New Rosemary Ainslie Demonstration Scheduled for Sunday, 4 August 2013

[poynt99]

Just a little addendum:

I am going to assume that the sampling rate by the DSO is fast enough to resolve the individual spikes in the oscillation.  So that means you can export the data and then analyze it with a spreadsheet.  This of course eliminates the sub-sampling and software algorithm issues related to massaging the data for the display.  With the exported data you work with the pure full-resolution of the DSO capture.

All that you have to do is cut out a time slice that corresponds to the pure oscillation phase.  Calculate the area above zero and calculate the area below zero to compare your discharging to the charging.  To do this, all that you have to do is sort your column of current sampling data points by value.  Split the data into above zero and below zero by inserting a blank row.  Then add up the positive values and add up the negative values and compare with the column summation function, the "sigma."  Then just compare the two values to see the average discharging current relative to the average charging current.

Once you figure out the start and end times for your time slice corresponding to the pure oscillation phase, you could do the calculation in less than five minutes.

You got it

[TinselKoala]

@MH: The scopes that .99 and the Ainslie team are using can sample fast enough to perform accurate math on the measurements of this circuit. The screens do compromise, but the scope's internal memory buffer holds every sample and the math is done on the buffer data, not the screen display.
The integrations and area measurements that you describe can be much more easily and reliably performed by the scopes themselves. I think their scopes can compute areas bounded by traces and selected by cursors, and this math is done on the memory samples, not the screen display.
So to compute the areas above and below the zero current reference one would simply position the horizontal and vertical cursors appropriately (one on the zero ref, and the other first above, then below, the positive and negative peaks of the waveform). Then when the areas are computed the values may be compared directly, using just a scratchpad. No fancy, errorprone spreadsheet manipulation need be performed.

Here is another test circuit that I would like to "run up the flagpole". What effect on the oscillations, and hence the battery recharge claim, would the insertion of a fast diode have? Consider if you will the following schematic. Will the MUR1560 ultrafast rectifier (600 V, 15 A,  60 ns recovery = about 17 MHz) block the oscillations? Will it affect the power seen at the load, or the battery recharging?

[poynt99]

TK,

I drew this up some time ago. You may want to try it as a variation on your theme. By all means replace the 1N4007 with whatever you wish.

I have not yet tried this.

.99

[Pirate88179]

So, has this "test" been delayed as I predicted?  Or, is it just taking place and none of us will be allowed to watch it live or see the videos later?

I see nothing on Youtube as of yet.

Bill

[poynt99]

Guys,

You see, Rose's contention is that the battery voltage actually varies in accordance with the oscillating voltage seen across the current viewing resistor. When in fact the true battery voltage does not vary one iota. If and when we see a varying voltage across the battery terminals, it is strictly because there is a parasitic inductance in series with the battery cells and between the measurement points.

She has always measured the battery voltage on the peg board where there is copious parasitic inductance in series with the batteries and in between the measurement points. I have shown, both via simulation, and empirically on the bench with her circuit, that this AC variance as seen across the batteries is not "real". In my video #8 the AC superimposed on the true battery voltage was almost completely eliminated simply by placing the Vbat probe directly across the batteries and by minimizing the inter-battery parasitic inductance through the use of short jumpers. If one had a single 36V or 72V battery, there would be even less AC component seen across the battery terminals. There will always be some observed residual AC component when measuring Vbat however, because internally every battery will  exhibit some parasitic series inductance. Check the electrical "model" of a SLAB.

But again, regardless of this residual AC component as seen across the battery terminals, the actual battery voltage, from a chemically-generated emf point of view, does not change. It is a fixed DC value that does not change, except for a gradual increase or decrease when charging or discharging respectively.

So it is not only fair game to average the battery voltage when making battery power measurements, it is good practice. This goes for both instantaneous and average Pbat measurements. Remember, the true VOLTAGE wave form of a DC battery is a flat line.

