SMPS for vacuum tube power amplifiers.(status: back at it)

[SK_Caterpilar_SK]

It definitely is a EMC hell, so far aiming for anything functionall.

I do have a solution for opto isolating the output with clasic method (TL431 and PC817). I can send you the schematic for that. It loads the supply at 500V with about 2W. It gives a little load for stability. I simply shunt regulated the opto diodes voltage so I dont kill the TL431. The primary side of the thing is simply text book grade stuff, sufficient enough.

Switching transformers provided by coilcraft are decent enough. I use the GA3459 for 12-24V to 500V conversion. Also the GA3460 does the same but the winding DCR is much lower and other things..

I planned using the HA4060 to try with line voltage but currently I just have the transformer and dont plan on doing such a dangerous project.
So far Im searching a suitable controller for the job. Not having bias windings in transformers is a pain, but solvable. If it would work out with temporary solutions I would go ahead and quote coilcraft for custom transformers but so far there is no point in doing that. Unless I find a viable solution for the first problem.

[mrjoda]

A feedback of smps is never so easy as it looks. You have to think in frequency domain, not in time.

check this article :

https://www.onsemi.com/pub/Collateral/TND381-D.PDF

it gives you a long no sleep nights and headache.

TL431 is the most straightforward solution, but not the best.

I based my ideas on Viper 53 from ST. I was not able to find any reasons why is not suitable for this task,  but maybe someone with better experiences could find one. 

[SK_Caterpilar_SK]

The 431 is the simplest all around solution and I like the Viper 53 unfortunately for me still trying to make a 12-24V input power supply and the switching transformers I have don't have aux windings. And in general I find PSR way more confusing than voltage feedback from the secondary side.
I would like to consider that option tho, I just don't have the literature to clarify some things

[scatterandfocus]

I saw this thread some days ago, but I don't login to any sites from work machines and network, so I didn't reply.

It's an interesting project.  At least one manufacturer did produce some tube amps using switching supplies - Crate V series amps.  They didn't go over well in the long run because of audio quality (noise) and generally poor construction.  Keep in mind that these amps were not aimed at the vintage amp crowd.  They were low priced tube amps, where the crowd isn't so picky as long as things work as they should and sound ok.  I haven't seen any other tube amps that use a switching supply, but I haven't gone looking either.  If done well enough that I couldn't hear a difference between it and a linear supply AND the thing is well constructed AND it doesn't introduce noise into other devices, I would prefer the more efficient supply.

Oh,and I forgot about some of the scary failures of those Crate V series amps.  Just went on a little reading spree about those amps, as I haven't heard anything about them in a long time now.  Apparently Crate went out of business a couple of years after the release of the V series amps.

[T3sl4co1l]

Dave Berning created a series of tube amps with DC transformers -- MOSFET choppers plus transformer coupling.  IIRC, they ran around 200kHz, and the primary side used as many windings in series to supply the voltage required for the tube, allowing otherwise fairly difficult-to-use types like transmitter tubes (e.g. 1kV 100mA, a high load impedance) to be used down to DC.

Tim

[SK_Caterpilar_SK]

Hey Tim, or anyone really who knows the answer to this.

How do I figure out what kind of capacitance I need on the input side of my powersupply (mains side).

Assuming ill be operating in DCM at 100kHz with peak current of 2.5A
Min voltage 120V DC up to 377VDC. What capacitance do I need after the mains rectifier?

[MagicSmoker]

Hey Tim, or anyone really who knows the answer to this.

How do I figure out what kind of capacitance I need on the input side of my powersupply (mains side).

Assuming ill be operating in DCM at 100kHz with peak current of 2.5A
Min voltage 120V DC up to 377VDC. What capacitance do I need after the mains rectifier?

