24V to 500-1000V, 20W floating DC-DC converter

bitrex wrote:

> Yes, the original flyback in the EDN article has as 0.01uF 1500V cap as 
> the main output cap and 0.022u 500Vs as the CW boost caps. this is a 4uF 
> 1000V cap, damn:
> 
> <https://i.ebayimg.com/images/g/LYEAAOSwiONYNMh1/s-l500.jpg>

And here is 50uF/900V, probably smaller:

https://pl.mouser.com/ProductDetail/Vishay-Roederstein/MKP1848C65090JY5?qs=zDkSFN9STR8qFDSFzHtCiw%3D%3D

     Best regards, Piotr

On 7/24/2020 12:50 PM, Tim Williams wrote:
> I made this some time ago,
> https://www.seventransistorlabs.com/Images/DCDC_800V.jpg
> UC3843 based flyback, very much a stock circuit, with an RCD snubber 
> added to the transistor I think.� 12V input, but 24V is fine too, with a 
> UC3842 and a couple component values changed.
> 
> Interesting part is the transformer, which has this windup,
> https://www.seventransistorlabs.com/Images/DCDC_800V_FoilWindup.jpg
> 3 layers of copper foil tape primary, with the secondary interleaved 
> between layers 1-2 and 2-3.� Each secondary is 15 turns 28AWG.� They are 
> wired in series, with the CT used as a ground point, and the rectifiers 
> being complementary.� Thus, effectively the output is (up to) +/-400V.
> 
> Core is EE33, slightly gapped.� (Hmm, weird, I thought those cores were 
> closer to 1uH/t^2 ungapped.� Did that particular one just happen to 
> overperform?..)� Way overkill for this power level (under 50W) but I 
> have a bunch of them on hand. :)
> 
> This module grounds the negative, so a positive output is obtained and a 
> feedback divider can be used to regulate.� (The divider is adjustable 
> for a 100-800V range.)

> For isolated application, the feedback divider has to be secondary side 
> only and a TL431 used for error amp, into an opto, in the usual way.� It 
> should probably be powered by an aux winding, say 10V (about one turn?) 
> so you don't have to draw several mA from the full HV to run it.

Could always do something kind and use an i2c isolator across the 
barrier and digipot on the secondary divider if anyone is expected to 
adjust the HV while the thing is running.

Maxim makes cheap ones if one is ok with using Maxim

> The balanced secondary design is essentially mandatory for an isolated 
> converter.� This cancels out most of the EMI; rather than 400V of delta 
> V across the isolation barrier, there's only ~40V due to the primary's 
> still unbalanced voltage.� This could be improved further by adding 
> shields, at some expense to leakage inductance.
> 
> I don't know where you would find such a transformer off-the-shelf.  
> They're not hard to wind if you just need a few, or you could ask 
> someone like Xfmrs to make them.� The interleaving, and balanced design 
> where possible, are critical to performance, and having sane EMI.

Do you remember the late Vladimir Vassilevsky's "antiseptic conveter"?

It was sort of like a SEPIC/flyback that used an off the shelf 
transformer with large leakage inductance and recaptured the leakage 
energy into the secondary.

But as I recall there was a cap across primary to secondary so not 
strictly isolated for AC.

> And by "sane" I mean, if the secondary is unbalanced, you'll literally 
> be running a low amplitude EFT generator or some bullshit like that.  
> Futile to filter.� And you probably don't want too much filtering 
> impedance so as to keep it reasonably well isolated at AC too..?
> 
> Tim
> 

On Fri, 24 Jul 2020 20:29:43 +0200, Piotr Wyderski
<peter.pan@neverland.mil> wrote:

>bitrex wrote:
>
>> Yes, the original flyback in the EDN article has as 0.01uF 1500V cap as 
>> the main output cap and 0.022u 500Vs as the CW boost caps. this is a 4uF 
>> 1000V cap, damn:
>> 
>> <https://i.ebayimg.com/images/g/LYEAAOSwiONYNMh1/s-l500.jpg>
>
>And here is 50uF/900V, probably smaller:
>
>https://pl.mouser.com/ProductDetail/Vishay-Roederstein/MKP1848C65090JY5?qs=zDkSFN9STR8qFDSFzHtCiw%3D%3D
>
>     Best regards, Piotr

20 joules!

LLNL considers something like 9J to be potentially lethal.

On 7/24/2020 1:51 PM, John Larkin wrote:
> On Fri, 24 Jul 2020 20:29:43 +0200, Piotr Wyderski
> <peter.pan@neverland.mil> wrote:
> 
>> bitrex wrote:
>>
>>> Yes, the original flyback in the EDN article has as 0.01uF 1500V cap as
>>> the main output cap and 0.022u 500Vs as the CW boost caps. this is a 4uF
>>> 1000V cap, damn:
>>>
>>> <https://i.ebayimg.com/images/g/LYEAAOSwiONYNMh1/s-l500.jpg>
>>
>> And here is 50uF/900V, probably smaller:
>>
>> https://pl.mouser.com/ProductDetail/Vishay-Roederstein/MKP1848C65090JY5?qs=zDkSFN9STR8qFDSFzHtCiw%3D%3D
>>
>>      Best regards, Piotr
> 
> 20 joules!
> 
> LLNL considers something like 9J to be potentially lethal.

A lethal dose of energy may be much less than 9J depending on exactly 
when during the heart rhythm the energy is received.

On Friday, July 24, 2020 at 4:38:55 AM UTC+10, John Larkin wrote:
> On Thu, 23 Jul 2020 10:14:28 -0700 (PDT), Matt B
> <matt.blessinger@gmail.com> wrote:
> 
> >On Thursday, July 23, 2020 at 11:31:26 AM UTC-5, jla...@highlandsniptechnology.com wrote:
> >> Any decent analog designer could Spice this in an houror two; I could 
> >> do it in 20 minutes, but I'm familiar with the part. All of the LT3083 
> >> boosts and flybacks that I've simulated have worked as expected. My 
> >> only problem has been inductor heating, which won't be an issue here 
> >> with the recommended transformer. 
> >> 
> >> If he wants to scribble or Spice a circuit, I offered to help. I'm 
> >> sure MPS would help too. 
> >
> >I'll be taking you up on that offer. Hopefully I can get around to the design this weekend.
> 
> 
> Cool. Group designs are fun. We rarely do that here.
> 
> You'll need a model of the transformer, magnetizing and leakage
> inductance and maybe some capacitances. Those are easy to approximate.

The idea that you ought to use flyback in a scheme for getting 20W at up to 1000V strikes me as wrong.

The problem with flyback is that you have to dump your quantum of energy into the inductance at low voltage, then wait for the inductance to ring up to the desired output voltage - using up some of the energy in charging the coil's interwinding capacitance - after which you have to wait a bit longer for what's left of that energy to get dump into the high-voltage reservoir capacitor.

Whenever I've tried to do it, I've ended up with bulky inductors which I've had to gap before they can store enough energy.

Forward converters - which do include centre-tapped inverter transformer systems  (as in the Royer and Baxandall inverters) - don't have to store anything like as much energy.

The catch is that you need a fairly high turns ratio to get the voltage step-up (but pi/2 less with the Baxandall inverter) which means a custom wound transformer, and the high turns ratio always means a relatively high inductance secondary and a relatively low self-resonant frequency (even if you can break up the secondary into stacked banks, which can dramatically reduce the inter-winding capacitance).

On the other hand you end up with a tolerably compact transformer and something like 95% efficiency, which Jim Williams wrote a lot of application notes to gloat about (Linear Technology application notes AN45, AN49, AN51, AN55, AN61, and AN65).

And you don't have to wrestle with the complexities of an integrated circuit switching regulator chip - none of them ever do exactly what you want them to.

Making the output voltage variable takes an effort - I like the idea of pulse-width modulating the drive into the inductor in a Baxandall class-D oscillator, but I've never made it work with a real circuit, though it simulates fine.

