Storing Cookies (See : http://ec.europa.eu/ipg/basics/legal/cookies/index_en.htm ) help us to bring you our services at overunity.com . If you use this website and our services you declare yourself okay with using cookies .More Infos here:
https://overunity.com/5553/privacy-policy/
If you do not agree with storing cookies, please LEAVE this website now. From the 25th of May 2018, every existing user has to accept the GDPR agreement at first login. If a user is unwilling to accept the GDPR, he should email us and request to erase his account. Many thanks for your understanding

User Menu

Custom Search

Author Topic: Testing the TK Tar Baby  (Read 1989213 times)

picowatt

  • Hero Member
  • *****
  • Posts: 2039
Re: Testing the TK Tar Baby
« Reply #30 on: March 27, 2012, 03:17:17 AM »
TK,

Why would you want any FG other than that beautiful Interstate.  I'd love to have it.  40volts into 50R!!  Nice...

How on earth are you switching on both Q1 and Q2 at the same time?  To turn on, Q1 needs the FG to go positive, Q2 needs the FG to go negative.  I can't get my picobrain around that...  Does that Interstate also perform magic? (if so, it's the perfect FG for the current task)

The RA circuit appears to only switch Q1 on very briefly followed by a several second duration wherein Q2 is biased on.

Reasons for staying with the IRFPG50:

1.  The much larger Ciss allows more AC current to pass when Q2 oscillates (Q2 is the "array", it is labeled as both Q1 and Q2 depending on which "paper" you read, I'll stick with Q2).  Draw in the Ciss (gate to source capacitances) and you will see that the Q1 and Q2 combined Ciss of 12,500pf or more is the AC path for current when Q2 is oscillating at HF (as well as the FG's 50R).

2.  The larger Ciss allows Q2 to have more AC gain

3.  I thought you were going to attempt a replication

Think about tacking a wire onto your Interstate prior to its 50R output resistor so you can measure across the 50R to get bias current, and, prove that current can indeed pass through an FG.....  Though I don't think much battery charging is going on...

PW

 

picowatt

  • Hero Member
  • *****
  • Posts: 2039
Re: Testing the TK Tar Baby
« Reply #31 on: March 27, 2012, 03:43:17 AM »
TK,

Reasons for staying with the IRFPG50:

4.  They're magical??

PW

MileHigh

  • Hero Member
  • *****
  • Posts: 7600
Re: Testing the TK Tar Baby
« Reply #32 on: March 27, 2012, 04:42:40 AM »
TK:

About the often repeated story about "18 months of continuous use."  I posted once before that all of this is purely anecdotal "evidence" and Rosemary should never have even stated it.  Then you have to ask yourself about what really happened and human nature.  Do you really think that every single day the students (or whomever) would say, "Time to set up the RAT circuit and stare at it again!"  What new insights could be learned by doing the same thing every day for months on end?  It just doesn't add up at all.  I would guess that the circuit was set up and running when Rosie was around and after the novelty wore off it collected dust when Rosie wasn't around.  I don't get the impression that she was there every day for 18 months.

It's all junk anecdotal data not even worth discussing.  I figure that she had somewhere between 10 and 20 megajoules to start with in her battery set, and when you average it out over 18 long months, the setup might have been on 10 to 15 hours per month.  Also, she has no idea whatsoever what the power consumption of the setup was because she measured "energy being returned to the batteries while the setup ran."  Without bothering to crunch the numbers, perhaps she burned off 5 or 6 megajoules during her testing.   And of course, we can't forget that it makes absolutely no sense to talk about draw-down test if you believe that the batteries are always recharging, none!

Moving on to more interesting stuff, Picowatt made a lot of interesting comments and really appears to know his stuff indeed.  I plumb forgot about an AC path at high frequency having less impedance back to the battery ground via the CSR as compared to the the path through 50 ohm resistor inside the function generator.  That may have saved Rosie's function generator resistor from burning out.  So indeed there are two paths for the power to return to ground, a mainly DC path through the function generator and an AC path through the various capacitances in the MOSFET array through the 0.25 ohm CSR.

