and with large heatsink blocking the flow :rolleyes:
 
... Anyway, I think you are going in the right direction by daring making a layout an publishing it to get some feedback (that's why I'm giving you these hints ;) )

Nice girl!
Many people here appreciate your support.
And it is nice to see that you are going on to give it, even if there are to many guys who are not like DJ_DUMY and who are just sucking information in order to copy and sell, without any own technical engagement.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.