IRS2092 + TC4420 + IRFB4227 (3x) DeadTime Problem

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi all,

I'm working on a Class-D Amplifier with 1kW@2Ohm. I first started with the reference design of the IRAUDAMP9 (BJT Totem Pole Drivers) and 3 IRFB4227 in parallel. Later I changed the transistor drivers to the TC4420 driver IC.
PWM Frequency is 200 kHz and Rail-Voltages are ±75VDC.
I'm using max dead time (105ns) but I got some Cross Conductions (shoot through) and the MOSFETs get quiet warm. My output filter is a SAGAMI 7G31A 22uH with a 2.2uF 275V capacitor. I carefully read the Application Note AN-1138 and the PCB-Reference Guide AN-1135 but I don't understand what the problem is.

When I measure the Output Signals of the IRS2092 (HO, LO) I get those Signals:
Picture "DeadTimeOK" = 105ns Dead Time and no Cross Conduction
Picture "DeadTimeProblem" = Falling Edge of High Side is too slow. It seems, the Signal wants to go low, but somehow it is pulled high for 200ns.
Yellow is High Side on Pin 14, blue is Low Side on Pin 11, Reference is GND on Pin 2, 2 Ohm Output Load connected.

I've already added an additional Load (1k Ohm and lower) between HO and VS, but no difference.

Does someone know what the problem could be? Thanks you for the answers.

I attached the Schematic, PCBs and Photos of the Circuit…

Best Regards Florian
 

Attachments

  • DeadTimeOK.jpg
    DeadTimeOK.jpg
    456.6 KB · Views: 799
  • DeadTimeProblem.jpg
    DeadTimeProblem.jpg
    469.2 KB · Views: 949
  • Driver V3.0A.JPG
    Driver V3.0A.JPG
    273.4 KB · Views: 1,745
  • Mainboard.jpg
    Mainboard.jpg
    544.6 KB · Views: 993
  • Driver V3.0A PCB.JPG
    Driver V3.0A PCB.JPG
    291.8 KB · Views: 889
  • IMG_5122.jpg
    IMG_5122.jpg
    528 KB · Views: 546
  • IMG_5123.jpg
    IMG_5123.jpg
    570.3 KB · Views: 456
  • IMG_5124.jpg
    IMG_5124.jpg
    482.1 KB · Views: 417
Hi all,

I'm working on a Class-D Amplifier with 1kW@2Ohm. I first started with the reference design of the IRAUDAMP9 (BJT Totem Pole Drivers) and 3 IRFB4227 in parallel. Later I changed the transistor drivers to the TC4420 driver IC.
PWM Frequency is 200 kHz and Rail-Voltages are ±75VDC.
I'm using max dead time (105ns) but I got some Cross Conductions (shoot through) and the MOSFETs get quiet warm. My output filter is a SAGAMI 7G31A 22uH with a 2.2uF 275V capacitor. I carefully read the Application Note AN-1138 and the PCB-Reference Guide AN-1135 but I don't understand what the problem is.

When I measure the Output Signals of the IRS2092 (HO, LO) I get those Signals:
Picture "DeadTimeOK" = 105ns Dead Time and no Cross Conduction
Picture "DeadTimeProblem" = Falling Edge of High Side is too slow. It seems, the Signal wants to go low, but somehow it is pulled high for 200ns.
Yellow is High Side on Pin 14, blue is Low Side on Pin 11, Reference is GND on Pin 2, 2 Ohm Output Load connected.

I've already added an additional Load (1k Ohm and lower) between HO and VS, but no difference.

Does someone know what the problem could be? Thanks you for the answers.

I attached the Schematic, PCBs and Photos of the Circuit…

Best Regards Florian

Sorry I attached the wrong schematic...
Here is the new one...
 

Attachments

  • Driver V3.0A.JPG
    Driver V3.0A.JPG
    223.3 KB · Views: 675
Administrator
Joined 2007
Paid Member
妳線路不太正確, 正負電壓應用在半橋系統中, 有一個非常重要的觀念, 就是相對地的工作電原, 您-75VDC FORGET VCC+15VDC, 所以VCC PIN 看不到HI

:cop:
English only please.

