Fun and games with pulse transformer

[throbscottle]

I've been trying to get the output half of this to work: http://www.kerrywong.com/2013/06/05/vfd-filament-driver-using-555/comment-page-1/ I particularly want to use a transformer so I can connect the centre tap to -24v or so (+ offset).
I have a green ferrite toroid 14 or 15mm diameter, originally had about 8 turns bifilar on it, came out of some sort of PSU. I tried some smaller ones first but didn't get enough output with them.
I wound it as per the web-site, 30T secondary CT, and 40T primary. Drove it using my pulse generator set to square wave, started with 100KHz but tried tuning up and down from there.
Output looked good open-circuit, dropped very low when connected to VFD filament. Tried more turns, voltage went down. Tried less turns, voltage went up. OK so impedance matching going on there. Tweaked to find best number of turns.
Fiddled with the primary next. Tried reducing the turns, but the secondary voltage went down! Ahhhhh, what's going on????
Ended up changing turns so much I lost count, but I think I've ended up putting an extra 20 turns on, so primary is now about 60T and secondary about 26T, and yet I have 5v going in and 2.2v coming out with the VFD filament connected. I can't get the voltage to go up any more except by increasing the supply voltage. Get a bit more output at higher frequency (about 142KHz), I'm guessing due to resonance, and a lesser increase at a lower frequency, somewhere in the mid 80's KHz. Also improved by changing transistor (don't have a BD131 like he uses, ended up with TIP31 I think)
I also experimented with the catch-diode. Tried putting it directly across the primary but the output drops completely, back in it's proper position for some reason it now makes no difference although it did before I fiddled with it
I think the VFD filament needs about 3v (don't have actual data for the display, it's a Samsung INRC09SS49T in case anyone is familiar with it.

So can anyone help me get more output from this circuit without having to so finely tweak the transformer (in fact, get it to behave more like a power transformer) and without resorting to a significantly higher supply voltage (increasing the supply voltage from 5V to 12V resulted in only 0.8V increase in transformer output). I don't mind making the circuit a little bit more complicated but not forgetting it's only to drive a filament so there's a whole bunch of stuff it doesn't need...

 

[schmitt trigger]

If you are running the transformer with an unipolar square wave, very likely it is saturation.

Other way to saturate the core is to exceed the volts-time product.
By your description that the secondary voltage does not change much as you increase the primary voltage from 5 to 12 volts, I believe that is a possibility.

 

[dr pepper]

No disrespect to mr wong but that circuit can be improved.
It looks like a forward converter but without the reset winding.
Both windings need to have Ac on them, ie as much time positive as negative, or at least the same volts x time for pos & neg cycles.
The circuit you showed us relies on the vfd heater for that, this is probably making it really hard to get the turns right, that and the fact your using a 'junk box' core.
2 things you could do, firstly get a core with power ferrit such as an N97 or N98 core, second try adding another primary with the same number of turns and put a diode directly across it, wired so that the diode coes not conduct when the primary transistor is on, then the trans should behave as a trans, ie turns ratio pri to sec should be what you expect.
Really you should calculate the core and turns, you could do this with software such as this (free):
https://320volt.com/en/smps-donusturucu-bobin-transformator-hesaplama/

 

[throbscottle]

So, perhaps I should try slower rise/fall pulses? Maybe that's why the site's author had good results with a 555 timer.
Saturation - am I right in thinking if I reduce the m/s ratio at some point it will be able to de-saturate between pulses?
Magnetics still a bit alien to me.
(edit) Ahhh dp your post appeared just when I posted this one! I will definitely try the extra winding + diode!

 

[throbscottle]

Hmmm, ok, how do you get it in English?

 

[dr pepper]

The pic on the link page shows the program in english, and mine is in english too, maybe theres a drop down you select english cant remember now.

You want the on pulse to be quickish and the off pulse to be a quick as possible (I'm talking about the rise time here), kerrys transistor drive circuit is probably good enough.

Saturation is when primary volts x seconds has gone above a vertain threshold, reducing on time or reducing voltage reduces the chance of sauration, another way to reduce saturation is to increase the number of turns, or replace the ferrite for one that can stand a higher magnetic flux.

 

[schmitt trigger]

Please google magnetic hysteresis curves, and understand where zero flux, remanence and saturation are located.

All these points are core specific. As Dr Pepper mentioned, taking an unknown junk box core, where you don't know these parameters, makes the design a trial and error approach.

 

[crutschow]

Magnetics still a bit alien to me.

One thing that can be confusing is magnetizing current and core saturation.

Applying a voltage to a transformer causes the (magnetizing) current to increase as determined by the applied voltage, its duration, and the transformer primary inductance [I = (V*t)/L].
This is unrelated to, and independent of, current due to any output load.
As long as the voltage is applied, this magnetizing current will continue to rise until the core saturates, at which point the inductance approaches zero and the current now becomes limited only by the winding resistance.
So you need to get the combination of the applied voltage and it's duration (volts-seconds) below that limit as given by the transformer inductance and saturation values.
These are determined by the number of primary turns, the core size and magnetic path length, and the core material magnetic characteristics.

And you also need to get that inductive magnetizing current back to zero before the next pulse, since that current doesn't instantly stop when the voltage stops.
Otherwise the current will continue to increase from each pulse, until the core saturates.
That's what the second winding and diode do.
They conduct the magnetizing current when the pulse voltage returns to zero, to dissipate that inductive energy (E = 1/2 LI²) in the winding resistance and diode voltage drop.

