[Open_electroporator] 1mm cuvettes?

[John Griessen]

On 09/12/2017 09:16 PM, Nathan McCorkle wrote:
>> OK, that means I need to undo the 12 supply and put something higher like
>> the 19 to 20V laptop supplies.
> Is that because you want the rising time to be faster than it would if
> you used the adjustments to push it higher? 

No, just to get high enough at all.  With a 39k load, and 12V DC in, any less pulse width than a(50,10,x) is too little.

pulses closer together would be needed to get a little more out.  Also that non-linearity
of the losses of volts in diodes seems to mater.

I would have guessed
> throwing the 12V on what I'm using (or what you have now) would just
> mean you need to push it a bit differently.
> 
> The non-linearity is something I too noticed. I was thinking it would
> be useful to have self-calibration via the ADC and a dummy load,
> shmoo-ing the variable space and being able to provide some sort of
> mapping for a "linearly_corrected_a" function, etc...

Hey, Intel shmoo plot reference!  I saw Noyce in a hallway at UT once in the 90's...


. . . and can get back to my duties of testing and
> reading/coding about the ADC and DMA.

Wow!  You feel duty bound?  Can I hold you to that?  :-)


On 09/12/2017 09:28 PM, Nathan McCorkle wrote:
 > I think the problem is here is the pulse period being 34... I think it
 > is getting to be sooner than the IRQ handler can start, thanks to
 > MicroPython overhead:
 > https://forum.micropython.org/viewtopic.php?t=3174&start=10#p18760
 > "It doesn't take much python to get 12 usec worth of code."
 >
 > and it isn't just our code, it's also the code we don't see.
 >
 > Notice in the "tim3_enable_tim1_tim2_tim5__tim1_shuts_down.py" file the comment:
 > to replicate John's Original period = 3194
 >
 > this is 2 orders of magnitude the original settings that "worked" at
 > least on the 'scope... so we have quite a range to work with. I'd bump

Careful...  that probably involves the scale factor I used at startup in my boot.py file....
which is not used by the program we use now, your contribution.

Let's look at theoretical, approximation with no losses in diodes, etc...  12V in is full AC peak value after being switched,
push and pull input windings give 24V peak to peak, but that is not visible in the output because of the way the doubler is one 
sided, and the pull cycle just recharges caps, not generating extra volts anywhere, so 12V in pushes, turns ratio = 609/8 = 76.

So the volts that get "doubled" are 76 * 12 = 913, so "doubled" = 1827, and reality doubled is more like 70% of that because of 
the load and losses, and pulse duty cycle being small.    That's why 12V is too little.  Because the transformer was redesigned 
since cs-hvt-epc19-01 to work with higher volts lower current on purpose.

I think your micropython code is good, no need to delve into using the timers with fewer interrupts to "get around" micropython's 
inherent  limits, and just use micropython for its strengths -- good coding, easy coding.  Using 15V for DC in is good for 1mm 
cuvettes plus a good extra 44%, and going up to try for 2mm and 3600V runs into design limits.  I think the transformer can take 
the volts, so capacitors and diodes in the doubler would need redesign...  What headroom can we expect?  If 15.2V gives 2600V 
loaded, what will 18V give?  (I don't want to go as high as 20V anymore because of some max specs for transistors)  The 
non-linearity is helpful in this direction, so approximate by direct ratio.   2600 * 18 /15.2 = 3078, not enough...

So aiming at 1mm cuvettes and 1mm spaced electrodes for zapping is still our design goal for this system. No easy tweaking will 
change it, and 15 to 17 volts for the DC in is good, and capacitors and diodes of 3000V rating are chosen with lots of parts 
hunting work that I won't afford to repeat soon.  With 17V DC in, the input to the doubler components would be 76 * 17 = 1292V, 
and "doubled" = 2584.   the caps' 3000V rating is 16% over that, which seems a good place to stop for parts tolerance life.

John