‎Your line is not quite correct, positive and negative voltage half-bridge system, there is a very important concept, is relatively power supply, -75VDC FORGET you VCC+15VDC, so the VCC PIN to see HI ‎
 
If iI interpret the sentence right it means that the driver supply would be done wrong. Wehn I have a look at the circuit posted then this should be fine (-> UG).

What I don't like is the layout: This would look very cool for a conventional (i.e. linear) amp. But for a class-d amp there is too much trace area and length where the switching signal is present. Also the leads of the output devices are too long.
Why didn't you just follow the IRAUDAMP9 layout ?

Regards

Charles
 
D

Deleted member 148505

With that kind of heatsink, only a pair of irfb4227 will do 1kw @ 2ohms if cross conduction is minimized (at 1kw output).

If you want to attain continuous 1kw (FTC testing), the inductor you are using (7G31A) will be the bottleneck.

Try to run it with only 1 pair to see if the waveforms will get better. Ringing is bad, at 1kw output, the waveforms will get worse so your current implementation definitely needs better pcb layout.
 
I agree with jlester87...try running it with only one pair of output devices and see if the problem still persists. Your schematic shows IRFB4227 but you are actually using 4127 which is a little harder to drive. Is there a reason that you switched from the BJT totem pole? Did you have a sample built with the totem pole buffer that was working correctly? Finally, what is the purpose of the resistors R7 and R10 tying the high voltage rails to the IRS2092?
 
I agree, layout is very bad, and 3 FET is too much.
I add: lower inductance would be better.
I don't really see cross conduction. From these waveforms some asymmetry can be seen, but to know if there is cross conduction at least the gate waveform should be seen, with low side drain voltage.
The "2 ohms load" info is meaningful only together with input or output signal. The most relevant info would be the output current.
 
Thanks you for all the answers!

Yes, UG is 12V referenced to -75V, so all the supply voltages are fine.

ungie found the problem! I order accidentally the wrong MOSFETs! After I replaced them all with IRFB4227, everything works fine and the MOSFETs have enough dead time, so they don't heat up.

As I started building my first Class D Amp I didn't know that the Layout is that important. I just copied the schematic from the IRAUDAMP9 and designed my own PCB. The result hasn't worked at all. I thought the problem were the BJT totem pole. I have read in other threads that some people are using the TC4420 and are very happy with it. So I changed them in my new design. It would probably work also with the BJT (ZXTN25100BFH & ZXTP25100BFH).

R7 and R10 exist also in the IRAUDAMP9 schematic (R50 & R52 on Page 29). I don't think they are realy necessary but i let them in my design.

Yes the inductor 7G31A gets realy hot (65°C+). Is there an alternative? Is T106-2 better? Maybe someone can give me a link to buy them from ebay?
 
I know, that the Layout is still not realy good. But I couldn't change the position of the heatsinks. So I tried to design the PCB as good as I can, for those MOSFET positions.
Now everything works fine and i get over 1kW RMS at 1kHz.

Maybe in the future I will completly redesign the amp and place all the MOSFETs very close together on one heatsink.
 
There is still some small room for improvement regarding the pin lengths: If you mount aluminium or copper bars where the output devices now are, then you can screw the output devices to these bars in an upright position.
As someone alread mentioned - these huge heatsinks are not really necessary.

Regards

Charles
 
D

Deleted member 148505

You need better output filter design if you only run it on 200khz. So it is better to just use a beefier gate buffer and "gate driver supply" and run the module at higher frequency.

There's diminishing returns when you add pairs of mosfet (more mosfet, more problems)

If you really want 3 pairs and operating the module at 300khz or higher, ZXTN25100BFH & ZXTP25100BFH are just weak for driving 3 pairs irfb4227 specially if copper plane in the PCB is not enough for heat dissipation. When the module is inside a chassis, everything will get hot compared to open air ambient temp. (unless there's active fan cooling)

If you want the ZXTN25100BFH & ZXTP25100BFH to survive, there must be a large copper plane for each transistor. Or try the buffers used by Markus on this thread. (SystemD_2kW)

http://www.diyaudio.com/forums/class-d/224052-systemd_2kw-any-interest-open-design.html
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.