Make sense?

 

[throbscottle]

I've read enough bits and pieces over the last couple of years, that yes, it does make sense. Always good to have the same picture presented many different ways when learning as different things come into the foreground.

I'm happy with the trial-and-error approach if I can understand what's going on. Probably be a week now before I can tinker again.

The drive transistor has been getting hotter than I thought it should so that could be another sign of saturation.

I know that the decaying magnetic field in the core is what keeps the current flowing, but why is the catch diode in the original circuit not enough to bring the magnetising current down to zero, and an extra winding needed? I've come across this before and still mystified.

 

[dr pepper]

Well put Cruts.

The diode only conducts in one direction, as Cruts was saying magnetization current needs a path in both directions, the diode only provides one, in a forward converter as we've been discussing the 'reset' winding with the diode across it is the other path.
The core is a bit like a bucket of water, charging it one way fills the bucket, and discharging it in the opposite way empty's it, and the rule is the bucket must be empty after each cycle.

 

[throbscottle]

Found a good explanation of the B-H curve so I'm slightly more educated!
Right, I tried putting another winding on it. Because I don't know how many turns I ended up with on the primary so it's only approximately the same. Checked the phase, connected a 1N5819 across the winding, powered on. Only 0.6v output! Waaaaaaa! Tried the diode the other way in case I'd got mixed up. Pretty similar result.
Is it worth re-winding to get the primary and tertiary windings the same, or should I just try a different core?

 

[throbscottle]

Ahh, I see now, the reset winding doesn't float with a diode across it, it has to be in series with the diode and across the supply.

 

[dr pepper]

The reset winding is only connected to the diode for a supply with a single switch.
You can trace the BH curve is you have a sig gen and a 'scope.
If the ferrite is meant for suppression it might be unsuitable.

 

[throbscottle]

When I finally build my sig jenny I'll have a go at that (the display I'm setting up is for the afforementioned jenny which BobW has been helping me with).
I think all the evidence points to one thing, the core simply isn't big enough. It came from a bi-filar power inductor from an SMP.
I found out this week you can open old ferrite transformers by boiling them, and I've got dozens, so I'll have a go with one of those.

 

[dr pepper]

If it was the o/p choke then it will be powdered iron, might be no use.
Powedered iron cores are often painted on one side, a colour code.
You can find out the maximum flux density of the core by building a saturation jig, I have one its very useful.
https://www.dos4ever.com/flyback/flyback.html (towards the bottom of the page).
The size of the core is important, but its max flux density and Al factor are more important.
I've boiled a few trannies & cores too, works well dunnit.

 

[throbscottle]

Interesting page, I wondered how you test these things. The test bench looks similar to an inductance checker I built ages ago (gives dc output proportional to L).
I boiled a tranny from an old wall-wart and wound it 50:30. Giving pretty close to 3V, which now I've got it I think may actually be a little high for the filament. No tertiary winding, just a diode as in Kerry's schematic. No noticeable heating of the transistor this time. Much less sensitive to frequency, and varying the m/s ratio gives some voltage adjustment.
I might have another go with a small toriodal core just for kicks...
(edit) It wasn't a choke, it was bi-filar wound with about 8 turns. Quite thick wire. Or do they make chokes like that too?

 

[dr pepper]

Chokes come in various guises.
The checker applies power to the choke via a switch, and the on time is adjustable, energy from the choke at switch off is dumped into a diode.
You vary the on time and observe the current rise across the sense resistor with a 'scope, when the core saturates theres a sudden change in the current slope.
Very handy gizmo if your into switchers.
The last time I used a Vfd the heater was just supplied via a couple of diodes from 5v, it worked, but Ac is much better for them. I did come across Kerry's cicruit when I was looking for something to power a Crt tube heater.
Using a Smps tranny is a better idea, still a bit hit & miss without doing any calcs, but good to see you got it to work.

 

[rjenkinsgb]

bi-filar wound with about 8 turns

That sounds like a common mode suppression choke.
Low frequency ones as used for mains input filtering in equipment are often made with deliberately "lossy" materials.

https://www.we-online.com/web/en/pa...m/blog_detail_electronics_in_action_61439.php
https://docs-emea.rs-online.com/webdocs/1526/0900766b8152611d.pdf

A similar winding could be used on an RF transformer or balun, but the core properties would be very different to a suppression choke. It's difficult to tell from an unmarked part in isolation.

 

[throbscottle]

It would explain why it's a bad performer. I've a few common mode chokes of the "side by side" variety, never realised there was a bi-filar version. I have a few of them too. (Been very busy dismantling!)
The thing is so low power (like driving a torch bulb really) I thought I'd be able to get away with pretty much anything. My first attempt was with a core taken out of a CFL. Fat lot of good that did! Well there was my wake up call...

So, I take it there is no point unwinding chokes to use the cores, then?

The VFD is going to be driven by one of those nice Princeton chips that takes a negative supply so there's no need to do level shifting, just ensure the filament has the negative HV supply, offset a bit to prevent ghosting.

 

[dr pepper]

Its doing stuff like this that gets you to know what to save from scrap boards, and how to use it.

For a power transformer even a little'un you need power ferrite.
The thing with Vfd fluo's is that the heater needs a reasonable current, mybe not a lot of power but a 100ma or so at low voltage.