Yet another UD2.x pulse skipper

[hammertone]

All,
Everybody wants longer sparks, I am one of them, and this directed my mind towards pulse skipping mode of operation. There are a few of these boards around, some of them rely on microprocessors and high-speed programming language, others use simple logic chips.
Some of them simply turn off all gate signals when the current exceeds the trip point, merely transferring the surplus charge back to the buss caps, and then turns them on again when the current drops below the trip point. Others cycle the gate signals, shutting down only one leg of the bridge at the time, in order to pass the extra charge through the tank.
As far as I am can gather, the latter scheme has been exclusive to the microprocessor equipped boards (UD3 and UD+) whereas the other, simpler scheme, is the one used in the logic chip equipped boards.
Since VHDL is out of my reach, I wanted a board equipped with logic chips, but I also wanted one that cycles the legs of the bridge, and this brought me back to Hydrons excellent post on 4HV.org back in November 2016:
https://4hv.org/e107_plugins/forum/forum_viewtopic.php?p=1&id=178230#post-178230
where he proposed exactly such a modification for the UD2.x boards.
A simulation it in LT-Spice confirmed that it worked:

Next logic step was a layout in Eagle, have it manufactured, and see how it would perform in hardware. My old Eagle V6.5 suddenly failed to produce the full set of gerbers, so I downloaded the most recent free version V. 9.6.2, and that was a good thing. What an incredible step forward in functionality the work from Autocad has brought to the package.
As Hydron suggested in the November 2016 post, I added an adjustable timeout watchdog, just in case that the interrupter should get stuck ON.
I also added a shunt potentiometer and a 3pin header for a led voltage meter, to display the current trip point on the front of the die cast case, since I tend to forget what I just set it to 5 minutes ago.
This is how the first iteration turned out:


The first layout has been performing very well indeed.
As you see in this capture of an unloaded tank circuit, the alternate on time performs as advertised:



It lacks the HFBR receivers (I swear by the IF fibers: the hardware is cheap, and the signal is visible as red light at the end of the fiber), but in the name of universality, I added them back in the second run.
All bypass caps moved to the bottom of the board, in order to get the positive end of it as close to the positive terminal of the chips. Perhaps not a big thing on such a small board, but still.



I never got used to soldering wires to the status leds so there is now a light pipe assembly to guide the light from the status leds out to the front plate of the enclosure.
Changed the diode footprints to something easier to solder by hand.

I like a non-ambiguous response from the indicator leds, and there were 2 Schmitt trigger inverters left unused in the 74AC14 chip. I used them to make the blue led light up to full volume, as soon as the interrupter signal goes high.  Later I added couple of fets, and a few passives, to do the same to the red led current trip indicator, as soon as the controller enters into pulse skip mode.
Testing this produced a messy board:



But even in that state, the board was fully functional, allowing me to test the tank with first a copper disk as a load:



And an iron frying pan as the load:



And finally, with the secondary in place:



I double routed the status leds, to bring back the header for remote status leds, light pipes may not be for everyone.
When I talked to Hydron about using the mod, he informed me about the quirks of the L8365 UVLO chip, so UVLO is now implemented with a comparator in the 3rd board run.
A note about the pinout of TL331 from Texas that I use. This was the first comparator in the sot123 package that was offered by Eagle, so I used it without a second thought.
 Later it came to my attention that this pinout is not very common, it looks like this:



More common pinouts are these:



I was sloppy and should have spotted this, as it would have given access to a much larger base of replacements, including the nice rail to rail type with internal hysteresis: NCS2200 from ONsemi.
Apart from TL331 from Texas, to my knowledge, replacement types with the present pinout include AS331 from BCD-semi, LM397 from Texas and TS391_R_xxx from ST-micro. They all offer common mode input voltage from a bit below Vss to Vdd-1.5 volts. So as long the reference voltage at the wiper of pot R20 is set to less than 3.5V (and it stops to have any useful meaning at above 2.38V) the comparator will function as expected, so plenty of headroom here. And the external hysteresis makes it possible to set a trip point that requires a forced power down to restart.
9V regulator exchanged with a 12V type, for more headroom in OC trip regime.
Finally, I changed the footprint of the output transistors from TO252-5-11 to TO252-4 because that is the footprint of the AOD609 transistors that I use. TO252-5-11 will still fit.

The 3rd board run looks like this



A note about diodes and UD2.xx's: all 18 diodes in the UD2.8PS are 60V 2A schottky's PMEG6020ER. It is always nice to reduce the number of different parts in the BOM.

Steven Ward has said he likes to see the UD2.X evolve, and still show its basic strength, Gao Guangyan has not returned my call, but my hat is off to him, for all his work. I used his schematic of the UD2.7A as the basis for this build.
At this point there doesn’t seem to be more that is worth changing, to justify another board run.
And yet, the same moment I wrote this, the urge to dump the adjustable inductor, and re-implement the undeniably elegant method of using a potentiometer to adjust phase lead crept into focus.
The Predictor was the first controller to use phase lead, with which I pioneered the concept back in 2009.
https://www.pupman.com/listarchives/2009/Jun/msg00216.html
I abandoned the Predictor when the UD2.7 appeared on the scene, because it was so much better.
In the meantime, an elegant front end has been proposed by David Knierim,

 https://highvoltageforum.net/index.php?topic=2054.msg15563#msg15563

and I intend to use it in a forth board order, which will then, with David on the list, be the most diligently credited board in tesla coiling history:



As soon as the 4th board run is in my hands, I will publish board layout, schematic, bom and gerbers for all to use.



Cheers, Finn Hammer

[davekni]

Others cycle the gate signals, shutting down only one leg of the bridge at the time, in order to pass the extra charge through the tank.
As far as I am can gather, the latter scheme has been exclusive to the microprocessor equipped boards (UD3 and UD+) whereas the other, simpler scheme, is the one used in the logic chip equipped boards.

Mike's version is a basic logic version that keeps extra current within the tank:
https://highvoltageforum.net/index.php?topic=2054.msg15348#msg15348
This is the thread of your link to my alternate input circuit with pot-adjusted phase lead.  Mike inverts phase of one leg rather than turning it off.  Has same effect on H-bridge output.  The leg with inverted gate drive is enabling IGBTs only when current is flowing through their anti-parallel diodes.  It makes no difference whether IGBTs are off or driven with inverted phase.  Would be trivial to modify Mike's version for turning off gate drive instead of inverted phase.  (Phase-inversion has one down-side that didn't come up in that thread:  GDT volt-second capability needs to be 50% higher to handle resulting longer times in high and low states.  Your version doesn't add anything to GDT requirements.)

Mike's version doesn't alternate legs for phase inversion, a nice feature of your circuit.  I mentioned that option in the other thread, but Mike decided not to include it.  Modifying Mike's to gate turn-off mode would make adding leg alternation easier.  Would still require adding one more gate package, so might not end up any smaller than your circuit.

As Hydron suggested in the November 2016 post, I added an adjustable timeout watchdog, just in case that the interrupter should get stuck ON.

Yes, I like that feature.  For my DRSSTC (designed before seeing this forum or anyone else's drivers), I added a timeout for feedback input too.  Idea was to handle a case where MMC failed shorted, resulting in current ramping indefinitely in one direction.  In hind sight, wasn't worth the extra work.

All bypass caps moved to the bottom of the board, in order to get the positive end of it as close to the positive terminal of the chips. Perhaps not a big thing on such a small board, but still.

Excellent feature, especially if using fast AC logic with its associated larger switching current spikes!  (Saw one reference in your post to 74AC14.  I presume that is used instead of 74LS14 shown in schematic.)  Are you using 74AC08's too?  With an added logic stage (compared to UD2.7), keeping delay low would help.  Reduces the delay that phase lead needs to undo.

BTW, is there a reason for IC13A, rather than using the output of IC3A directly?  The OR gate is just recreating enable FF output.  Is the delay helpful?

Later it came to my attention that this pinout is not very common, it looks like this:

The lack of standardization for SOT23-5 pinouts is quite frustrating.  I like that package for single comparitors and opamps and logic elements.

But even in that state, the board was fully functional, allowing me to test the tank with first a copper disk as a load:

Nice clean results!

[Hydron]

RE the OR gate, it might indeed be that the delay is helpful - it seems that I used an RC on its output as well in the original version I posted on 4hv.
At this point it's so long ago that I'm having to figure out how my own circuit works from the description I wrote way back then! To be honest I'd have to give it a good hard look to really fully understand it again.

