TVS diode selection for 400v transistor

[John123]

Hello again, I'm in lockdown and have gone flyback mad!

What TVS reverse stand off and breakdown voltage would you guys select for a 400v transistor?

I've gotten myself confused when it comes to TVS datasheets. I'm running a TV flyback transformer like in the pic at 12v, it's actually quite efficient with nothing getting that warm drawing about 2 amps from the 12v supply and a 2 inch arc (still working on 24v mode). Frequency is about 15khz.

The TVS will not conduct under normal operation as the capacitor is selected to limit the peak voltage to something sensible (about 170v peak right now when the output is unloaded), the TVS is just added protection for unloaded flyback output and capacitor charging situations.


However by reducing the capacitor value and allowing the primary voltage to peak higher I can get more hv on the output without needing to increase the supply voltage, so I want to select a TVS which protects the transistor but doesn't conduct under normal operation.

But here's where I'm stuck, I've currently got a 200v TVS in there and when I retuned the capacitor to allow for more peak primary voltage the TVS clamps much harder than I expected. I'm looking at the datasheet for the P6KE200 and was expecting maximum peaks of 274v, but reality it clamps pretty close to the 200v breakdown rating rather than shooting up that large exponential curve.


I'm guessing the energy in the circuit isn't enough to push the TVS to such a high clamping level? And if that's the case what TVS voltage can I safely go up to for protecting my 400v NPN transistor (MJE13007)? Maybe the 300-350v standoff versions would be best suited, they're pretty expensive so I wanted check before ordering something which might not be a good choice, then there's the whole transistor safe operating area to further complicate things.

tl;dr which TVS diode breakdown would be a good choice for protecting a 400v transistor switching 15khz tv flyback at 2 amps average primary current?

Thanks.

[ritaismyconscience]

You want higher voltage across primary for obvious reasons but the voltage at peak current is still smaller than 400. The 250V one definitely will work, but you could probably go up to 300 to 350V because the current probably won't be close to 1A.

[John123]

You want higher voltage across primary for obvious reasons but the voltage at peak current is still smaller than 400. The 250V one definitely will work, but you could probably go up to 300 to 350V because the current probably won't be close to 1A.

Yeah I that's what I'm getting at, the energy in the primary appears to be low enough that the TVS does a pretty good job at clamping near its breakdown voltage (under forced conditions by reducing the capacitor value), I think 210v was the highest I saw using a 200v TVS and the TVS doesn't really get warm doing that.

But the TVS diodes all have such a wide clamping range in the datasheets that it makes selection confusing for amateurs like me. Like I said under normal operation it won't be allowed to conduct, but I still want it to protect the transistor from going over 400v in the event of something going wrong, the oscillator section is still pretty experimental so lots to go wrong.

[davekni]

Yes, worst-case specifications for TVS diodes don't allow tight clamping.  I've purchased quite a range of TVS diodes, finding them handy to have for my personal stock.  All have been quite close to their nominal voltage at low current and room temperature.  So, a P6KE300 is likely to have a breakdown right around 300V.  However, that voltage will depend on current, and especially on temperature.  The clamping voltage and current is specified with a total energy that heats the diode to its maximum 150C or 175C.  As with most silicon avalanche-breakdown voltages, it goes up with temperature.  So, the P6KE300's worst-case clamping voltage of 414V might not quite protect your 400V FET.  That depends heavily on the TVS current pulse duration and repeat frequency - in other words how hot the TVS diode gets.  Also, if your FET is running hot, it's breakdown voltage will be above 400V.  And, some FETs have some avalanche energy capability.  Overall, P6KE300 is probably the best choice.  Definitely not P6KE350.  Use the P6KE250 if you want to be extra safe with clamping voltage.  Still, be careful about letting the TVS do too much clamping, keeping it under its average power dissipation rating.  Or, go to 1.5KE300 for more power and higher peak current capability.

[John123]

Thanks, other than the higher power rating are there any other tradeoffs between the P6KE and 1.5KE versions? I'm guessing extra capacitance and leakage current. Then there's uni polar vs bi-polar, there's going to be tradeoffs between them too I take it.

It's actually a 400v NPN bipolar transistor (MJE13007) which I'm using hence the extra measures with the TVS being taken, it forms part of a cheap and nasty spare parts flyback driver (apart from the TVS).

Update: Found a P6KE250CA and when that is allowed to clamp it clamps pretty sharply around 258v, safe to proceed to the 300v version then?

[davekni]

The 258V clamp voltage will increase if the diode is required to clamp repetitively, enough to get it hot.  But that is a great ratio from 250V rating to 258V clamp even for single-pulse, a good low-impedance part.

Yes, I'd go for 300V.  It isn't quite guaranteed worst-case (414V), but likely to be fine.

The best I can tell, the 1500W diodes are equivalent to 2.5 paralleled 600W diodes.

The only real down-side to bidirectional TVS diodes that I'm aware of is slower clamping in the ns range due to forward recovery time of the forward-biased diode.  Capacitance is generally lower for bidirectional TVS diodes due to being two in series.  The clamping speed should be no issue for your use, since the flyback capacitor limits voltage slew rate already.

[SteveN87]

It's actually a 400v NPN bipolar transistor (MJE13007)

The datasheet quotes VCEO = 400V and VCES = 700V. What do you have to do to get the 700V rating? Is it as simple (!) as having the driver clamp the base voltage to 0V during the off time?

[John123]

The 258V clamp voltage will increase if the diode is required to clamp repetitively, enough to get it hot.  But that is a great ratio from 250V rating to 258V clamp even for single-pulse, a good low-impedance part.

Yes, I'd go for 300V.  It isn't quite guaranteed worst-case (414V), but likely to be fine.

The best I can tell, the 1500W diodes are equivalent to 2.5 paralleled 600W diodes.

The only real down-side to bidirectional TVS diodes that I'm aware of is slower clamping in the ns range due to forward recovery time of the forward-biased diode.  Capacitance is generally lower for bidirectional TVS diodes due to being two in series.  The clamping speed should be no issue for your use, since the flyback capacitor limits voltage slew rate already.

Thanks again! I learned something here.

BTW don't you just hate it when circuits perform better as a point to point rats nest of wiring and alligator jumpers than on a nice PCB?

It's actually a 400v NPN bipolar transistor (MJE13007)

The datasheet quotes VCEO = 400V and VCES = 700V. What do you have to do to get the 700V rating? Is it as simple (!) as having the driver clamp the base voltage to 0V during the off time?

Yeah I think it is something to do with that, in my circuit the base is clamped to just under its negative breakdown voltage via some zener diodes.

It's a bit confusing at any rate, during my testing I accidentally went over the 400v rating before putting the TVS on and it broke down and went short instantly. It seems very fragile to breakdown compared to other bipolar transistors I tried which'll go into a destructive clamping mode and get very hot.

[davekni]

Yes, Vces (700V) just needs the base pulled down to the emitter voltage to sink the collector-to-base leakage current.  I'm rusty on my semiconductor physics, but having the base below emitter voltage by almost Vbe breakdown might actually make things worse than 0Vbe.  I need to consult my brother, as he is more versed in such matters.  I'd suggest clamping the base voltage to one diode-drop below emitter just to be safe.  There's no value in going farther negative.

The Vces is typically a static (DC) test.  There may also be issues during turn-off, when there's still some minority carriers in the base region (base storage charge).  Again, I'll need to think about this more and/or consult my brother.

[John123]

having the base below emitter voltage by almost Vbe breakdown might actually make things worse than 0Vbe.  I need to consult my brother, as he is more versed in such matters.  I'd suggest clamping the base voltage to one diode-drop below emitter just to be safe.  There's no value in going farther negative.

Thing is it eats into the output of the flyback if I just clamp it with a single diode, there needs to be some wiggle room there to prevent this from happening. The zener + diode idea was to clip it just before the emitter-base reverse breakdown happens (not related to the TVS in my original question).

BTW are TVS diodes better than zeners for low voltage repetitive waveform clipping in the hundreds of milliamps range? I'm using an 8.2v 1N5344B in anti series with a UF4004 diode between base and emitter to clip before the emitter-base junction does.

[davekni]

The MJE13007 data sheet has a graph called "Maximum Reverse Bias Switching Safe Operating Area", figure 7.  It shows the allowed current at turn-off vs. Vce, with different graphs for different levels of Vbe, 0V, -2V, and -5V.  I see that it's better with more negative Vbe.  Hadn't ran into that situation before.  It may improve farther to -9Vbe, but it's also possible that the curve would reverse direction and allow lower Vce at -9Vbe.  If the Vce punch-through voltage is only slightly above the 700V avalanche-breakdown voltage for Vces, then the more negative base could bring the punch-through breakdown voltage below 700V.  So, -5Vbe is good, and -9Vbe MIGHT be better.  (Punch-through is when the reverse-biased depletion regions for C-B and B-E meet in the middle of the base region.  It's a different voltage breakdown mechanism than the more common avalanche breakdown.)

Notice that -9Vbe is the maximum allowed.  An 8.2V nominal zener plus diode forward drop will be typically about 9V.  Worst case it could be a bit higher.  For a one-up hobby project, that's likely fine.  Might drop to a 7.5V zener if making more than one.  One other consideration from my brother Dan:  Many bipolar transistors have their properties, particularly beta, degrade when operated with reverse Vbe for extended periods.  This is true especially when Vbe is close to breakdown voltage.  Dan's experience is with high-frequency BJTs, not power devices, so this may or may not be significant for this device.

That zener looks plenty capable for several hundred mA.  No need to use a TVS instead.

[John123]

Wow thanks for the deep level info!

I can confirm what your brother said about the gain degrading, originally the MJE13007 measured a gain of 25 on my cheap chinese tester, but now its only 19 (permanently). Definitely some reverse breakdown damage being done, although it hasn't further degraded below 19.

I'm starting to think the MJE13007 is a standout case, I just swapped it for an MJ15003 to test (retuned for less peak collector voltage of course!) and noticed the flyback output didn't change if I used the zener diode on the base in series with the UF4004 (as opposed to just the UF4004 on its own clamping it to 0.7v negative).

In fact the circuit with an MJ15003 performs much better than the MJE13007 ever did (less waste heat too), more output voltage for a lower peak collector to emitter voltage.

I notice the MJ15003 is designed for linear applications such as audio amplifiers and boasts a large safe operating area, whilst the MJE13007 is technically geared toward switching power supplies. In ATX power supplies I've seen the MJE series used in they've always got them in pairs with base drive transformers supplying base current, makes me wonder if the weird reverse base voltage effects are down to this design requirement.

Still seems weird that an audio output transistor performs better than a SMPS part, I guess it's down to how the cycle is terminated via the transistor coming out of saturation or core going into saturation.

I think my topology might technically be some kind of ringing choke converter or blocking oscillator.

[davekni]

Do you have any MJE13007 parts you haven't used yet?  You could try it at -5V instead of -9Vbe.  The beta degradation is tied to reverse BE leakage current, which typically rises sharply as you approach the -9Vbe spec limit.  I don't know if switching speeds or any other parameters degrade with beta.  You might find that an MJE13007 part operated within recommended conditions works well.

Yes, MJE13007 is an unusual part relative to my experience.

[John123]

Do you have any MJE13007 parts you haven't used yet?  You could try it at -5V instead of -9Vbe.  The beta degradation is tied to reverse BE leakage current, which typically rises sharply as you approach the -9Vbe spec limit.  I don't know if switching speeds or any other parameters degrade with beta.  You might find that an MJE13007 part operated within recommended conditions works well.

Yes, MJE13007 is an unusual part relative to my experience.

I've got 2 MJE13009's untouched by my transistor abusing hands, I guess I could try them and just adhere to a strict negative turn off bias from the get go with suitable zeners.

Do you think the gain degradation happens instantly over a couple cycles or gradually over the course of a few minutes, seems a bit odd how according to my chinese tester the gain dropped from 25 to 19 and no more after that. Its like its reached an equilibrium, maybe more reverse turn off current would push it further.

Unclamped the reverse voltage e-b acted like a zener with very poor regulation, I think I saw -11.3v across it at one point with a 12v supply.

Update: I finally plucked up the courage to scope the circuit with an arcing flyback connected, firstly you were very right about the 8.2v zener being too high. I'm seeing it clamp at 8.6v + UF4004 so yeah too much for the transistor, I've dropped to a 7.5v zener now and it clamps the base with some safety margin.

Secondly the MJE13007 seems to be very slow to turn off in this circuit without additional base drive components, the MJ15003 works great out the box and produces a much higher dv/dt rate for more power on the output. The peak drain voltage can be the same but the rate of collapse seems to be key for bigger arcs.

[davekni]

John,

Probably best to switch to a part that is working well.  Turn-off speed is certainly important for flyback use.  The MJE13007 inductive turn-off characteristics are specified with -2.5A base drive.  With lower base-drive currents, other transistors designed for higher-frequency might function better.  (My guess as to why -5V is better for this part is that internal base resistance requires that much voltage to get the -2.5A recommended for inductive load turn-off.)

I don't know if the beta degradation has a plateau or not.  In general it's a cumulative process, so more use should make it worse.  Since the goal is usually to avoid degradation, data on extended degradation probably isn't valuable enough to measure.

[John123]

Thanks!

btw a question about hv flyback transformer safe unloaded output voltages, is there a safe maximum in relation to their rated final anode voltage where it won't stress the insulation too much? As they're getting scarce around here and CRT TV's and monitors are nearly all gone, I don't want to overvolt and damage my remaining good ones like I've done in the past and destroyed them chasing after the big volts.

Let's say the HR diemen equivalent flyback is rated for 25kv, which in terms of arc length is just over 2cm between pointed needle electrodes (ignition, not drawn out). What would be a safe maximum ignition distance I could push it to, 3cm? I had another 25kV rated flyback jumping a 7cm gap for a brief time before the insulation failed, but that was for like 10 seconds, I guess the max safe distance is going to be lower for continued operation.

