Reverse diodes in full bridge

Status
Not open for further replies.
Jump to Latest
C

ChocoHolic

Member
Joined 2003
#1
Hi Class D DIYers !

Some months back, when I was doing my first thought’s regarding my class D sub woofer amp, I did not think too much about the reverse diodes. But I had chosen to go for some nice IRFB52ND15 MosFets. They offer a low gate charge of 60nC, while being capable of 150V / 60A and offering a max Rdson of 32mOhms.
OK, now I am one step closer…
http://www.diyaudio.com/forums/showthread.php?s=&threadid=55880
… and during my last measurements I found, that my power requirements are much less. A full bridge coming from +/- 40V (or just 80V, still under consideration) and being able to deliver 400W is sufficient.
So the IRFB52ND15 are a little bit oversized, but I have them already with me and will use them.
Currently I am thinking about external reverse diodes, -or not.
When I look at the complex load of the filter and sub driver, then I come to the conclusion that there is a high portion of switching events which will not force the half bridge to start sloping during the dead time, but will stick to the rail while drawing the load current through the reverse diodes until the opposite MosFets are turned on. In this case the switches, which are turned on will have to discharge the conductive reverse diodes. Resulting in some current peak and some losses and some EMI-radiation.
I was thinking about some decent fast schottky diodes in parallel to the intrinsic body diodes.
I.e. 30CPQ100 from International Rectifier.
But I know that some colleagues had bad experiences with the reliability of schottky diodes in
a SMPS, even without coming close to the limits of the data sheet… They never found the reason for the defects. Also the additional schottkys would add some additional undesired capacitance to the bridge…. and costs…. and assembly efforts…
….also additional solder joints, each itself causing a potential drop of reliability….
Furtheron the reverse diodes of the IRFB52ND15 seem to be quite good:
Reverse recovery time 140ns and reverse recovery charge around 800nC at 36 A & 100A/us.
Also my switching frequency will be only 80kHz (Please note: It’s for a sub woofer.)
In short form: I am tending to do it the simple way. The gate turn off could be done with the max speed which the driver can deliver. Gate turn on could be softened by a gate series resistor around 47 Ohms, this should limit the du/dt around 2V/ns, well below the allowed value of 5.5V/ns for the diode recovery.

Reverse diodes are often discussed and from my perception, the best trade off seems to depend on the specific load, the switching frequency and the chosen MosFets.
Looking forward to get your views!
Markus
 
#2
Full bridges are a way to go to much higher powers before you need to add external diodes. Fairchild are a good place to start looking for FETs with low charge diodes. Do check out all data sheets because they have old and new types in the portfolio. I can't exactly remember the type but they have a 38A (or so) 150V fet with <200nC body diode.

If you do choose to add external diodes, don't forget to add a schottky diode in series with the MOSFET to insure it'll never conduct reverse. Otherwise the parasitic inductances will keep the external reverse diode from catching the current when the FET is turned off, even if it has vastly lower forward drop than the FET's internal diode.
 
#3
Hi Bruno,
great input with the additional series diode to the FET.
...had this discussion just some hours back, but we did not come to a clear conclusion, if the low forward voltage of a schottky diode alone would be sufficient to avoid current through the body diode, or if an additional diode in series to the FET in order to block reverse direction would be necessary....
So your statement is clear. A schottky in parallel without series diode will not be sufficient. Thanks.
Your comments do encourage me even more to go for the simple solution and use the body diodes on purpose.
So you would propose to forget my fat 60A devices and to look for some less heavy types, in order to get body diodes with less charge?
Thanks for the hint with fairchild.

...hm, up to 5 minutes back I loved my IRFB52ND15... :bawling: 😉

BTW:
My reasons for the full bridge were:
a) 100V voltage rating of my planned drivers (LM5104).
b) Improved supply voltage rejection, compared to half bridge.

Thanks
Markus
 
#4
You shouldn't make the FETs much "fatter" than necessary to deliver the required power into the required load. If you insure that rated current at Tc=100deg > Imax, you're fine.

And you've got an extra reason for using a full bridge, namely that you can get to higher power levels without having to bother about reverse recovery.
 
#7
IRF540Z

Hi Bruno / IVX,
I searched for nice MosFets.
...and found several nice types at Fairchild...
100V/30-40A/40mOhms/Qrr around 200nC

BUT the most outstanding device I found at IR:
IRF540Z please note, the Z is essential !
Without Z you will get a well known oldy.

The Z-type seems to be a completely different device.
100V/36A (at Tc=100°C still 25A)
Rdson= 27 mOhms
Total gate charge: typ. 42 nC
Reverse Recovery Charge: only typ. 41nC

Looks lovely, but seems to be hard to get.....

Bye
Markus
 
#10
If it is a "typo".. ==> 😡

...otherwise great!
I will try to get some.
The data sheet looks like the perfect match for my application.
May be we can share experiences in some months.
:yes:

IVX:
Sounds like you will be the first to play around with this device..
Right? 😀 😎
 
#11
There's also seen from datasheets that junction-to-case thermal resistance and power dissipation for IRF540Z are about 1,5 times worse comparing with conventional IRF540.
This says that they managed to develop new manufacturing technology when die size is much smaller for same current handling and - amazing! - even lower channel on-resistance.
So smaller die size helps to explain such low gate and diode recovery charges :smash:
 
#12
...and IR seems to be able to do this reasonable manufacturing efforts !
I can get this device in a standard electronic shop for EUR 1,10.
First it looked hard to get them and I never expected to find it at
CONRAD.DE, but they offer it on their web page !
Do we live in heaven???!
:nod:
 
#14
Bruno Putzeys said:
Full bridges are a way to go to much higher powers before you need to add external diodes. Fairchild are a good place to start looking for FETs with low charge diodes. Do check out all data sheets because they have old and new types in the portfolio. I can't exactly remember the type but they have a 38A (or so) 150V fet with <200nC body diode.

If you do choose to add external diodes, don't forget to add a schottky diode in series with the MOSFET to insure it'll never conduct reverse. Otherwise the parasitic inductances will keep the external reverse diode from catching the current when the FET is turned off, even if it has vastly lower forward drop than the FET's internal diode.
Click to expand...

That seems like the measure one would want to avoid at all costs, due to a fairly substantial hit on efficiency. Could there be any other possibilities, perhaps layout, certain package types.. anything that would allow the avoidance of the extra series diodes?

Thanks
 
#15
classd4sure said:
That seems like the measure one would want to avoid at all costs, due to a fairly substantial hit on efficiency. Could there be any other possibilities, perhaps layout, certain package types.. anything that would allow the avoidance of the extra series diodes?
Click to expand...
On losses are not the principal loss factor. The losses in the series schottky count as on losses.
Suppose you're running a half bridge amp from +/-80V supply rails into a 4 ohm load (800W) and the voltage drop on the schottky becomes maximum 1V at 20A (0.2V plus an internal resistance of 40mOhm), you lose about 10W in the series schottkies. That's 1.25%. Incurring this loss buys you the right to use larger FETs with disregard for the body diode.
I'd say there's no net efficiency loss when using current-steering diodes compared to a full bridge.

Experiments have been done to integrate a schottky onto the MOSFET die, by "simply" extending the source metal. Unfortunately, the effective resistance of the schottky becomes rather high (a drawing would help here but that's where my energy peters out). So that didn't fly. Co-packaging only helps if you can use one of those cute bond-wire free packages (IR), and even then the takeover speed is not fast enough for normal class D use.
 
Status
Not open for further replies.
Top Bottom