It's a bit complicated and I acknowledge that when I talked about the "fake" voltage measured across the CSR possibly being from capacitive-inductive coupling, I honestly wasn't thinking about the capacitive coupling through the various MOSFET gate-source-drain AC paths  (I plumb forgot - a dumb mistake).  So there is very likely some real AC power flow there with real current, and not just a "fake induced tingle voltage" with no associated current.

I read somewhere that the various capacitances inside a MOSFET structure are also a function of bias voltage.  The width of the gate channels change as the voltage changes.  (I am outside the envelope of my direct experience here and can't even remember the proper terminology).  However, putting that aside for a second, let me mention a test that may help you get a feel for this.

You could try to use a second signal generator and a inject high-frequency sine wave into the common MOSFET drain node.  As you sweep the sine wave higher and higher in frequency, observe what's happening at the CSR.  I think Picowatt is dead-on and above a certain frequency you should see a very strong sine wave at the CSR.  Call it an "anecdotal" or "get a feel for it" experiment.

Anyways, don't let too many cooks spoil your soup and your fun.  The good news is that you can measure the AC power across the CSR and that is just one component of the "power pie."  As long as you know the total power the circuit is consuming, then the fun can be finding and measuring (or inferring) what all of the "power slices" are.  Picowatt astutely reminded us that some of those power slices are of the AC-only variety.

MileHigh

TinselKoala

  • Hero Member
  • *****
  • Posts: 13958
Re: Testing the TK Tar Baby
« Reply #33 on: March 27, 2012, 04:51:26 AM »
@picowatt:

I'd like to use the 830a for several reasons. First, I can get them for $1.30 each locally, and the PG50 costs about 5 times as much. Second, I think it would really be nice if the 830a turned out to work "just like" the PG50 in TarBaby. But of course I'll be using the PG50s when they finally arrive.

The optoisolators work too well. They cause the mosfet(s) to switch cleanly and more precisely, if with a bit of slow turn-on. This allows a lot more power to get to the load; in oscillation mode my inline DMM reads about 170 -200 mA, but when I use the optoisolators and get a clean turnon on the drain signal the current goes up to 1.8 or 2.0 amps or more, and the load heats like crazy. This also happens when "tuning", trying to get the oscillations to appear, when a clean turnon is achieved. I am formulating a theory about the chain of events during a typical NERD test session, if there is such a thing.

Right now I'm powering the optoisolator output side with a 9v battery; the only connection to the DUT is by the gate drive leads coming off the isolator collectors. I'm using an LED and its resistor as a "pulldown" for the optoisolator's phototransistor stage, so I can also monitor the pulses visually when they are slow enough. Very nice switching. Next will be to hook up the optoisolator's power needs to the main board battery supply, to see if the oscillations will return.

The only time Q1 and Q2 are both on at the same time is when I really crank the FG's output to some stressful amplitudes and offsets. I can get them both to saturate and stay on and that lets 3 or 4 amps thru to the load. Or if I use a symmetrical pulse positive and negative, at very fast frequencies there is "shoot thru" when both are briefly on during the zerocrossing of the gate signal. But when I mentioned it earlier, I meant both could be turned on, but alternately, with a symmetrical gate drive signal. I think. It is also possible to get the oscillations in both phases of the drain signal and I think both sets of transistors are partially on during this time.

If I can't reliably get the oscillations with the FG isolated like this, then I'll have to look for some other control condition that will allow me to compare _with_ function generator current path and _without_ FG current path.

MileHigh

  • Hero Member
  • *****
  • Posts: 7600
Re: Testing the TK Tar Baby
« Reply #34 on: March 27, 2012, 04:58:37 AM »
TK:

I did my last edit in the previous post I swear.  There is a code freeze on that posting.

