[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
OL-DRSSTC 11
- To: tesla@xxxxxxxxxx
- Subject: OL-DRSSTC 11
- From: "Tesla list" <tesla@xxxxxxxxxx>
- Date: Thu, 13 Oct 2005 15:39:07 -0600
- Delivered-to: testla@pupman.com
- Delivered-to: tesla@pupman.com
- Old-return-path: <vardin@twfpowerelectronics.com>
- Resent-date: Thu, 13 Oct 2005 16:49:58 -0600 (MDT)
- Resent-from: tesla@xxxxxxxxxx
- Resent-message-id: <DlYn1.A.qfH.VSuTDB@poodle>
- Resent-sender: tesla-request@xxxxxxxxxx
Original poster: Terry Fritz <vardin@xxxxxxxxxxxxxxxxxxxxxxx>
Hi All,
I took some fascinating waveforms!!! Pretty technical stuff though %:-)
Here is the drive voltage from the H-Bridge (blue) and the primary
loop current (yellow):
http://hot-streamer.com/temp/OL-DRSSTC-2005-10-13-001.gif
If your really interested, here is the raw data file that can be
opened in a spreadsheet or MathCad (text too):
http://hot-streamer.com/temp/OL-DRSSTC-2005-10-13-002.CSV
Here is a close-up at the highest current peak:
http://hot-streamer.com/temp/OL-DRSSTC-2005-10-13-003.gif
The DRSSTC designers will note that this wave form is pretty
odd!! The voltage on the buss caps is dropping linearly as expected
in this system during the firing cycle. Since it is run off a
variac, the buss voltage is 52V volts at start in this case.
First you will notice the rather extreme dip in the drive voltage in
the center of what should be a squarewave. The dV/dT and current
matches about 5uF which is what the big poly buffer caps are. This
suggests that the electrolytic array is pretty soft into 100
amps. It looks like they can hold their own again at about 60
amps. This might be a real problem at higher currents. Here is a
detailed picture:
http://hot-streamer.com/temp/OL-DRSSTC-2005-10-13-003a.gif
The droop seems less on the negative side, so perhaps the
electrolytics can supply more negative current than positive
current... But it looks like I might need better electrolytics or
more poly buffer caps at higher current.
Next we look at the IGBT switching cycle:
http://hot-streamer.com/temp/OL-DRSSTC-2005-10-13-004.gif
It looks like the IGBTs are shutting off about 50nS after the zero
current crossing!! That is very good since it means the bridge is
not "fighting" the primary loop oscillation. Note how the bridge
actually switches almost 1uS after the current crossing!!
For the next ~700nS, the voltage has just jumped to the rail voltage
limited by the reverse diode. This is very interesting because the
circuit is free of drive at this point. All the IGBTs are
independently controlled by independent CTs. It looks like the buss
caps are being recharged a bit here since the voltage rises some in this zone.
This is a close up of the IGBTs switching from on open circuit
limited by the reverse diodes to full IGBT "on" in 370nS:
http://hot-streamer.com/temp/OL-DRSSTC-2005-10-13-005.gif
I suppose one could figure switching energy (or push a bunch of
buttons on the scope to calculate it directly), but I have not
yet. The actual switching time is about ~370nS so it is very fast,
just pretty late. Note that the primary current (amplified here)
shows no switching glitches at all!
http://hot-streamer.com/temp/OL-DRSSTC-2005-10-13-007.gif
Here is the gate drive signal in relation to the primary loop current:
http://hot-streamer.com/temp/OL-DRSSTC-2005-10-13-006.gif
For a current transformer driven gate with a resistor across it, the
switching time is given fairly closely by this equation I derived. I
wrote the derivation down in my notes here next to Fermat's theorem :o)
T = SQRT ((V x C) / (K x D x I x F x pi))
T = switching time
V = Voltage from initial state to IGBT turn on (25 + 10 = 35)
C = Gate capacitance (3.3nF)
K = CT ratio (1/100)
F = Frequency (96000)
R = Gate parallel resistance (100)
I = Peak current
D = Division ratio (0.195)
D = R / SQRT(R^2 + (1 / 2 x pi x F x C)^2) = 0.195
T = SQRT ((35 x 3.3e-9) / (0.01x 0.195 x 100 x 96000 x pi)) = 1.40uS
So one can play with these variable to try and reduce the late
switching time. If there were no parallel resistance at all, it
would be 619nS. Lowering the gate rail voltage to 20 volts would
reduce it about 100nS but that is probably a bad idea for such a
small gain. Lower frequency helps, but the coil is as it
stands. Probably do not want to double wrap the CTs since they are
already taking a LOT of current already ;-)) Custom CTs might do
better here, but I like the COTS ones ;-) IGBTs with lower gate
capacitance helps, but the IGBTS are as they are here.
The time definitely goes down with higher primary current. At 500
amps it is 626nS. With 500 amps and no parallel resistor, it is
276nS. So the time will pick up naturally with higher current but I
need to look at a higher parallel resistance for sure!
The current at switching is:
Is = I x 2 x pi x F x t
So at 500 amps and 276 nS:
Is = 500 x 2 x pi x 96000 x 276e-9 = 83 amps
With a 100 ohm resistor that is 188 amps!!
So the gate drive resistor needs to go higher which is the next order
of business....
We can also estimate IGBT heating:
P = BPS x t x Is x Vbuss / SQRT(2) x cycles
P = 60 x 276e-9 x 83 x 250 / 1.414 x ~100 = 24 watts
Happily, that is easily within the range of the H-bridge.
If we leave in the gate resistance (as it is right now):
P = 60 x 619e-9 x 188 x 250 / 1.414 x ~100 = 123 watts!
So the OL-DRSSTC really is simple, when it is all done ;-))
Cheers,
Terry