Now, some may think that when using the following method: AVG(vcsr) * AVG(Vbat) to calculate Pbat, Vbat must be measured directly across the battery terminals, but this is in fact not the case. Vbat can be measured at the peg board where Rose and her team have always measured it, and the resulting value will be the same as if it was measured right across the battery terminals. And this regardless of how much AC component there is seen at the measurement point.

The battery current on the other hand is a very different matter; it MUST be measured with a low inductance CSR right at the resistor body, regardless if you are using scope probes or DMM probes. The reason? The battery current DOES vary, i.e. it has a varying vector, and any parasitic inductance in series with the CSR and in between the measurement points will skew the measured average value.

Once we have a reliable Vcsr measurement, we can either average it and multiply it by the average battery voltage, or we can use the instantaneous samples and multiply them by the instantaneous Vbat samples, which later will be averaged by the scope to produce an average Pbat value. Both methods will produce exactly the same results if executed correctly.

Since we know the battery voltage is truly a flat line measurement, any notion of multiplying an erroneously-obtained Vbat measurement exhibiting wild oscillations, by an equally erroneously-obtained Vcsr measurement from a CSR array fraught with disaster, is a pipe dream. Any "anti-phase" relationship observed between the AC portions of these two traces is a fantasy. The battery voltage is truly NOT doing what is being measured at that test point. Therefore any Pbat value obtained from this flawed measurement configuration is doomed to be erroneous at best. There is no half-witted academic anywhere that would buy that measurement, and thankfully to date we've seen that they haven't.

Now, how does one go about appeasing Rose's misplaced distrust of using AVG(Vcsr) to calculate Pbat? There is probably no avenue available that will lead to success in that venture, but we can try to illustrate the contention by graphical means to a level even a grade 3 student would likely understand. Let's give it a go.

Below you see a scope shot of the oscillation phase as generated in my simulation of her circuit. It's a very nice sine wave, mostly symmetrical in its own right (slightly compressed positive excursion), but there is something noteworthy about it...do you see it? It seems to be offset somewhat from the zero reference line. And indeed it is. I have drawn in the zero reference line for your....reference. It would seem that the wave form is shifted upwards in the positive direction, wrt to the zero reference line.

Now, when using the AVG[p(t)] method to obtain the average Pbat value for the circuit, the Vcsr samples are multiplied by the Vbat samples, then averaged. Since we know Vbat is a flat line trace, we replace that with a constant K. So in effect, we are left with the CSR trace to determine p(t). In other words, every single sample in the Vcsr trace is simply multiplied by K. As such, for purposes of illustration, we can leave K out of the "equation". So now, the Vcsr trace is a direct reflection of the battery power Pbat.

What is the average value of a pure sine wave with no offset? Well, it is zero of course. What would be the result if our Vcsr wave form was in fact symmetrical about the zero reference line and we multiplied all it's samples by K? Is 0 * K still not 0W for Pbat? Now, what if we shift our Vcsr sine wave upwards as is shown in the scope shot? Do we now not have a simple situation of a constant (the offset) times another constant K? indeed we do.

We can also see graphically, that there is a larger area within the sine wave curve in the positive portion of the sine wave, vs. the negative portion. This too tells us that the "average" of that trace is going to be a net positive value. The sine wave has not been evenly-sliced (horizontally) by the zero reference line, and that tells us that there is a net value associated with its existence, as opposed to an evenly-sliced sine wave that has no net average value at all.

So the point being, even if using the AVG[p(t)] method to obtain Pbat, it really comes down to averaging of the Vcsr trace; either there is a net current flowing in to the battery, or out of it. And this holds regardless if we are measuring the Q1 ON time, or the Q2 OFF time, or both phases of the cycle.

I have included a scope shot indicating the scope-computed average of that oscillation trace, and it results in 206.6mA. If one were to obtain the areas of the positive and negative portions of the sine wave shown in the previous scope shot, their difference would equal this 206.6mA.

.99