I just wanted to say congrats on the big fat capacitor.  It can always be used as a low-pass filter in conjunction with your batteries to make a rock rock steady power consumption measurement for the entire circuit also.  In that case of course you lose the "purity" of powering of the circuit by a set of batteries and it is instead powered by the batteries transferring current into the capacitor first.  So in theory the circuit is being powered by the big fat capacitor.  You will still see the identical magic oscillations though.

MileHigh

TinselKoala

  • Hero Member
  • *****
  • Posts: 13958
Re: Testing the TK Tar Baby
« Reply #35 on: March 27, 2012, 05:05:16 AM »
Oh, I want to make sure that it's clear that I'm not monitoring the CVR quantitatively yet; the current values I'm giving lately are just the rough figures from the cheapo DMM inline with the battery.

It's nice to have more than two channels sometimes. Sigh.

@MH:
It's late and I don't think I'm up for much more tonight. I'm going to have to think about this AC issue a bit more. Would one expect a LED to light up in both polarity orientations, if inserted where the CVR is? Or a back-to-back LED pair?

MileHigh

  • Hero Member
  • *****
  • Posts: 7600
Re: Testing the TK Tar Baby
« Reply #36 on: March 27, 2012, 05:16:12 AM »
TK:

Go "dodo" and get a good rest.

For the LEDs, you want to try putting a back-to-back pair in series with the CSR.  When you get the oscillations, both LEDs should light up.

MileHigh

picowatt

  • Hero Member
  • *****
  • Posts: 2039
Re: Testing the TK Tar Baby
« Reply #37 on: March 27, 2012, 05:22:18 AM »
TK,

Using the opto's would allow full turn on, and indeed slower turn off, so I can see how they could overlap, but, that deviates from the RA circuit operation.

Are you using the similarly short Q1 on time/very long Q2 biased on time?

Your alternate FET's should work, but for duplicating as close as possible the waveforms and Fosc of the RA circuit, the PG50 would be a closer replication.  You should be able to swap them out with the 830's but the lower capacitances of those devices will flavor the results slightly.

If RA knew what the open circuit votage of the FG was during a test deemed "successful", (i.e., Q2's bias current) an appropriate battey supply equal to the FG open circuit voltage in series with a 50R replacing the FG would have performed similarly (as far as Q2 is concerned).  Of course, the new bias battery would discharge at a rate equal to the bias current and possibly from any assymetrical AC not fully bypassed by the Ciss of Q1/Q2 (though if assymetrical the correct way, possibly the new bias battery would get a slight boost now and again).

I haven't heard if .99 is going to finish his duplication or not, it would be interesting if the two of you could do similar reps and acheive similar measurement results.  But hey, its a lot of time and expense.  More power to both of you.

To infinity and beyond.........

PW



TinselKoala

  • Hero Member
  • *****
  • Posts: 13958
Re: Testing the TK Tar Baby
« Reply #38 on: March 27, 2012, 05:59:35 AM »
Blah... I made a "tad bit" of a math error earlier. My new load resistor stack of 5 x 50 Ohm 12 Watt resistors in parallel is of course a 10 ohm (measured at 10.3 ohms on the Simpson) resistor capable of handling 60 Watts, not the 250 I somehow came up with earlier. It's immersed in mineral oil though, so hopefully it will be OK.

But I caught the error first.....  or at least you lot are too polite to razz me about it.....   :o

Anyhow, yes, and good night.

I'm uploading a video describing the optocoupler test so far, but it will be an hour before it's ready, probably.

