diyAudio

diyAudio (http://www.diyaudio.com/forums/)
-   Class D (http://www.diyaudio.com/forums/class-d/)
-   -   poor switching (http://www.diyaudio.com/forums/class-d/35939-poor-switching.html)

classd4sure 12th June 2004 06:06 PM

poor switching
 
Hi,

The load: IRFP250

I calculated 1 amp drive current based on 100nC gate charge for 10Mhz. Yes? This is for a PWM application.

That should turn the sucker on nice and fast in the worst case I thought.

Now then..for the driver transistor...
After many hours of searching, I came up with the zetex FCX1149A, which is identical to their ZTX1149A, different case, handles twice the dissipation, slightly better Hfe.

I chose this one for it's ultra high beta and current rating, I can't go with a darlington as it would leave me with too high a base current requirement.

I couldnt' get it to work very well at all....couldn't figure out why it was switching so poorly..

So I stripped everything else away (I'm in spice land), and just tried switching the driver transistor through a 1 amp resistive load of 11ohms.

It switches so slowly the best I get is a sort of triangle wave out of it, even though I get the current I want, it's useless switching so slowly.

Ft is 135Mhz according to the data sheet, Ton=150ns Toff=270ns. Output capacitance 50pF. It should be able to handle a 1Mhz frequency I thought....it doesn't.....it doesnt' even handle half of that well.

What did I do wrong in the selection process, I thought this would work great.

I need some input here because if I can't select them better than that....may as well give up now.

Thanks

Chris

jan.didden 12th June 2004 06:08 PM

Chris,

Can you post the (spice) schematic?

Jan Didden

classd4sure 12th June 2004 06:24 PM

2 Attachment(s)
Hi,

The mere fact you asked me that tells me my selection wasn't all that bad?

I set the circuit back up as the test switch, 2n5401n handles this same circuit far better....only can't get 1 amp out of it. Strange power scheme is because of how the prior circuit was..

Thanks
Chris

jan.didden 12th June 2004 06:40 PM

The fact I asked you is based on lots of experience that what one *thinks* is in the circuit is not. I don't want to start to count how often this ****ed me off....

Jan Didden

jan.didden 12th June 2004 06:45 PM

OK, question: you are assuming that the puls generator floats atop the negrail. Did you check that? Many simulators *automagically* assume that any source is sitting on ground. Try to do it with all those voltage sources deleted, the puls generator against ground, the collector resistor to ground, just a pos battery for the emitter supply. And that 1.1k in the base lead doesn't help either, 50 ohms is better if you want just to test the transistor.

Jan Didden

jan.didden 12th June 2004 07:17 PM

2 Attachment(s)
....like so:

The switch-off is slower than the switch-on, as expected. A B-E resistor should improve that. And it's not the same transistor, but good enough for government purposes.

Jan Didden

Pafi 12th June 2004 07:57 PM

classd4sure!

- At fT transistors has 0dB current gain.

- BJT's switches off veeery slow, if they were saturated. Saturation must be avoided. Check out shottky-transistor!

- 1 A is not enough for 10MHz, since it has to turn on, and turn off during a period! 1/(100ns+100ns)=5 MHz.

- The charge must be loaded to, or sinked from gate at Ugs=4...5V. At this voltage only 0,4A flows on 11 ohm.

- This problem is already solved. Solution is MIC4421, or UCC27321 9A MOSFET drivers (for example).

- Maybe you should choose faster MOSFET. For example IRFB31N20D, W34NB20, or under 150V: FDP2552...

I hope I helped You!

classd4sure 12th June 2004 08:39 PM

2 Attachment(s)
Hi,

Yes, that's a very fair question, and a good one...I saw it comming. I know multisim would never accept that ..this program does, I of course, did check that.

The reason I chose 1.1k..

Under these conditions I reasoned min Hfe according to the data sheet graph =200.

I used 100 for calculation...more than safe I thought.

so for the 10mA required at the base=1.1k....half of what "should" work. IRL...I'd take required base current X3..but this is spice, and I reasoned I had already left enough of a margin of error by using Hfe 100 instead of the 200 it actually should be.

Am I wrong? Looks like it..but why?
50Ohm or 1.1k ohm ....still gives me my 1 amp of course. I think the 1.1k should do a far better job of switching it though. I will attach a pic of the results given the same circuit with the two different base resistors driven at 500Khz and an easy duty cycle of 50%, I think you'll agree it's useless for PWM at even that slow a speed, even with the 50ohm.

At 50Ohm it measures 325n off delay, worse than the 270n it's rated at. It will have a baker clamp ...should I really have to overdrive it like this just to get it to fully switch? I would think not.

Regarding the 10Mhz, working frequency of this circuit will be around 350k PWM full modulation. So again it seemed to me that 10Mhz was enough of a margin to calculate required drive current with, to allow it to swith on /off in 100ns.

Pafi, You are correct Sir, the gate resistor calculates to 5ohms in my actual circuit, the reason it's 11 ohms here is because it's not the actual circuit. 11V/1amp load =11ohm..just to see how good it can switch this current.

Jan, that's a respectable looking wave, do you only get 200 to 500mA out of it? Too bad..Guess that's not my transistor either.

Thanks

Chris

PS: If this makes no sense ...I'm simply burnt....and will edit later;)

jan.didden 12th June 2004 09:04 PM

Well, my graph showes 8V across 11 ohms from a 9V battery, which for a small signal transistor is good (VCEsat=1V), and the switching speeds show that it works. It's really a trivial cct of course. If it doesn't work in your case, it's not the transistor, believe me.

Did you try with the puls gen against ground? How did you check that the floating puls gen works?


Jan Didden

Jaka Racman 12th June 2004 09:55 PM

Hi Chris,

I do not know if your requirements are really serious (10Mhz switching), but presuming that they are I do not think you have chosen the right way. I have been simulating switching designs almost to death, and I have found out that component models usually do not work very well in Spice. Sometimes FET models lack package parasitic inductances etc. So for longer simulations (over many switching cycles) I prefer to use S (switch element).

If your requirements are really 10MHz, you are candidate for IXYS RF (former Directed Energy) FETs. Also for 10Mhz switching, you would not use classic drive circuit, but some kind of resonant driver. Those usually use gate inductor instead of gate resistor. Such drivers have low losses, since they conserve energy in resonant circuit instead of dissipating it in the gate resistor. There are many variants of such designs, and I may try and find some examples, if you are interested.

Best regards,

Jaka Racman


All times are GMT. The time now is 12:02 PM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio


Content Relevant URLs by vBSEO 3.3.2