HV flyback design theory?

Years ago I made what I call my Raster-Hack to do just what you are looking for - running an XY video game (Star Wars) monitor from a raster HV as the XY monitor's HV circuit was toast and there was (at the time) no replacement for the xformer.

I caution you that whatever you decide to power your monitor to beware of the X-Ray potential if you get near the picture tube's rated maximum voltage. That was one reason I used the raster chassis to drive the tube, it had built in X-ray over-voltage sense and shutdown built in and would thus be safe for home use.

So, depending on the size and type of your XY tube, you may be better off finding a good video game chassis and using it for your project.

At least if it is a one-off...

John :-#)#

--
(Please post followups or tech inquiries to the USENET newsgroup) 
                      John's Jukes Ltd. 
MOVED to #7 - 3979 Marine Way, Burnaby, BC, Canada V5J 5E3 
          (604)872-5757 (Pinballs, Jukes, Video Games) 
                      www.flippers.com 
        "Old pinballers never die, they just flip out."

in

ed

You need to characterise the transformer. You need the inductance of each w inding, the mutual inductance between each pair of windings, and the stray capacitance associated with each winding (which can be hard to measure sinc e getting current flowing through one capacitance means having current flow ing through all the others).

The transformer equation is

V1= L1.dI1/dt + M.dI2/dt

V2 = M.dI1/dt + L2.dI2/dt

where M is the mutal inductance of the two coils and less than the square r oot of the product of L1 and L2

M = k. (L1.L2)^0.5

where k can be as high as 0.999 for high-permeability cores, and can get do wn to 0.98 for gapped ferrite cores.

LTSpice lets you use these numbers to model coupled inductors.

The self-capacitance of a single layer winding can be quite low - around 1p F. The self-capacitance of a multiplayer winding can be quite a lot higher. If you got some idea of the dimensions of the windings and the wire gauge used you can generate useful rough estimates. High voltage transformers can be quite bad.

Peter Baxandall seems to have invented the class-D oscillator while working up an inverter to generate high voltage photomultiplier voltages (around 1 kV) from 12V rails. It doesn't show up in his paper

but I worked for a guy who had trained under him, and that was the story I got from him, along with tales of dropping aluminised ping-pong balls from the bomb-bay of a Canberra bomber flying up in the stratosphere. Peter was working at the UK Royal Radar Establishment at Malvern at the time.

--
Bill Sloman, Sydney

Replace the yoke with an equivalent R+L (or maybe even not much R) and drive it at roughly the expected frequency. You can use a 3843, sure, get some nicely regulated A2 with a big fat resistor divider, without needing the tricks/hacks they used back in the day. Trace the circuit, use the original primary winding and voltages. If it's offline, that's probably 160VDC at

15kHz. Easy enough, if a bit of a pain if you're using a low voltage supply instead (use a boost? add external winding? isolator for SELV control to offline side drive?).

Bonus points for tracing the circuit to find what HV regulation feedback it had in the first place. Usually there's a focus/screen divider chain that returns to a not-quite-ground pin that can be used for sensing.

Incidentally, a lot of Trinitons did this already, a separate HV supply, since the horizontal sweep range was so huge. Some multisync monitors managed all in the same circuit (impressive!). Last one I had, still have the service docs for, HV supply was a MOSFET of all things, driving what looks to be a fairly ordinary flyback transformer, with a BA9756 controller. Not current mode, really weird circuit; it's parafeed through a fixed value inductor (not an RFC, its value is significant). Well, I suppose that's as good an indication as any about the first thing I said -- the inductance effectively in parallel accounts for the lack of yoke, storing energy that the transformer wouldn't otherwise. (FBTs aren't gapped very much, remember. Air gap is what stores the energy.)

And if you are using a different supply voltage, there's always the self-excited oscillator on some turns on the back leg of the core. May be harder to regulate though. Mind that, if you do use the same kind of circuit (3843 + such), you MUST deal with the leakage inductance of the shitty coupling from primary being outside the main windings. A resonant drive is preferred for this reason.

Even with good coupling, note the secondary is significantly resonant by itself; combined-HV multisync monitors pushed over 100kHz fundamental, with

10% or so of that being retrace -- the waveform is roughly class E, i.e. flyback but with a resonant hump instead of a square clamped pulse. That means the self-resonant frequency was around a MHz. Which itself is impressive for such high voltages; they used special winding techniques to achieve it. Regular TV FBTs won't, and will have a lower SRF (again, likely near a useful harmonic of sweep). A resonant load means awful performance for hard-switched flyback; it's best to take advantage of it instead.