Glad to see that it seems to do the job (I never built it myself) - looking forward to seeing some sparks!

[hammertone]

All bypass caps moved to the bottom of the board, in order to get the positive end of it as close to the positive terminal of the chips. Perhaps not a big thing on such a small board, but still.

Excellent feature, especially if using fast AC logic with its associated larger switching current spikes!  (Saw one reference in your post to 74AC14.  I presume that is used instead of 74LS14 shown in schematic.)  Are you using 74AC08's too?  With an added logic stage (compared to UD2.7), keeping delay low would help.  Reduces the delay that phase lead needs to undo.

BTW, is there a reason for IC13A, rather than using the output of IC3A directly?  The OR gate is just recreating enable FF output.  Is the delay helpful?

I am using 74HC's all through, and never thought about it. The AC's are faster, I see that now that you prompted me to look at the datasheets, and perhaps a good idea to shift to them later.
With regard to the extra OR gate in the chain, again, did not think too much about that, I just followed Thomas' schematic, and since it worked, saw no reason to change anything except:

RE the OR gate, it might indeed be that the delay is helpful - it seems that I used an RC on its output as well in the original version I posted on 4hv.
At this point it's so long ago that I'm having to figure out how my own circuit works from the description I wrote way back then! To be honest I'd have to give it a good hard look to really fully understand it again.

Glad to see that it seems to do the job (I never built it myself) - looking forward to seeing some sparks!

There were indeed RC-delays in line with the preset inlets to the flipflops



and I thought they were perhaps intended to secure the flops from entering in the so called unpredictable state described in the function table:





I tried to eliminate them, and saw no obvious change, so left it out from v.3 of the board.

Oh, sparks:
Just a short clip as proof of concept. 1.2 meters from 0.45 secondary 300uS ontime 100BPS and really not a setup to make long sparks due to space restrictions. I think it is time to commit the controller to one of the larger coils that I have handy.
https://youtube.com/shorts/-x4gwBCWoeo?feature=share

Cheers, Finn Hammer

[Hydron]

OK so I've figured it out - the reason that the output of IC3A isn't used instead of the OR gate (IC13A) is because this was originally designed as an add-on to an existing UD2.x board. I went for the minimum amount of connections to the existing board, so added a new gate (which I had anyway due to the need for OR gates elsewhere) instead of using the existing signal from IC3A.

If another design spin is done then yeah it should be taken from IC3A - this would eliminate the possibility of a glitch where both outputs from the phase-select jumper/switch are low for a short period from resetting the pulse-skip flip-flops. This potential glitch (which doesn't seem to occur in Finn's interrupter) is the reason for the RC filter in my original schematic sketch. If any problems are noticed with the current version then a small cap to ground on the output of IC13A should solve it (value is a compromise between not getting big current spikes on transitions and having enough C in conjunction with the R of the output driver in the IC to form a sufficient RC filter).

[hammertone]

OK so I've figured it out - the reason that the output of IC3A isn't used instead of the OR gate (IC13A) is because this was originally designed as an add-on to an existing UD2.x board. I went for the minimum amount of connections to the existing board, so added a new gate (which I had anyway due to the need for OR gates elsewhere) instead of using the existing signal from IC3A.

If another design spin is done then yeah it should be taken from IC3A

David and Thomas, thanks for your sharp observations. I have now parked IC13A, inputs tied to ground, and routed directly from IC3A pin5 to IC3B pin 10 and IC12A pin4



This goes to the board house together with the new phase lead front end.

Cheers, Finn Hammer

[hammertone]

After another 14 days, waiting for the boards to show up, I have now populated and tested the final version of the UD2.8PS, which features "Predictor" style phase lead adjustment with a fixed inductor and a pot, instead of the variable inductor seen in the previous versions.
I can report that it functions as intended, and this is what it looks like, preset for 300A of trip current and lodged in it's cast Hammond case:



And here you can see the phase lead arrangement:



As promised, here is a zip with the BOM in PDF, the schematic and board layout, as well as the gerber files, in case you should want to make a run for your own project.

UD2.8PS.zip

This has been a great project, My thanks goes to all who has assisted, and hopefully it will be usefull to others than myself.

Cheers, Finn Hammer