Generating 300VDC

[Digibin]

I have a requirement to generate approximately 300VDC (low currents, no more than a few mA probably) from a 5V USB supply. It's a prototype for a portable system so total footprint, efficiency and weight are important. The prototype will be an Arduino shield hence USB supply.

I've looked at two approaches; 1) Cockcroft Walton voltage multiplier and 2) flyback boost converter.

1)
I looked at boosting 5V up to 24V using a standard DC-DC converter and then driving a voltage multiplier up to 300V. I simulated this in LTSpice and due to diminishing returns of each successive stage, it looks like it would need over 20 stages, which would need a large board area. And this is using ideal components. Increasing the frequency of the AC input appears to increase the output voltage, how else can I improve the efficiency of this topology? Increasing the initial voltage would of course help, which could perhaps be done by chopping the DC into a step up transformer prior to the voltage multiplier. Not sure if this would save board space, or how big/heavy it would need to be, but I'm less concerned with weight than I am with board size and efficiency. I'm happy to share the simulation.

2)
I then looked at camera flash circuits which are essentially flyback converters. In terms of component count this approach looks much better. Using a (small) 1:100 transformer the 5V input should easily reach 300V. But how does the efficiency of this compare with the voltage multiplier?

For me the flyback is more favorable at the moment. Which leads me to ask the question, is the voltage multiplier used much in practice? I guess it might be used for x2, x3, x4 applications, and for higher voltages perhaps where each stage gains vastly more voltage.

Any advice welcome!

[dom0]

Which leads me to ask the question, is the voltage multiplier used much in practice?

For 300 V? No. Add two zeroes and you want a multiplier.

[uncle_bob]

I have a requirement to generate approximately 300VDC (low currents, no more than a few mA probably) from a 5V USB supply. It's a prototype for a portable system so total footprint, efficiency and weight are important. The prototype will be an Arduino shield hence USB supply.

I've looked at two approaches; 1) Cockcroft Walton voltage multiplier and 2) flyback boost converter.

1)
I looked at boosting 5V up to 24V using a standard DC-DC converter and then driving a voltage multiplier up to 300V. I simulated this in LTSpice and due to diminishing returns of each successive stage, it looks like it would need over 20 stages, which would need a large board area. And this is using ideal components. Increasing the frequency of the AC input appears to increase the output voltage, how else can I improve the efficiency of this topology? Increasing the initial voltage would of course help, which could perhaps be done by chopping the DC into a step up transformer prior to the voltage multiplier. Not sure if this would save board space, or how big/heavy it would need to be, but I'm less concerned with weight than I am with board size and efficiency. I'm happy to share the simulation.

2)
I then looked at camera flash circuits which are essentially flyback converters. In terms of component count this approach looks much better. Using a (small) 1:100 transformer the 5V input should easily reach 300V. But how does the efficiency of this compare with the voltage multiplier?

For me the flyback is more favorable at the moment. Which leads me to ask the question, is the voltage multiplier used much in practice? I guess it might be used for x2, x3, x4 applications, and for higher voltages perhaps where each stage gains vastly more voltage.

Any advice welcome!

Hi

Ok, there is a 60:1 ratio between your 5V USB and the 300V output. If a "couple of ma" means 10 ma, a 100% efficient device will pull 600 ma of of the USB. That may (or may not) be available. It depends on exactly what you intend to plug this into.

The same current comes out to 30W. That's a pretty big converter as these things go. If it's 70% efficient, it will be dumping almost 10W of heat. It also will be pulling closer to 900 ma than 600.

No, you never said 10 ma. It's a nice convent round number that is easy to scale things from. If your actual current is 3 ma, cut everything by a factor of 3.

If you are willing to get into the DIY magnetics stuff, there are a lot of ways to go. If you want to do this with stock parts, not so much. The simple (and likely safest) way to go is a simple transformer. Wind it up with a 10V center tapped primary and a 300V secondary. Driving it is just finding a couple of MOSFETs and few other bits and pieces.

Bob

[dom0]

It's just 3 W, not 30 W. So power dissipation is not really an issue, but the 600 mA would still be... problematic. At least for any single port. USB PD is probably not widespread enough yet.

[timb]

There's a few Linear app notes from Jim Williams on high voltage converters, and LT has good parts for this as well.

AN118 High Voltage, Low Noise DC/DC Converters: http://cds.linear.com/docs/en/application-note/AN118fb.pdf

AN81 Ultracompact LCD Backlight Inverters: http://cds.linear.com/docs/en/application-note/an81f.pdf

AN98 (Has a 0 to 300V Output DCDC Converter circuit that's super simple!) http://cds.linear.com/docs/en/application-note/an98f.pdf

I've adapted one of the AN81 (LCD backlight) circuits to a 500V power supply before. I'm currently working on a high voltage low noise power supply from some of the AN118 examples as well.

Anyway, those should be a good starting place.

[Digibin]

Ok, there is a 60:1 ratio between your 5V USB and the 300V output. If a "couple of ma" means 10 ma, a 100% efficient device will pull 600 ma of of the USB. That may (or may not) be available. It depends on exactly what you intend to plug this into.

Milliamps is probably an over-exaggeration. Truth is I don't yet know the current, but it's likely much less than mA.

There's a few Linear app notes from Jim Williams on high voltage converters, and LT has good parts for this as well.

AN118 High Voltage, Low Noise DC/DC Converters: http://cds.linear.com/docs/en/application-note/AN118fb.pdf

AN81 Ultracompact LCD Backlight Inverters: http://cds.linear.com/docs/en/application-note/an81f.pdf

AN98 (Has a 0 to 300V Output DCDC Converter circuit that's super simple!) http://cds.linear.com/docs/en/application-note/an98f.pdf

I've adapted one of the AN81 (LCD backlight) circuits to a 500V power supply before. I'm currently working on a high voltage low noise power supply from some of the AN118 examples as well.

Anyway, those should be a good starting place.

This was very useful, thanks. The circuit in Fig 44 in AN98 looks like the simplest solution. By the looks of it these photoflash chargers would be more efficient at low output currents than a typical flyback converter.

[uncle_bob]

Milliamps is probably an over-exaggeration. Truth is I don't yet know the current, but it's likely much less than mA.

========

This was very useful, thanks. The circuit in Fig 44 in AN98 looks like the simplest solution. By the looks of it these photoflash chargers would be more efficient at low output currents than a typical flyback converter.

Hi

Unless you have ripple or regulation limitations, a "spark coil" circuit is an easy way to go. A photoflash setup is not sensitive to ripple and regulation is a "dropped 10% fire back up" nature. If those parameters are ok in your application, go for it.

Bob

[Zero999]

What's your budget?

If you don't mind splashing out, what about a ready built DC-DC converter?


http://docs-europe.electrocomponents.com/webdocs/135f/0900766b8135f19a.pdf
http://uk.rs-online.com/web/p/isolated-dc-dc-converters/7331109/

[timb]

Ok, there is a 60:1 ratio between your 5V USB and the 300V output. If a "couple of ma" means 10 ma, a 100% efficient device will pull 600 ma of of the USB. That may (or may not) be available. It depends on exactly what you intend to plug this into.

Milliamps is probably an over-exaggeration. Truth is I don't yet know the current, but it's likely much less than mA.

There's a few Linear app notes from Jim Williams on high voltage converters, and LT has good parts for this as well.

AN118 High Voltage, Low Noise DC/DC Converters: http://cds.linear.com/docs/en/application-note/AN118fb.pdf

AN81 Ultracompact LCD Backlight Inverters: http://cds.linear.com/docs/en/application-note/an81f.pdf

AN98 (Has a 0 to 300V Output DCDC Converter circuit that's super simple!) http://cds.linear.com/docs/en/application-note/an98f.pdf

I've adapted one of the AN81 (LCD backlight) circuits to a 500V power supply before. I'm currently working on a high voltage low noise power supply from some of the AN118 examples as well.

Anyway, those should be a good starting place.

This was very useful, thanks. The circuit in Fig 44 in AN98 looks like the simplest solution. By the looks of it these photoflash chargers would be more efficient at low output currents than a typical flyback converter.