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Design 
Website: https://www.seventransistorlabs.com/ 

 wrote in message news:r8fd0l$jl6$1@dont-email.me... 
> Hi all, 
> 
> I apologize for the probably too stupid questions, but it's a totally new 
> electronic territory for me and I don't seem to be finding the needed 
> theory articles and/or books to help me. 
> 
> I would like to re-use some old BW TV's HV transformer to power the CRTs  
> in 
> X-Y mode (all the yoke changes and drivers are already done). I'd just  
> need 
> the Anode supply and the G2/G4 supplies that are usually obtained with 
> a few taps on a "low" voltage secondary winding on the HV transformer. 
> 
> Now, what would be the step to design a proper driver circuit around a 
> given transformer? 
> I have quite a lot of test equipment and I can measure inductances at 
> various frequencies, I have pulse/function generator (HP-8116A), HV probe 
> and most of what it's probably needed, but can't find a good text about 
> this subject. 
> Of course I don't want to copy a typical horizontal yoke driver plus 
> flyback driver circuit. I've tried to use a big-ish air-cored inductor 
> in place of the horizontal yoke coil and it kind of worked well enough 
> but the inductor was getting hot quite soon. Also the typical TV driver 
> circuit has no voltage feedback, probably since the horizontal yoke is the 
> big load there and CRT's anode load won't be changing much the power  
> budget 
> of the converter. 
> My next attempt will probably be designing a PWM driver based on the 
> old UC3843 (current mode, with a voltage feedback from a low voltage 
> secondary). But I'm afraid there're too many unknowns still for me to 
> make a good design. 
> Is there any way to get the right knowledge to design such a supply? 
> 
> Thanks in advance. 
> Frank

G2 and G4 are "screen" and "focus" right? usually those are tapped of the trippler.

Use the transformer and circuit you already have, Find a different way to drive the yoke.

--
  Jasen.

The original TV is a 220V AC/12V battery 11" black and white design. All was powered with a regulated 11V rail. I have traced the original flyback circuit, it used germanium PNP driver and a germanium diode. It doesn't appear to have any voltage feedback. One low voltage secondary tap was brought to a TCA511 pin to probably sense the end of flyback period. I've tried to substitute the horizontal coil with another similiar inductance air-wound coil, but it became quite hot and anyway the HV wasn't quite right. The 11" CRT needs 11kV nominal anode supply and I was getting a bit less than

8 kV with that method. G2/G4 supplies were however correct more or less (no loading, just the tube capacitances, no filament supply when I did the test). So one thing that was really puzzling is: how come I get the correct low voltages and the HV is so much off?

G2/G4 are derived from a single secondary tap, rectified by a BA158 diode filtered by 22nF to ground then divided by a couple of 2M2 trimmers to set the two voltages. It was 1972...

One approach I've attempted was to use an astable 555 circuit, tuned to approximately 16 KHz, then I've used that secondary tap that went to the TCA511 IC to derive a voltage feedback to make the 555 "wait" for the transformer flux to go back to zero before starting a new cycle. That produced nice driver collector and base waveforms, no ringing almost, but also way too low HV (around 6kV), that was with no replacement yoke coil and experimenting with capacitor values in parallel to the flyback diode.

So are you saying that the transformer alone can't store enough energy without the original yoke coil?

I've seen self oscillating circuits, for example the HV supply for electrohome G05 XY monitor. It needs a driver's base secondary winding though, that I might add to the transformer, but well, how do you calculate the correct turn ratio? Coupling would also be way less than perfect and so on... Also the G05 has voltage feedback from a 90V secondary and it uses this feedback to regulate the DC input to the primary side, clever...

Is there a way to properly design a resonant drive without using an added secondary? Maybe using that low voltage tap I already have? Though that would be referenced to ground.

I definitely can see multiple peaks in the driver's collector waveform, no matter how I tried to drive it. Is there a good way to properly measure this secondary resonant frequency? Aren't all the windings contributing to resonance (and their loads too)?

Thanks a lot! Frank

Really, just drive it. If in the US, or who cares just measure the inductan ce and make it resonate at about 70KHz. The you can regulate by frequency easily. Not a problem but most of the transistors that handle that will be Asian origin, if that is a problem, if so you can feed a smaller winding. T hen European styles get cheaper as well as US ones. Hell I haven't bought o ne in a while so don't take that to the bank necessarily, but voltage is on e of the prime things for transistors.