The waveforms that Ainslie showed are symmetrical, 50 percent duty cycle square waveforms, with the gate drive signal going from 0 down to -5 volts or so. I don't know quite what you mean, PW, about "the similarly short Q1 on time/very long Q2 biased on time". 
With only two channels on my scope I'm really only guessing, based on heating, which mosfets are on when. If I see high currents and the drain signal staying low without sign of pulsing, then I think that at least one mosfet is staying totally on. If I see the oscillations and partial drawdown in "both" phases of the drain signal, then I think that all the mosfets are at least partially turning on, Q1 and Q2 alternating. But I don't have the precision with this kit to be able easily to draw out a full timing diagram for the switching...yet. I have a Tek DPSO at the lab that I'll torture with the Tar Baby later this week and that will tell me a lot... a lot that you lot probably already know.


TinselKoala

  • Hero Member
  • *****
  • Posts: 13958
Re: Testing the TK Tar Baby
« Reply #39 on: March 27, 2012, 06:45:56 AM »
Musings engendered by me watching my latest video...

So you come into the lab and get set up. You still don't understand the reason for the oscillations but you know if you "tune" long enough they will eventually appear and if you don't touch anything they will stay.
So the load is in the water in the insulated teapot. You log the temperature. You turn on everything and begin tuning. As you tune for the oscillations, you will be going through periods of mosfets totally off with drain trace HIGH, periods of clean switching with drain trace in phase with gate signal from fully ON to fully OFF without oscillations, and periods of mosfet(s) constantly fully on, with drain trace at baseline and no sign of switching even though the gate signal indicates normally.
Then, suddenly, the oscillations appear. You cream your jeans, then you look over at the load temperature. It is amazingly high, like 80 degrees when ambient was 16. You log this, and back off to watch. Hmm.... the oscillations continue, with the drain trace oscillating around the _HIGH_ level as is normal for these oscs. The CVR trace indicates whatever weirdnesses. This is the data you log. Meanwhile, the temperature continues stable, no further rise. So you... retune,change freqs or something. First you lose the oscs, you fiddle around, there are periods of clean switching and  continuous on. Then you reestablish the oscillations and notice that the temp is much higher now, maybe even boiling. Whoopee, the magic oscillations are doing their thing!!

But what you don't realize is that the heating happened during the TUNING, when periods of high current flowed from the battery to the load, and this data wasn't recorded, since you didn't have oscillations and your circuit isn't operating in the magic mode. The load warmed up while you were tuning, and the slight power transfer during the magic oscillations is enough to _overcome the insulation leakage_ and keep the load from cooling off.

Retune, lather rinse repeat. And you've fooled yourself into thinking that the oscillations contribute to the heating of the load, when the major part of the work was actually done during the unmonitored and unanalyzed tuning stage.

It's just a musing, just a guess, a conjecture.

hoptoad

  • Hero Member
  • *****
  • Posts: 1009
Re: Testing the TK Tar Baby
« Reply #40 on: March 27, 2012, 11:29:37 AM »
snip..
The optoisolators work too well. They cause the mosfet(s) to switch cleanly and more precisely, if with a bit of slow turn-on.
 ...snip..

 LOL - Damn those pesky things doing what they're designed to do, and doing it so well !!.
 
 Cheers

picowatt

  • Hero Member
  • *****
  • Posts: 2039
Re: Testing the TK Tar Baby
« Reply #41 on: March 27, 2012, 01:54:27 PM »
TK,

It is difficult to know which circuit/claim to test from all the ambiguities and versions.

Regarding the duty cycle I mentioned, I was indeed incorrect with regard to "seconds".

In an RA paper, RA states that "the offset of FG was set to its extremel negative limit".   Also, "the duty cycle is also set to the limit of the FG's shortest on time within each switching period of 2.7 minutes".  (Additional data like FG hi/lo open circuit voltage swing, duty cycle in absolutes, etc are not provided)

I believe this is the setup that RA and .99 were discussing regarding a test of COP=infinity and the scope shots I have seen discuused regarding same.  I could, however, be wrong, as there is a lot of "noise" out there.  Q1 is basically a small player (its Ciss does add to the AC current path during Q2 oscillations) and the Q2 array could have been just as easily biased on with a PWR supply through 50R.