-- 
Bill Sloman, Sydney

On 7/24/2020 10:35 PM, Bill Sloman wrote:
> On Friday, July 24, 2020 at 4:38:55 AM UTC+10, John Larkin wrote:
>> On Thu, 23 Jul 2020 10:14:28 -0700 (PDT), Matt B
>> <matt.blessinger@gmail.com> wrote:
>>
>>> On Thursday, July 23, 2020 at 11:31:26 AM UTC-5, jla...@highlandsniptechnology.com wrote:
>>>> Any decent analog designer could Spice this in an houror two; I could
>>>> do it in 20 minutes, but I'm familiar with the part. All of the LT3083
>>>> boosts and flybacks that I've simulated have worked as expected. My
>>>> only problem has been inductor heating, which won't be an issue here
>>>> with the recommended transformer.
>>>>
>>>> If he wants to scribble or Spice a circuit, I offered to help. I'm
>>>> sure MPS would help too.
>>>
>>> I'll be taking you up on that offer. Hopefully I can get around to the design this weekend.
>>
>>
>> Cool. Group designs are fun. We rarely do that here.
>>
>> You'll need a model of the transformer, magnetizing and leakage
>> inductance and maybe some capacitances. Those are easy to approximate.
> 
> The idea that you ought to use flyback in a scheme for getting 20W at up to 1000V strikes me as wrong.
> 
> The problem with flyback is that you have to dump your quantum of energy into the inductance at low voltage, then wait for the inductance to ring up to the desired output voltage - using up some of the energy in charging the coil's interwinding capacitance - after which you have to wait a bit longer for what's left of that energy to get dump into the high-voltage reservoir capacitor.
> 
> Whenever I've tried to do it, I've ended up with bulky inductors which I've had to gap before they can store enough energy.
> 
> Forward converters - which do include centre-tapped inverter transformer systems  (as in the Royer and Baxandall inverters) - don't have to store anything like as much energy.
> 
> The catch is that you need a fairly high turns ratio to get the voltage step-up (but pi/2 less with the Baxandall inverter) which means a custom wound transformer, and the high turns ratio always means a relatively high inductance secondary and a relatively low self-resonant frequency (even if you can break up the secondary into stacked banks, which can dramatically reduce the inter-winding capacitance).
> 
> On the other hand you end up with a tolerably compact transformer and something like 95% efficiency, which Jim Williams wrote a lot of application notes to gloat about (Linear Technology application notes AN45, AN49, AN51, AN55, AN61, and AN65).
> 
> And you don't have to wrestle with the complexities of an integrated circuit switching regulator chip - none of them ever do exactly what you want them to.
> 
> Making the output voltage variable takes an effort - I like the idea of pulse-width modulating the drive into the inductor in a Baxandall class-D oscillator, but I've never made it work with a real circuit, though it simulates fine.
> 

tangentially related but this is an idea I had (probably not novel) for 
a medium-power high voltage series battery balancing charger based on an 
isolated/bidirectional Cuk converter:

<https://www.dropbox.com/s/vh8mo7nw7ysczg2/multi_cuk.jpg?dl=0>

The bulk charge supply charges the series stack, when a sense amp 
detects detects one of the batteries in the stack is above the others 
the appropriate Cuk converter secondary on the transformer is cut in and 
pushes energy from that one back into the bulk charge supply cap through 
the primary.

Tim

For your balanced flyback, your output windings are  effectively in parallel, right?

So each winding share the flyback energy, so the positive and negative rails has quite similar voltage?

"Klaus Kragelund" <klauskvik@hotmail.com> wrote in message 
news:58b3b626-f42b-4f3e-8509-be176da224deo@googlegroups.com...
> Tim
>
> For your balanced flyback, your output windings are  effectively in 
> parallel, right?
>
> So each winding share the flyback energy, so the positive and negative 
> rails has quite similar voltage?

Well, they're wired in series, but they act in, well, a balanced manner, so 
as far as the transformer is concerned, they act in parallel.  If that's 
what you mean then yes. :)

The foil primary helps enforce flux balance, so yes the two halves will have 
good cross-regulation, if you were to wire it as a bipolar source instead.

Tim

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

I can't figure out how to post new message with the new google group.  A search of TL431 came up with this thread recently.  So, i am high-jacking this for a quick question:

> For isolated application, the feedback divider has to be secondary side only 
> and a TL431 used for error amp, into an opto, in the usual way.

I need a 12V 5A High Current Shunt Regulator, using fig. 19 of page 7

https://www.onsemi.com/pub/Collateral/TL431-D.PDF

I am using 10K for R1, 1.1K for R2.  What should i use for Rbe of the pass transistor (TO-220 PNP)?

On 09/09/20 13:21, Edward Lee wrote:
> I can't figure out how to post new message with the new google group.  A search of TL431 came up with this thread recently.  So, i am high-jacking this for a quick question:
>
>> For isolated application, the feedback divider has to be secondary side only
>> and a TL431 used for error amp, into an opto, in the usual way.
>
> I need a 12V 5A High Current Shunt Regulator, using fig. 19 of page 7
>
> https://www.onsemi.com/pub/Collateral/TL431-D.PDF
>
> I am using 10K for R1, 1.1K for R2.  What should i use for Rbe of the pass transistor (TO-220 PNP)?

Only problem with that is that the base drive current of the power
transistor may be greater than the 431 can handle. May need to
use a darlington on the output, which would also improve the
gain around the 431. Or, perhass replace the transistor with a
power mosfet.

Depending on required precision, you could do that with a single zener,
resistor and power transistor. Drop the 431 altogether...

Chris