Get you a signal generator that does square waves. Get a piece of metal for a heatsink and a variable DC power supply. Got your transistor at what the y intended as the primary (doesn't have to be but better if it is) and can pulse it with the generator and gradually bring up the voltage.

I approach it this way because you got a "bunch" or whatever of them. There was never any implication they were all the same. therefore it can be adap ted. But if they are all the same once adapted you know what you have to do and it is a piece of cake.

I the end if they are the same you got OK, 22KHz, 55 volts and it should pu ll XX mA. (those are just off the top, don't use them just figure it out fo r yourself)

NOW, that is licked,l what about the focus voltage and possibly a high G2 v oltage required ? Divide it down ? High resistors have a high price,l so it all depends on what you want to sink into this.

If you got low focus CRTs, and they exist down to zero volts, but can be as high as 25% of the anode, the lower voltage ones are advantageous in the d esign but in my experience the lose their focus as they age.

You can stack your own, but it really isn't much cheaper. You have to conne ct them and then they need to BE somewhere. The big square ones cost a fort une.

There are ways but to go into that we need to get into the parameters of th e CRT.

We can find out by the number. And IF your flybacks are all the same we can figure out what they do, many have resistors built in for focus and G2.

And if you want to control the intensity which you do, those voltages are m uch more manageable. You don't want to burn the screens. Now if you control it by G1 you get near infinite impedance, if you use K you get to use a bi t lower voltage but then there is a load. Not a huge load but something.

You work with what you got.

All the windings are coupled together. Just drive the secondary with a sine wave - through a resistor (or some other) impedance and see a what frequency the voltage across the coil peaks.

If the resistor is too low you will get a very broad flat-topped peak, too high and you won't see much voltage swing across the coil.

Compare the phase of the sine wave across the coil with the phase of the drive waveform - they will be in-phase at resonance.

There's not a lot of point in trying to work out where the parallel capacitance is - most of it is going to be between the turns of the secondary winding.

--
Bill Sloman, Sydney

Ah...

Probably didn't even have regulation, on account of B&W not giving a crap. Or possibly some sneaky primary side or boost feedback, who knows.

How similar? Hot means low efficiency. Maybe you were shorting something out. What did the voltage and current waveforms look like?

Keep in mind they liked to use coupling caps everywhere in those things, too, guess I should add that. Could well be that the FBT is designed for full bipolar flux swing, not unipolar switching directly into it. Which means it should still work at about twice the frequency without caps, which will need even less (about half) the inductance in parallel to keep the same peak current.

Yeah, that's the right kind of waveform. Mind if the transistor is BJT, it needs an antiparallel ("damper") diode (the original will either be right beside it, or integrated with the HOT). MOSFET has body diode, is fine. What's missing then is either enough on-time to charge to the required energy, or low enough inductance to do the same in the first place.

And yep, that's exactly what missing the yoke should do.

Because of poor coupling, a push-pull circuit is best. The reactive energy of the primary and leakage inductances are "stirred" into the supply, just as the damper diode clamps resonant energy in the quasi-resonant case. Whereas a single-switch circuit might have to dump all that energy into a snubber.

The base winding can be in the same place. Just put on two CT windings, one with few turns and wire size that doesn't matter, and another with a few more turns and a few amps worth of wire size.

Typical guess would be a few volts per turn, so out of a 12V supply, 3 or 4 turns (for each half of the primary) should do. Base off voltage should be above -6V peak (to avoid breakdown, causing runaway operation), so about half the turns or 1-2 will do there. Connect the CTs to supplies and you're off running. The base winding is usually supplied with a resistor, the current setting on-time or power output or the like.

Alternately, use MOSFETs with the gates driven from opposite drains, no need for a separate drive winding. Usually a series diode and pull-up resistor is used to limit gate on voltage.

Example:

This was a custom transformer so the voltage isn't incredible, but the circuit overall is actually a bit relevant...

The series inductor to the primary CT is generally beneficial, but it depends on what type of circuit one is building. Baxandall (as BS will regale you) needs it. Royer doesn't (commutation driven by transformer saturation).

You may find you need much higher supply voltage (or fewer turns) to get the desired peak output voltage this way. And if it's resonating at a much higher frequency (100s kHz), well, that's just what it's going to do; if you need to keep PRF below 20kHz say to avoid overheating diodes and stuff, you'll need to set up a gate circuit to do that.

Random windings and taps have random ratings. Bad idea. Stick to known primary.

