[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: DRSSTC Noise source found



Original poster: Terry Fritz <teslalist@xxxxxxxxxxxxxxxxxxxxxxx>

Hi Robert and Antonio,

At 10:01 AM 2/17/2005, you wrote:
Hi Terry,

Its Probably the inductance of the caps resonating with the C and not
related to the fact the dielectric is polypropylene.

Yep!! I have heard poly turns into resistive mush at say 10MHz, but that is typically far past poly cap's resonant frequency. We have fancy spectrum analyzers at work, but they have more buttons than I have brain cells 0;o)) Maybe in a dull moment I could try to measure them if I can get someone there to help me.


But the lead inductance and normal C resonance does seems to be a big deal!! My new caps will top at 250MHz so that should blow the resonances away. The H-bridge is very compact with very short distances so it is well capable of 10MHz stuff... But it certainly did not have capacitive bypassing for that!! Pretty easy to fix now that I see what is going on *;-))

Keep the leads as short as you possible.  I don't recall your particular
construction

http://drsstc.com/~terrell/pictures/F-H-Bridge-Drive.JPG

http://drsstc.com/~terrell/pictures/F-H-Bridge-Drive-Big.JPG

http://drsstc.com/~terrell/pcbart/H-Bridge.gif

Not to terrible MHz++ short buss and dual wiring stuff... But very "under bypassed" for HF =:O

but if you can allow the caps to sit in the same plane as the
buss rails with less than a 1/10in of lead to the soldered connection to the
supply rails.

They are still wire leads. I have some "super" 2200pF ATC 4040 SM caps but I hate to use them since they are not at all easy to get. Rather solve the problem with plain ol' DigiKey stuff...


Your new caps may do better not because of the fancy dielectric but because
they have a smaller capacitance.

Yep!! Two 10nF in parallel. Should do 225MHz with the lead inductance.

Ceramics should do just fine but again keep the lead length as small as
possible, several in parallel will be much better than one large one with
just one pair of leads.

I hope the cheap ceramics work since they are only 1$. The ruby mica are $16!! each. Right now the H-bridge is only $80 total. I got the ruby's just to be sure I am covered in any case. Be interesting to see how they compare to the standard ceramics.


Impressive that the IGFET's switching can generate 50Mhz stuff.
I thought shoot thru would be controlled by the dead time.

I apparently can go up to 17 ohms on the gate right now and that is enough to pull down to stop shoot through. The dV/dT is terribly fast!! 3000V/uS stuff...


http://drsstc.com/~terrell/pictures/NoCrossDelay-02.gif

No wonder it rips right through those big poly caps!!! The current is in yellow. You can see the IGBT turning off and going up a few volts at turn off (~250nS). It floats against the reverse diode for 200nS more till 450nS when the lower IGBT rips to voltage down 300V in <20nS!! The logic seem to be opening the top IGBT within 100nS... The 200nS delay seems to be a waist... The lower ring should get fixed by the new caps. With 5% duty cycle the switching thermal load is a none issue. So why not switch it super fast >;-)) I guess the noise can also be eliminated if the circuit simply works "faster" than the noise does ;-))) Steve Ward runs it all "fast" too and his stuff works great ;-))

It is probably best if you increase the gate drive resistors to reduce the
switching speed, which is the source of the noise.

That will only really increase turn off time really.

http://hot-streamer.com/temp/IGBTswitchtime.gif

I am running 25 Volts to force high collector currents so the transconductance region is raced past way too fast any way (might be better if the gate drive were a sine wave...) If I put in too much gate resistance ( >17 ) shoot through will kill it. I could by pass turn off with a diode, but I only turn off to ground level instead of say -15V so the diode voltage drop gets scary... The mess can take one heck of a lot of noise right now! Hopefully this capacitor fix on the buss with just fix everything real nice. Hope hope...

It appears to me your switching way to fast.  You have soft switching anyway
so why generate all the hash with fast switching for no reason???

Those IGBTs switch &%^%& fast >:o))) Those silly 150kHz numbers in the data sheets are "thermal" limited since the switching losses are high. But we work at say 5% duty cycle. So, in our case, we can switch them in the 100nS range!!! Weeeee!! Every 10nS is about an amp added to the switch current at 500 amps peak. So really cutting down the switching time seems important for us. Apparently, the IGBTs can do those speeds. IR IGBTs have lowered gain to allow that. Not sure about the 40n60s, but they seems to do fine too!!


So if we can handle the noise, no reason at all not to switch them ultra fast!!!

The present parasitic oscillations may be dissipating much power than the
increase you may get with reduced switching speed.

Have to know the impedance of the caps at 25MHz to know just how much power there is behind those glitches... My feeling is that it is not much.


But again if your soft switching its the ON current dissipation that will be
dominating so you can probably reduce the switching speed dramatically
without significant dissipation problems.

Hard to say what is best. Just have to try things... Right now, "number one" is to get them 10nF caps in that buss!!! I missed the shipping dead line last night, so they will be here tomorrow...


Cheers,

        Terry


Robert

----- Original Message -----
From: "Tesla list" <tesla@xxxxxxxxxx>
To: <tesla@xxxxxxxxxx>
Sent: Wednesday, February 16, 2005 8:23 PM
Subject: Re: DRSSTC Noise source found


> Original poster: Terry Fritz <teslalist@xxxxxxxxxxxxxxxxxxxxxxx> >
.....