Testing the TK Tar Baby

[TinselKoala]

TK,

You have already spent a lot of time and effort on this, and I for one greatly appreciate it. 

But, if I could trouble you for one more measurement:

With the circuit oscillating similarly, please post a scope shot similar to your #490 post timebase and with both vertical channels set as you did for Vbat, but with the scope probes on opposing ends of the load resistor.  I would like to see just how much Vdrop there is at both DC, and more particularly, at AC.  This should provide us with some idea of the battery AC impedance at Fosc.

If you need to get to that "real work", I will understand.

PW

The HP scope does not have isolated ground references nor do I have a differential voltage probe ... so if I'm following you correctly, I can't make that measurement across the load with probes back-to-back.

Or do you mean one probe on one end of the load, one probe on the other end, and both reference leads to the normal circuit ground point? That I can do, of course. But that's equivalent to what we've been calling "battery" trace for the one probe and the "common mosfet drain" trace for the other probe, isn't it?

Later... I have to get the dogs herded to the vet for annual vaccinations this afternoon.

[picowatt]

The HP scope does not have isolated ground references nor do I have a differential voltage probe ... so if I'm following you correctly, I can't make that measurement across the load with probes back-to-back.

Or do you mean one probe on one end of the load, one probe on the other end, and both reference leads to the normal circuit ground point? That I can do, of course. But that's equivalent to what we've been calling "battery" trace for the one probe and the "common mosfet drain" trace for the other probe, isn't it?

Later... I have to get the dogs herded to the vet for annual vaccinations this afternoon.

TK,

Yes, I meant probe grounds to batt- and the probe tips at opposing ends of the load resistor.  And yes, absent any interconnect inductance, this would be similar to Vbatt versus Vdrain.  Your scope shot from post 490 was really clean, and a similar shot across the load resistor would be telling.

Dogs?  (as in more than one?)

PW

[MileHigh]

TK:

I understand that you can't do everything and suggestions are coming at you from all angles.  So I have another "make work" suggestion for you but don't let me break your stride.

The suggestion is for getting a more accurate manual integral calculation.  I am going to give you an image-processing-based solution assuming that you may have some software.  Also, I am going to give you a more basic way of doing it also.  Here goes....

You take what you consider your best picture of your scope display.  It would be preferable to have a pic with a faster time base showing one or two complete cycles, but the picture you are working with now will also work.  My discussion will be based on a single cycle, but you may prefer to do it with two cycles.

You take your image and then you crop out a single cycle.   Then you do a horizontal image stretching so that it has the same aspect ratio as an 8 1/2" x 11" sheet of paper in landscape mode.

Now for some image processing software tricks.  You invert the image so the trace is dark and the background light.  [I am adding one more IP step from the original posting:]  Then run a low-pass filter to soften the image to give it a "slightly out of focus" look.  You are doing this step to make the next step, the image thresholding "cleaner."

Then you run image thresholding such that all pixels above a certain intensity are white and all pixels below a certain intensity are black.   So at this point you should have a thick solid black line that represents the waveform on a white background.  The line will not be clean may have some jagged edges, etc.

Then you run a few passes of an erosion filter and you erode the thick black line away until it becomes a thin black line.  Then you superimpose a grid on top of that line so that you can make easy measurements.   Finally you print that out in landscape mode on your printer.

I assume you can see the logic here.  You stretch out the waveform so that you can extract more information from the more gradual slopes.  You are doing a kind of interpolation.

Without the image processing functions you could pretty much do the same thing manually.  Just invert the image then crop and stretch so that your full cycle will print nicely in landscape mode.  You might want to increase the contrast in the image or change the gamma of the image.

Then you print out a stretched and fuzzy waveform, and you just do the "image processing" by eye.  You take your pen and ruler (or curve fitting template) and draw a nice line down the center of the fuzzy waveform.

Either way, your manual crunching should be easier and more accurate, and obviously you can now reduce the delta-t and get more precision.

Another make-work project as part of the "New Deal" to invoke closure....

MileHigh

[TinselKoala]

@MH: You must be crazy.   

Results for the first two cycles. I measured the distances off the screen at high magnification with a divider from my drafting kit. Then I used the yellow scale on the left to count the divider's width in gridlines. There are 7.5 gridlines per scope screen vertical division. The rest is just math.

ETA: IMPORTANT:
The first spreadsheet shot I posted had a formula error and the means weren't correct. Here is the correct mean calculation for the first two full cycles.
Sorry if there was any confusion.

[TinselKoala]

TK,

Yes, I meant probe grounds to batt- and the probe tips at opposing ends of the load resistor.  And yes, absent any interconnect inductance, this would be similar to Vbatt versus Vdrain.  Your scope shot from post 490 was really clean, and a similar shot across the load resistor would be telling.

Dogs?  (as in more than one?)

PW

Maggie, Murphy, and Mommadog. Momma is my housemate's old bitch, Murphy is another rescue from the park who has been here about a month now.