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Design 
Website: https://www.seventransistorlabs.com/

ok, I've driven the smallest secondary tap with a 2k2 carbon comp resistor, sine wave output at about 3V.The peak is definitely at 36 KHz Though, if I drive the primary and measure the secondary voltage on the same smallest tap, I get the highest amplitude at 38 KHz, I'm not sure about this difference. Driving the primary with the same sine wave and measuring the anode output, through a 1Gohm probe not to load it too much, I get a peak again at 38 KHz (87V DC with 690mV RMS drive on the primary). How can we explain this resonance difference?

they remain pretty much in phase on a broad range around the resonance

so this 36 / 38 KHz is the self resonance of the high voltage secondary?

Thanks Frank

What is M for 3 coils?

this is as much as I tracked of the original circuit:

The horizontal yoke is 88uH, the schematic has an error (not so relevant) as R504 is a pullup on the TCA511 driver pin which is open collector, so it is infacts on the other side of the 220 ohms resistor.

I've found a big-ish 95 uH inductor, probably a power supply filter of some sort and that was wired in place of the horizontal yoke coil and the linearity coil was shorted out of circuit. Voltage waveforms here:

Top one is the driver transistor's collector (a BU124), 50V/div bottom one is the transistor base, horizontal is 10us/div. I didn't make more checks as the EHT was still too low to give me confindence in tweaking the circuit further.

well the original circuit seems enough simple, I don't think I've missed a series capacitor to the primary.

well, I've played quite a little bit with the pulse width and repetion frequency, but that didn't really give any good result.

thanks so much for the example! Should I use litz wire for the primary turns? Can these turns be added over the existing primary? That would make them quite far away from the ferrite. Or maybe it's better to try to "squish" them between the end of the existing winding and the start of the top ferrite horizontal segment?

Should I expect that this resonating frequency be close to the current resonance peak I've measured on the unmodified transformer?

yes makes a lot of sense, and I think it's feasible on the old 1970's high voltage transformers that weren't all potted.

Thanks again Frank

4+4 turns on the first random HV transformer I have, 2xIRF640 (I don't have the 644, but looks like it shouldn't matter much). HV is at 8.6kV huge step forward!!! The oscillation frequency went to 29.3 KHz. Now, to increase the output, can I safely go to 3+3 turns? The mosfets don't even get warm after one minute or so. Here's one drain waveform:

I've used a 150 uH inductor, doesn't even get warm, but I'm not loading any supply other than the HV (connected to the CRT to have the DC smoothed by the internal capacitor) with the 1Gohm HV probe.

I guess BA159 fast recovery will be ok for the low voltages? As for the HV diode string, that's internal to the potted part of the flyback, I hope it's not too upset for the 29 KHz now.

So, thanks a whole lot again! I was really scratching my head since many weeks on the wrong circuits :)

Frank

new

Ts in

need

ch winding, the mutual inductance between each pair of windings, and the st ray capacitance associated with each winding (which can be hard to measure since getting current flowing through one capacitance means having current flowing through all the others).

re root of the product of L1 and L2

Different. Basically, k is that proportion of the flux generated by the cur rent in the first coil that threads the other coil.

k for two coils that are wound as twisted pairs of wire are going to be clo ser to one than k for two coils that are wound separately, on top of or nex t to one another.

For a three coil assembly you'd have M12 = k12.(L1.L2)^0.5, M13 = k13.(L1.L3)^0.5, and M23 = k23.(L2.L3)^0.5 .

LTSpice lets you specify different k values for each pair of windings - if you want to. See the manual on "K mutual inductance".

--
Bill Sloman, Sydney

ok I've tried 3+3 turns as primary and the PC power supply that I'm using tried to start but went immediately into shutdown. I could see a +12kV indication at start. I then had the idea of trying a 1 ohm resistor in series to the positive supply to the transformer center tap, but that made the mosfets (both) very hot and made something smoke. I'll investigate later what smoked and I'll try someting like 3.5 + 3.5 turns. I'm not using the voltage and current limiting circuit *yet*, I think that would do the job of making the oscillator start and regulate the DC input, but I would first measure what voltage can I get from a secondary tap, then calculate a divider to serve as voltage feedback. Any better idea? Well, lots of experiments before I reach the goal, but I think I'm on the right track.

Frank

I have a few questions about this circuit, if you don't mind:

1) I don't have such a high Hfe BJT as the 2SD1273, I think a sziklai pair in place of it should be fine, since it seems to me it will never be saturated anyway, unless I'm missing something?

