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

Re: Terry's DRSSTC - Frequency/Pulse Width Controller

Original poster: Terry Fritz <teslalist@xxxxxxxxxxxxxxxxxxxxxxx>

Hi Bob,

Thanks for your thoughts here!! I am definitely a "ground plane" kind of guy when it comes to circuits. I left out the IC bypass caps and such on the schematics I posted. The wires to the temp sensor and the IGBT drives will just be twisted pair.

200++ amps of 100kHz is exciting, but suppose I trip the thing off at a current peak! The voltage will smash up against the anti-parallel diodes and create all kinds of "noisy excitement" as I wildly try to shut the mess down ;-)) I try to keep things very low-Z to reject stray noise too. R3 and C1 form a filter at about 16kHz. I did not want to use more capacitance since draining the cap during power down might hurt the LM339 if it were too big. I will have to resistor and cap the +5V to the LM35 since it is on the heatsink but I will stick it on the buss side so the big caps will absorb "shock" there too. The heatsinks have no isolation thermal pads so they can get maximum thermal transfer. Thus, they float at full voltage... Ground plane "singing" is an issue. I will have to be carful there since everything is relative to that plane. I think McMaster Carr as 12x12x1 inch copper plate :o))) But if I do run into any big noise issues, they will have to be studied and corrected. Best to solve such problems at their source rather than trying to band-aid against them... I have not used ferrites and mica caps "yet" ;-))

The current sensors have to be really fast!! The zero crossing switch cannot have any phase lag. The IGBT over current might have to protect against a shorted IGBT (Hahahahahah!! I am "dreaming" there :o))) The spike is about 8000 amps in that case!!! But the "other" IGBT will fold at about 800 amps so there is a slight chance a limit circuit could stop it.... I "could" reduce the gate drive voltage too to limit the fault current, but that is usually a "bad" idea. Best to transfer them fault joules to the "exploding" IGBT... Of course, best not to have the IGBT blow at all ;-)) If I can keep them cool and limit the peak current to less than 750 amps, they will be fine... I am not really planning on needing more than 100 amps though *;-)))

I am going to have a "hand box" now to make the timing signals and it will be fiber optic. I am using my old fiber probe ideas. I was stunned to see that they are at 125MHz now!!! The fancy fiber optics add about $50 but I will not get electrocuted if the mess goes nukies... I am still tempted to make it a basic stamp thing (not quite good enough to use PICs yet 0:-)) The serial updating of an LCD display is too slow and would mess up the timing... I will probably go with "Terry's wonderful digital counter divider thing" yet to be drawn out. Based off my ThereminVision sensor work in robotics... It would have a bit of "jitter" to it to mess up the FCC standard metering >:-)) The Rabbit RCM3000 "jitters" the uP clock signal so the EMI harmonics jump all around and do not register on a spectrum analyzer. The noise is at 100..no 95, no 103, no 92.... MHz thing ;o))) I am thinking of Finn here ;-))

I did test the cheap DigiKey current transformers and they look very nice!! I wrote a little thing here about them:

http://drsstc.com/Design/CurrentTransformerTest.pdf

They can do what a Pearson current monitor does "to a point" at 1/100th the cost ;-)) The top yellow trace at the end is the $7 digiKey part and the Blue trace is the $850 (retail. Not E-bay ;-)) Pearson 411. Before Rick asks, It is the canned "cardiac" waveform from the HP signal generator ;-)) Very nice for this stuff too!!

Cheers,

        Terry





At 01:30 PM 12/15/2004, you wrote: 
........
>
> http://drsstc.com/Design/OverTemp.gif
>

To reduce EMI sensitivity  it's good design practice  to add a low pass
filtering to signal where its connected to the board. At say a time constant
of  1/20 or even 1/100 of the coil frequency.
You can also add  a second one to the last active element of the signal
path. You don't want more bandwidth than you need.
If you want to get real fancy add a rate of change term i.e. a bit of phase
lead to compensate for the thermal lag.  That should give a better
indication of the actually die temperature during transient conditions. This
will make it more sensitive to noise so you will also need the low pass
filter.

You should do the same on the current sensor to reduce false triggering
during switching transients. More tricky but say three or more times the
coil frequency.

You should also consider trying to make the inputs differential.  That looks
like it would only need a few more resistors. The signal reference points
should return to a stare point on the circuit board.

Bob