At it's simplest you can rearrange/modify the capacitor energy equation to solve this problem:

C = [(2 * Pin) / (Vpeak² - Vtrough²)] * tdischarge

Where tdischarge is the period of half a line cycle (e.g. - 10ms in most of the world, 8.33ms in North America), Vpeak is the max voltage the capacitor charges to each half cycle and vtrough is the minimum voltage the capacitor discharges to each half cycle. A typical value for vtrough for full-wave rectified single-phase mains is 75% of Vpeak. Note that while the capacitor size is reduced as Vtrough is reduced, the peak current in the switch goes up. No free lunches and all that.

For example, if you have 220VAC nominal mains and 500W input power then you only need 240uF to keep peak to peak ripple under 25%. That said, keep in mind that electrolytic capacitors have some of the loosest tolerances and lowest lifespans of any electronic component and you might find you need to use a lot more capacitance just to handle the ripple current and/or to achieve a reasonable life.


EDIT - I haven't read all of this thread.

[SK_Caterpilar_SK]

So

2x 56uF 450V and one 10uF low ESR all in parallel (perhaps a SMD high voltage cap to bypass it a little better. I dont know if that makes it any better but I expect it will )

Max output power I expect is 70W peak 50W nominal.

[T3sl4co1l]

One 100-220uF snap-in type will suffice.  They have surprisingly excellent performance.  Even a smaller value, give or take ratings (it may be conventional can style; always check the datasheet, in any case!).

These flyback supplies use the same range, for 120V input (the one on the left has a 400V cap so should be able to run from a 240V input as well; or, I forget if the transistors were 600V+ rated, if they're only 250 or 400V, that's a bit of a problem with that).
https://www.seventransistorlabs.com/Images/100WPowerSupplies.jpg
So at 240V only, even half capacitance (47-100uF) would be acceptable.

Tim

[SK_Caterpilar_SK]

I'm using snap ins for the output filtration so I indeed plan on using snaps at the input side to keep the low profile look with them. Also I wanted two 56uF to give me total 112uF and another 10uF low impedance just to give it some extra security of it not failing in short order. The load would split evenly across the caps hopefully not stressing them too much.

[MagicSmoker]

2x 56uF 450V and one 10uF low ESR all in parallel (perhaps a SMD high voltage cap to bypass it a little better. I dont know if that makes it any better but I expect it will )

Max output power I expect is 70W peak 50W nominal.

70W output max, so estimate about 100W input and I assume this is a flyback, but is it just for operation at your mains voltage (220VAC/50Hz) or "universal" (typically 90VAC to 260VAC or thereabouts), and so the input capacitance needs to be designed for the low-end of the voltage range?

High frequency bypass of the input capacitance to a buck or flyback is always a good idea, but something like a 100nF film capacitor placed right across the power loop (e.g. - top of flyback primary and input ground) would be more appropriate; there's no help - and sometimes harm in the form of ringing - from placing the low-ESR bypass capacitor right next to the electrolytics doing the low-frequency ripple filtering and reservoir duty.

[T3sl4co1l]

It really won't matter because as long as the electrolytic is placed where the film bypass would be placed, it'll perform just as well.  Bypass isn't critical on a flyback anyway, leakage inductance acts in series with the supply.  The switch snubber is the only required low-inductance path.

Tim

[MagicSmoker]

It really won't matter because as long as the electrolytic is placed where the film bypass would be placed, it'll perform just as well. ...

Bolded the key part; my point wasn't dissimilar, just more that 10uF of "low-ESR" (unspecified, but assumed film) capacitance directly in parallel with the bulk elkos located some distance away from the switching loop would be kind of a waste of money here.

[m3vuv]

stick an hf transceiver near it if you think its noise free!!

[SK_Caterpilar_SK]

Low ESR electrolitic (low impedance caps)

[SK_Caterpilar_SK]

Alright this is how far i got. I dont have aux windings so I had no idea how to power the chip, so I decided to go with a pass tipe regulator, after all the worst it can get is 2W loss (5mA limited current draw).
there will be a mains filter before the rectified, just externally. I have not chosen values for everything yet. There are two mosfets on the schematic (so I can either use TO247 or TO220).

I dont have a snubber network yet. I dont know how to make one at all. And if its at all necessary. (Fsw 100kHz, VinDC 100V-400V and Ipk=2.55A)

I have utilised all my dirty tricks there in the feedback to not burn the 431 in the instant the PSU is turned on.

[SK_Caterpilar_SK]

Also I have seen in many SMPS designs capacitors going from primary ground to the secondary ground (small capacitance high voltage rating cermaic caps). Rumor says its for noise reduction but I dont really know if it will be necessary. I can add them onto the PCB and during testing I can just leave try them and if it wont work nicely then not populate the places. I just dont know what capacitances to look for, and if its even necessary.

[T3sl4co1l]

Snubber and Y-cap (GND-GND) depend on transformer characteristics.

Ideally, a shielded transmission line transformer could be made, giving very little common mode noise, and a modest amount of leakage which may be tolerable without a snubber.  This is not easily made, however: personally, I'd sit down and take about an hour to design such a thing, checking against available cores and bobbins to see how much winding area and wire length is possible, and therefore what impedance, and how much leakage inductance, I can get.

But if you're just putting windings on a core, count on needing both, and common mode chokes.

Typical snubber is just the drain pin into a diode into a capacitor to ground.  The cap is relatively large (>= 10n?), so that the peak voltage gets clamped by it.  This "boost" voltage is discharged into the DC+ supply with a power resistor.

The capacitor can be DC+ referenced, if the DC+ bypass is nearby.  This is often seen in commercial designs, since it's easier to lay out (the R and C are simply in parallel).

Note that you want a TLVH431 or the like -- lower minimum current.

Tim

[mrjoda]

For bobbins and cores, check the semic trade (semic.cz) it is the best distributor for these things as i know in our countries.

[SK_Caterpilar_SK]

Thank you Tim for your reply.

No the transformer is bought from coilcraft. Part number is HA4060. So far I did not make my own. Its been pre made for a different but similar purpose.

The specs of the transformer are here. I will be slightly overdriving the transformer (2.5A Ipk instead of 2A)
https://www.coilcraft.com/pdfs/ha4060.pdf

Im not looking forward to wind my own transformers. I rather quote Coilcraft for something custom or the local manufacturer..

Do I need a snubber network at all and if yes how do I design one ?

[T3sl4co1l]

That transformer looks alright, 2.7uH is pretty typical and should be easy to snub.

Look up SMPS circuits, you'll see how the snubber is wired.  Typical values would be something like, UF4007, 10nF and 22kΩ.

Tim

[SK_Caterpilar_SK]

Also what are my options using a zenner clamp snubber? What re the drawbacks? Can I combine zenner clamping with the "traditional" way of snubbing? Is there even a point of doing that?

[T3sl4co1l]

RCD clamp snubber and zener clamp snubber are about as good.  They are both better than RC damper, because the peak voltage is limited more rigidly.

RCD rate (dV/dt) snubber is acceptable, and can provide reduced turn-off losses (in the transistor at least; with optimal values, the overall losses can actually be reduced slightly), and can be dimensioned to dampen the ringdown waveform (in DCM), but doesn't provide as good clamping.

The only differences between RCD rate and peak snubbers is the RC time constant, and where the resistor is connected: in the rate snubber, the resistor is in parallel with the diode, and the RC time constant is small (a fraction of a cycle).

Tim

[SK_Caterpilar_SK]

So what would be the possibly best all arround solution. With the values you suggest and if possible a simplified schematic too.. Because at this point I have seen soo many variations that I have no idea which one is the correct one to use in my speciffic case.

[MagicSmoker]

So what would be the possibly best all arround solution. With the values you suggest and if possible a simplified schematic too.. Because at this point I have seen soo many variations that I have no idea which one is the correct one to use in my speciffic case.

What does your current design look like? Sampling a few random posts including the first one it seems that somewhere along the way this went from being a 12V input switcher of indeterminate type to an offline (aka mains-supplied) flyback.

BTW - switchers are difficult, but you are doing much better at 17 than I was. Then again, you had to request datasheets by fax when I was 17, and the internet of the time were bulletin boards reached by 300bps modems... Good times...