I do not know the "minimum" ON time of RA's FG, but from her paper and at a 10% "on" time, that would be 16.2 seconds of Q1 "on" time alternated with and followed by 145.8 seconds of Q2 biased on time, ad infinitum.

Setting your FG to a long period and minimum duty cycle, output set to -15 volts open circuit during its lo period, and who knows what during the hi period (+5??) would likely produce similar results.  Depending on RA's FG "hi" period voltage setting, Q1 may have never been turning on.  RA states that no current flow was measured during the FG hi period, so possibly the FG output during that time was below the Q1 gate threshold voltage.

MH is correct about the variability of MOSFET capacitances with voltage, that complicates analysis/prediction a bit...  As per some of the RA statements, the FG's offset/level was used to tweak the oscillation/power level.  This would affect both the DC bias setting and vary capacitances a bit a well

At a DC bias setting below 300 ma, the pwr dissipated in the load resistor is minimal at DC, so I suspect the bulk of any greater power dissipated at the load would have been via AC curents from the oscillation.

Ever work on a 60-70's audio amp with a load connected and a dreaded AC oscillation screaming away?  Things get very hot very fast (until one of the outputs hang and then "poof").  Ah... the early days of discrete, those were the days...

PW





 

TinselKoala

  • Hero Member
  • *****
  • Posts: 13958
Re: Testing the TK Tar Baby
« Reply #42 on: March 27, 2012, 02:55:13 PM »
I'm starting to get confused here. The Ainslie demonstration video shows their function generator set to produce a symmetrical square wave, negative going only, with a frequency of 10 Hz. This is confirmed by the shot of the Instek's panel, showing the figure "10" and some change and by the timebase of the scope shot (13) which is set to 40 ms/div, and the period of the FG's pulses is twelve minor divisions or 100 ms, as confirmed by the cursors (21, 20, 17). They claim in the video that this setting produces heating and battery recharging.  Where did these LONG periods come from that PW is referring to? 

I would much prefer to operate with the same parameters that they have actually _shown_. We know that their reports are unreliable, but scope traces do not lie. They may misdirect and provide cumulus-cloud-like material for projected imaginings, be rife with artefact and irrelevancies... but they do not lie. We know that much of what is "reported" by the NERD RATS is tainted in one way or another, by misobservation, improper interpretation of instrument indications, bad "calculation", math errors, anecdote and post hockery. This is why I tend to disregard everything (especially their "explanations") except what bits of raw reproducible data that can be gleaned from the dross.

Of course I _can_ operate at such weird duty cycles and long periods.... it's just a hassle with this analog scope I have here. And I don't want to generate miles of chart recorder paper like I did with Steorn's eOrbo farce.

Here, once again, is the scope shot from their demonstration, showing what they claim to be a fully operating NERD RAT device in oscillating, battery-recharging, load-heating mode.

Or will they  now claim that this does NOT show an operating NERD device.... ??


picowatt

  • Hero Member
  • *****
  • Posts: 2039
Re: Testing the TK Tar Baby
« Reply #43 on: March 27, 2012, 03:09:18 PM »

TinselKoala

  • Hero Member
  • *****
  • Posts: 13958
Re: Testing the TK Tar Baby
« Reply #44 on: March 27, 2012, 03:15:48 PM »

 LOL - Damn those pesky things doing what they're designed to do, and doing it so well !!.
 
 Cheers

 8)
I actually was able to get some oscillations while using the optoisolators. They are less consistent and make a more interesting (to me) pattern than the constant squeal of the certified Ainslie NERD oscillations, and are harder to "tune". I think that if I can figure out how to get the opto power and ground return from the main battery supply, this might help to reestablish the genuine NERD feedback. Also, of course the PG50s may behave differently.

Two scope shots showing the oscillations obtained _with_ the opto isolators: