TV flyback number of primary turns rule of thumb

[John123]

Hi, is there a rule of thumb for selecting the minimum number of primary turns when driving these things using external primary windings on the exposed part of the core? For example in flyback topology to produce arcs what would be a good volt/turn rule of thumb to use for extended run times to reduce losses and saturation etc.

Also does switching frequency and max on time/off time ratio count with selecting turns? Right now I'm using 40khz and a max on time of 60% (audio modulated PWM limit) with 24v and 8 turns, arcs get bigger and power consumption increases if I reduce turns but obviously I assume there would be problems going down to say 3 turns. The driver currently doesn't have any peak current control so I'm being careful about keeping the arc continuous and steady rather than pulling them until they extinguish or letting the flyback run unloaded.

[davekni]

Sounds like you are running a non-resonant flyback (no capacitor across primary).  Will be more efficient if using existing primary winding under secondary since leakage inductance will be lower.  However existing primary likely has more turns, designed for ~200Vdc supply and ~1500V switch transistor, with correspondingly lower current.  For ZVS-driven flybacks, some leakage inductance is helpful.  That's an ideal case for winding a new primary on other ferrite leg.

No simple rule-of-thumb I'm aware of.  For a lower limit on turn count, calculate maximum volt-seconds/turn to stay under maximum flux density.  0.3T is probably OK for most ferrite materials.  Some can go to 0.4T (or even 0.5T if kept close to room temperature).  Core cross-sectional area in meters^2 * flux (0.3T) gives volt-seconds/turn.  Max_on_time * 24V / turn_count needs to be under that limit.  This calculation avoids saturation, but doesn't guarantee reasonable ferrite heating due to AC losses.

[John123]

Sounds like you are running a non-resonant flyback (no capacitor across primary).  Will be more efficient if using existing primary winding under secondary since leakage inductance will be lower.  However existing primary likely has more turns, designed for ~200Vdc supply and ~1500V switch transistor, with correspondingly lower current.  For ZVS-driven flybacks, some leakage inductance is helpful.  That's an ideal case for winding a new primary on other ferrite leg.

No simple rule-of-thumb I'm aware of.  For a lower limit on turn count, calculate maximum volt-seconds/turn to stay under maximum flux density.  0.3T is probably OK for most ferrite materials.  Some can go to 0.4T (or even 0.5T if kept close to room temperature).  Core cross-sectional area in meters^2 * flux (0.3T) gives volt-seconds/turn.  Max_on_time * 24V / turn_count needs to be under that limit.  This calculation avoids saturation, but doesn't guarantee reasonable ferrite heating due to AC losses.

Thanks Dave. Core measurements are as follows plus the small 0.3mm core gap if that counts, the legs to the left which run down through the center of the flyback are rounded rather than square however.

I'm having trouble calculating the volt-seconds/turn from these numbers. The core and flyback do warm up, after 15 minutes of audio modulation it was like a little hand warmer.

It is indeed running in single ended flyback mode, output stage is the same as this except with a more modern fet and without the current sensing part and a smaller 33nF drain capacitor since the frequency is higher at 40khz (peak drain voltage is well under the fets maximum rating).

[klugesmith]

>>I'm having trouble calculating the volt-seconds/turn from these numbers.

To save you and Dave the trouble: I got 38.4 volt-microseconds using Dave's formula in a spreadsheet. The cells in green have values in SI units.   

When converting between, say, square mm and square meters, I always dedicate a spreadsheet cell and formula. Instead of polluting a physics formula with a power-of-1000 multiplier, to do 2 unrelated steps in 1 cell. Cells are cheap.

Can you draw primary and secondary voltage and current waveforms (ideal), and know whether you get to double the flux limit (i.e. to swing between -Bmax and +Bmax) ?
Flyback topology is one case with nonzero DC current in primary and secondary windings, but the time-averaged voltages are still zero if we neglect winding resistance.

[John123]

>>I'm having trouble calculating the volt-seconds/turn from these numbers.

To save you and Dave the trouble: I got 38.4 volt-microseconds using Dave's formula in a spreadsheet. The cells in green have values in SI units.   

When converting between, say, square mm and square meters, I always dedicate a spreadsheet cell and formula. Instead of polluting a physics formula with a power-of-1000 multiplier, to do 2 unrelated steps in 1 cell. Cells are cheap.

Can you draw primary and secondary voltage and current waveforms (ideal), and know whether you get to double the flux limit (i.e. to swing between -Bmax and +Bmax) ?
Flyback topology is one case with nonzero DC current in primary and secondary windings, but the time-averaged voltages are still zero if we neglect winding resistance.

Thanks, how would I convert that number into minimum turns? I'm really bad at maths and these numbers tend to confuse me, every time I try using the calculator I get different results (mostly garbage) as I'm not sure if I'm doing it right.
Drain voltage waveform from a simulation Dave uploaded in 2020.

Sense resistor current.

[davekni]

Thanks, how would I convert that number into minimum turns?

60%/40kHz = 15us maximum on-time.  15us * 24V = 360uVs.  360uVs / 38.4uVs/turn = 9.4 turns.  Or, in other words, your 8 turn coil is generating peak flux a bit higher than 0.3T.  Specifically:  360uVs / 8turns = 45uVs/turn.  45uVs/turn / 128mm^2 = 0.35T.  Probably fine for your core.  If the core is designed for flyback use, it likely has high saturation flux, so perhaps 0.5T at room temperature dropping to 0.4T by 100C.

Drain voltage waveform from a simulation Dave uploaded in 2020.

That simulation has some different part values.  Most important, it has no secondary winding with its coupling factor and parasitic capacitance and arc load.

[John123]

Thanks, how would I convert that number into minimum turns?

60%/40kHz = 15us maximum on-time.  15us * 24V = 360uVs.  360uVs / 38.4uVs/turn = 9.4 turns.  Or, in other words, your 8 turn coil is generating peak flux a bit higher than 0.3T.  Specifically:  360uVs / 8turns = 45uVs/turn.  45uVs/turn / 128mm^2 = 0.35T.  Probably fine for your core.  If the core is designed for flyback use, it likely has high saturation flux, so perhaps 0.5T at room temperature dropping to 0.4T by 100C.

Drain voltage waveform from a simulation Dave uploaded in 2020.

That simulation has some different part values.  Most important, it has no secondary winding with its coupling factor and parasitic capacitance and arc load.

Thanks! That's much easier to understand after seeing a working out example. BTW the 60% is just the max the PWM can swing to, it's default "center" is set to 30% and the audio signal PWMs the duty cycle between close to zero and 60% to vary the arc power and make sound.

Doesn't sound too bad and can get louder than my phone without distorting/clipping with the right kind of music (80's stuff seems to fit that criteria best). Actually I recently found out the background hissing sound comes from the bluetooth to aux receiver I'm using, the arc is silent on its own or if I use an old phone with a direct 3.5mm to 3.5mm aux cable.

I wonder if some kind of feedback from the output could help cancel out the non linearities of the TV flyback and make the sound even better.

[John123]

Turns out I've been running it in forward mode by having the primary coil connections reversed, are the calculations still the same for forward vs flyback?

Output voltage in forward mode is lower but the arc is thicker and current draw is about 500mA higher from the 24v source with all other parameters being the same.

[davekni]

Turns out I've been running it in forward mode by having the primary coil connections reversed, are the calculations still the same for forward vs flyback

Yes.  That calculation presumes no flux remaining in core at start of each FET on period.  If flux has reversed due to the resonant cap, then peak flux will be a bit lower.  The calculation gives peak-to-peak, which is the same as peak for one-sided waveforms such as typical of non-resonant flyback topologies.