2) the 1/2 358 opamp has no DC negative feedback, that's well, unusual to me, I would like to understand better that part. Thanks

Frank

This is what we call an error amplifier. Its output can indeed be saturated

+/- if the load is taking its sweet time to catch up, or if the input is sudden and large. The 10k in, and 0.01 + 100k across it, sets the response time and gain (compensation). 2SD1273 can of course be replaced with anything of suitable rating, Sziklai, Darlington, MOSFET if you don't mind the lost Vgs(on); or even better, a buck converter -- the purpose is simply to make voltage at the primary CT. Or current at the CT more specifically, since the oscillator is current mode. Which means a current-mode buck, and usually a relatively large inductor so the current ripple is small, is ideal.

Eventually the voltage output responds to the error amp's output (changing the oscillator's supply voltage), and the voltage divider closes the DC feedback loop.

The feedback is also buffered, so there's an accurate x1000 output sense (I don't have HV probes, as it happens).

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Design 
Website: https://www.seventransistorlabs.com/

well yes I figured out it would be closing the loop with that time constant of 1ms, it was just the lack of DC path that was looking odd to me, but it's just me :) Nothing bad happens if the output saturates. New things are always welcome for learning.

ok I understood, I think I'll just stick with a sziklai pair since that will allow me to have a prototype sooner. I'm still a long way from the needed 11 kV.

yes I've noticed that, I won't be needing it since it's a bad idea to wire a long string of resistors down from the CRT anode. I'll be using a low voltage G2/G4 supply for feedback, if I can make it to work in the first place that is.

Thanks Frank

Ok, just a followup to complete the thread with results (failures), should anyone try the same route. I've built Tim's circuit (using a complementary pair in place of the high beta power NPN, so I don't lose too much Vbe if I had chosen a darlington for example). The current sense resistor has been left at 0.22 ohms (and there was no point in tweaking it, I'm going to explain why). I've employed a 150 uH 3A (at least) inductor, but I've tried some transformers without it (usually with same result or poorer). Input voltage is 12V and mosfets are 2 x IRF640. All added primary windings have been made with 0.8mm diameter enameled copper wire. I've tried a few HV (or flyback, as they call them) transformer from various BW TV chassis ('70s era), including the chassis that's the donor of the 11" CRT I wanted to use for XY display. These transformers are the "two windings in two legs" kind, with one leg having the primary plus low voltage (G2/G4 and maybe K, maybe heater) secondaries and the HV potted block on the other leg.

Transformer n.1: with 4+4 turns over the original primary was giving 8.6kV DC and oscillating at a bit more than 29 KHz. I then killed this one trying 3+3 turns before building the complete circuit with the current sense resistor. So I deserved to burn it I think.

Transformer n.2 (original from the CRT's chassis): 6.6 kV DC with 4+4 turns, goes to 25 KHz. Can go up to 7.2 kV DC lowering the current sense resistor. With 3+3 turns (and 0.22 ohms sense resistor) goes to 8.5kV DC but the core (not the windings) gets quickly hot. Now, on one of these transformer, the ferrite is engraved with "N27", so it might be indeed the TDK's N27 material. According to the datasheet, this should saturate at 500 mT flux density and with the approximate core's cross section that I could measure (vertical legs are almost all covered, plus they have a hole for the screw), turns out that I should have a bit more than 6 turns on 12V to not exceed the 500mT density. So it makes sense that I got the hot core with 3+3 turns. Even with 4+4 turns the core gets just a bit warm anyway. So even running the core at the saturation limit doesn't give the rated output (can the old HV diodes be too hopefully slow at already 25 KHz?).

Transformer n.3: very similar apparence to the n.2, goes to 42 KHz and outputs 5.8 kV DC, with 4+4 turns primary. No further tests made since it sounds and smells like it's lightly arcing probably internally in the potted winding.

Transformers n.4, 5, 6: oscillations at around 100, 160 and 206 KHz, in all these cases the DC output is around 1.2 kV (funny?). These have bigger core cross-section and apparently bigger HV winding, so probably were made for bigger CRT sizes. Low output is probably because of too slow HV diodes or too high frequency for the core's material.

So I think my best option here is to buy a ready made 10-15kV settable supply.

Hints and comments are welcome of course. Thanks

Frank

One more thing to try: put enough capacitance across the primary to make it resonate at about the secondary resonant frequency. Probably that's the extra peaks you were seeing. Because again, the leakage is high, so the primary can kind of do its own thing at high frequencies, but you want it to do what the secondary is doing.

Tim

-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Design Website: