Bob Boyce TPU thread

[eldarion]

Eldarion, you have not mentioned what if anything you are reading as outputs from your other longitudinal windings.

Applied pulse width may be set for 500 ns, but what is the actual pulse width at the primary? Have you checked this to see? You may have to increase pulse width or decrease pulse width based on the response time of your drive electronics.

Bob

Hi Bob,

The pulse width is 550ns at the primary (550ns of actual "full on" where the voltage is constant), but there is about 1us rise and fall time.

I am reading a small signal at the longitudinal outputs, it looks like a smaller version of the secondary output.

Do you still have the information on the G-strain energy absorber?  I would like to study it if you don't mind--it might help me learn a bit more about this technology, seeing it from a different angle.

Thanks!

Eldarion

[eldarion]

Bob,

I have one other question for you:
I am seeing a lot of ringing on the primary coils directly after each pulse.  Is this desireable, or should I be attempting to suppress it?

Right now, I am just letting it ring...
I have corrected the ringing issue, and power input has dropped.  Now, there is very little difference between open-ended drive and closed drive.  However, there is now only a tiny output signal on the secondary for any frequency between 1KHz and 200KHz.  (I didn't test beyond 200KHz.)  I seem to be getting good drive into the primary coils for the power levels involved, so I think any impedance mismatching is minimal.  Also, the rise/fall times are now in the nanosecond range.

And this is probably related.  I wonder if this could be the problem?

If you do not see the effects, then your pulses may not be of sharp enough rise/fall times, or there may be too much of an impedance mismatch between the drive and the transformer.

Thanks!

Eldarion

EDIT: So much fun to talk to myself here!   At least this might help other experimenters going down the same path...
Bob did mention that unipolar DC pulses were important.  With the current MOSFET-based drive setup (the same type of setup that is used on the hydroxy system, i.e. MOSFETs used as pulldown "switches"), the impedance from the pulse generator is swinging wildly from about zero to almost infinity as the pulses are generated.  When the line is released, and the impedance is therefore nearly infinite, the primary coil will ring.  I do believe this is very bad, so I will see if I can eliminate it. 

If I am wrong in my thinking here, please say something!  It would seem that the hydroxy system is vastly different than the pure electrical version, despite similarities in coil construction.

Also something to think about, the pulse width is 500ns as you know.  That corresponds to a frequency of 1MHz, which causes the impedance of the primary to be quite high.  The impedance mismatch may be bad enough to require some sort of capacitive matching, but I am not certain at this point.

EDIT2: Hmm...maybe the signal generator and output circuit / HV potential grounds do NOT have to be separated, as I had originally thought:

Oh, all applied potentials are in reference to earth ground.

(Here Bob was referring to the HV potential and the pulse generator)

FINAL EDIT: Well, as you can see, I tried the above and still nothing.  So, I know the pulse generator is now operating properly, but something else is still wrong!

[Earl]

Eldarion,

There are some things to keep in mind when wiring a FET switch.

1)  Regardles of the switching frequency, the physical layout must be as if you were building an amplifier for 1 GHz.
Think of all FET switching circuits as if they were a GHz linear amplifier that is being over-drriven at xxx kHz.

2) The consequence of building a GHz amplifier is that all lead lengths must be EXTREMELY short.  Every mm, every tenth of a mm is important.  All of the current loops shown in the attached drawing must be very, very, VERY short.  Do everything you can to keep the total loop lengths short, shorter, and still even more shorter.  NO effort is too much to reduce lead lengths to close to zero.

3)  The center of the universe is the source pad on the die.  Everything turns around this.  Unfortunately, the source pad is not available.  We must live with the inductance of the bonding wire to the lead.  So we make the center of the universe as close to the package as possible.  Each and every nH works against you, so keep the series nH's as low as possible.

It is not the ground that is important, it is the FET source.  Everything that needs to be grounded goes directly to here; short lead lengths to the source, as close to the package as possible.

In one consultation, it was necessary for the client to mount the SMD FET driver on the opposite side of the PCB as the FET so that its ground and the FET source had very small connection distance.

4)  All by-passing consists of 2, better 3 or more ceramic capacitors.  Use SMD if at all possilbe.  Different values.  1nF, 10nF, and or 100nF, in parallel with a SMD tantal is nice to use.  Only the caps supply the current to charge gate and coil primary, the power supply doesn't supply any current at all for circuit operation (if it does, you have problems).  All the power supply does is recharge the caps.

5)  To discharge the gate quickly, the driver IC ground must be connected directly to the FET source, keep this length as close to zero as possible.

Eldarion, the ringing should be small, are you using 3 different ceramic caps in parallel and using current loop lengths close to zero and star grounding to the source, as show in the attached schematic?

Regards, Earl

[eldarion]

Hi Earl,

Thanks for the tips!  The ringing I was talking about was in the coil itself, when the FET opens.  I was able to damp that out, and now the waveform is a series of spikes as it is supposed to be.

All,

I have just realized something with regards to my DC potential supply, and thought I would post it here so that others do not make the same mistake.  See attached diagram of my output stage; this might explain my poor COP.  The impedance of the HF short circuit is extremely low in comparison to the impedance of the load and DC blocking capacitor, so most of the secondary output power is being shunted away into the HV bias supply and ground.

Eldarion

[Grumpy]

Potential