I can't really tell until it's too late lol. Are pins and arcing to the core a good indication of over voltage stress? One of them starts trying to arc from around the HV out area to the core at 35kV unloaded.

Here is the chart I'm following

[davekni]

John,

For how far to push voltage, I don't have any good advice.  There's likely an exponential function of life time vs. voltage, perhaps something like 10th the life for every 10% above rated voltage.  But I have no idea on the actual factor, and it likely is a steep function of temperature as well.

Concerning pins arcing, I'm guessing there's a pin that needs to be grounded.  Many CRT flyback transformers have an internal high-value resistor from either the HV output or from a lower-voltage tap (ie. after the first diode in the output winding/diode string) to a pin.  This pin is used with an external usually-adjustable resistor for focus voltage and/or other grid voltages.  )Some flybacks have the adjustment POTs integrated with the flyback.)  The resistance is 268meg on one part I checked.  I think some parts had two such resistors, but none I have now do.  This pin should be tied to ground, or through a resistor to ground to measure output voltage.  If left floating, it does tend to arc, and that may be internal rather than external, causing failure even below rated output voltage.

[petespaco]

I wonder if electron microscopes might use/still use a flyback transformer.
I googled:
flyback transformer for electron microscope
and got several hits for companies that appear to make and/or sell them.

We used to have a machine that worked sorta like an electron microscope, at 20,750 volts (supposedly, just below the level where x-rays would become a problem).  Instead of scanning, it wrote characters about 120 microns high onto our special dry processed microfilm  (3M).  At the time, one of my jobs was to make sure that the flyback transformers were up to spec.   We had them made by Setchell-Carlson in St.Paul, Mn., in small quantities, of course.
I visited their lab a couple of times to work out specific issues.  They really knew what they were doing.

FYI, in case you are interested, here's a webpage that describes that machine:
https://spaco.org/History/3M-Graphic-Systems-Hardware-History/115EBR.htm

Pete Stanaitis
---------------
 

[John123]

Ah thanks! My thinking with leaving it floating was to reduce loading on the output end.

The pin for the bottom of the divider resistors and HV return (ABL) pins 11 and 7 are about 1cm apart so that makes for about 10-15kV between them when driven hard, this is for a TV flyback. I think this is the schematic.


Although it could also connect to the final anode like this one, this is my thinking for the reduced performance of monitor flybacks when we use them in these homemade drivers. The HV out is always being loaded down (this schematic is also from a TV model however).


Do PC monitor ones have any differing drive requirements such as frequency and drive waveform, maybe the core material is of a different composition. I've noticed they never seem to put out as much when external primary coils are wound on the exposed part of the core for low voltage driving. Like you're lucky to get more than an inch ignition out of a monitor flyback for the same amount of input power as a TV flyback putting out 2 inches +.

I wonder if electron microscopes might use/still use a flyback transformer.
I googled:
flyback transformer for electron microscope
and got several hits for companies that appear to make and/or sell them.

We used to have a machine that worked sorta like an electron microscope, at 20,750 volts (supposedly, just below the level where x-rays would become a problem).  Instead of scanning, it wrote characters about 120 microns high onto our special dry processed microfilm  (3M).  At the time, one of my jobs was to make sure that the flyback transformers were up to spec.   We had them made by Setchell-Carlson in St.Paul, Mn., in small quantities, of course.
I visited their lab a couple of times to work out specific issues.  They really knew what they were doing.

FYI, in case you are interested, here's a webpage that describes that machine:
https://spaco.org/History/3M-Graphic-Systems-Hardware-History/115EBR.htm

Pete Stanaitis
---------------

Wow that looks interesting, cool website too! What sort of thing went into designing and coming up with those HV flyback transformers? I assume they were specific and unique to every CRT or else there wouldn't be so many different makes, models and variants out there!

[SteveN87]

This might be of interest (I bought one a while back):

https://www.amazon.co.uk/ZREAL-Flyback-Transformer-Engraving-Cutting/dp/B07F6FRX6C

No data supplied and couldn't find much on the web.

Haven't destroyed used it yet, but have done some measurements:

Primary inductance: 7.7 mH
Primary DC resistance: 0.6 ohm
Core: probably 3C90, Le=186mm, Ae=215 mm^2, no gap

The core is just held in place with a bit of hot melt, so it's quite easy to remove. I added a 0.85 mm (per leg) spacer, which makes the primary inductance 475 uH. With that gap, a separate winding of 22 turns measured 92 uH.

As a lockdown exercise, I'm simulating a general-purpose flyback driver; I might build one and try it with this victim flyback...