Yeah, they are good parts. However, Fig. 44 has about 2V of ripple due to the hysteresis used in the feedback loop. If you can't live with that, Fig. 46 adds a simple post regulator which brings it down from 2V to 2mV. The post regulator is very simple and uses an Op Amp, Zener Diode and 4 Transistors, two of them cascoded. Basically a discrete high voltage linear regulator. Since it's only 5mV max output, you don't need massive heatsinks or anything.

[Someone]

Milliamps is probably an over-exaggeration. Truth is I don't yet know the current, but it's likely much less than mA.

Up to several W of power (1mA @ your 300V) it is easy enough to use a boost convertor, it can be done in a single stage but you can use an output autotransformer, or an output voltage multiplier to reduce the stress on your switching element.

http://pdfserv.maximintegrated.com/en/an/AN1109.pdf
http://www.iaeng.org/publication/WCECS2010/WCECS2010_pp904-908.pdf

[Psi]

These things are mostly unregulated, so with no load they put out like 150V on the 72V output. And they are isolated.
So you might be able to get away with two in series for 300VDC at very small currents.
Otherwise you would need 3.
http://www.digikey.com/product-detail/en/NMT0572SC/811-1583-5-ND/1927144

[Digibin]

Unless you have ripple or regulation limitations, a "spark coil" circuit is an easy way to go. A photoflash setup is not sensitive to ripple and regulation is a "dropped 10% fire back up" nature. If those parameters are ok in your application, go for it.

Yeah, they are good parts. However, Fig. 44 has about 2V of ripple due to the hysteresis used in the feedback loop.

This is fine - the ripple isn't important.

What's your budget?

If you don't mind splashing out, what about a ready built DC-DC converter?

These things are mostly unregulated, so with no load they put out like 150V on the 72V output. And they are isolated.
So you might be able to get away with two in series for 300VDC at very small currents.
Otherwise you would need 3.
http://www.digikey.com/product-detail/en/NMT0572SC/811-1583-5-ND/1927144

This is an option perhaps for the prototype, but I'd rather design something to reduce cost.

Up to several W of power (1mA @ your 300V) it is easy enough to use a boost convertor, it can be done in a single stage but you can use an output autotransformer, or an output voltage multiplier to reduce the stress on your switching element.

http://pdfserv.maximintegrated.com/en/an/AN1109.pdf
http://www.iaeng.org/publication/WCECS2010/WCECS2010_pp904-908.pdf

c. Flyback with miniaturized magnetics, such as ATB322524, in DCM or CCM (mostly preferred, high efficiency and low noise due to lower peak current).

I strongly recommend you to read R.W. Erickson's lecture notes.
http://ecee.colorado.edu/copec/book/slides/slidedir.html
Chapter 2, 3, 5, 6. You can go through them within as short as 2 hours, and learn essentials of power electronics that everyone designing SMPS should know.

This is also interesting - thanks for the links. The main concern here is efficiency - what sort of efficiency could I except from a flyback converter operating at very low currents? Efficiency vs output load graphs always show dropping efficiency at lighter loads, but these are designed for a particular output current. If designed for say <1mA could you still achieve good efficiency?

Another requirement I forgot to mention is that it does need to be isolated. So an isolated flyback circuit would fit the bill. Otherwise I could adapt the photoflash charger feedback for isolation.

[uncle_bob]

This is also interesting - thanks for the links. The main concern here is efficiency - what sort of efficiency could I except from a flyback converter operating at very low currents? Efficiency vs output load graphs always show dropping efficiency at lighter loads, but these are designed for a particular output current. If designed for say <1mA could you still achieve good efficiency?

Another requirement I forgot to mention is that it does need to be isolated. So an isolated flyback circuit would fit the bill. Otherwise I could adapt the photoflash charger feedback for isolation.

Hi

Best guess on an isolated supply at low currents would be in the 30% to 50% range. The closer you get to zero current, the closer you get to zero efficiency. A much better way to look at it is - how much current (or power) does it pull at zero output?

Depending on the circuit (and the isolated feedback approach) that could be 10's or 100' of mw. If you have 10 ua out at 300V, that is 3 mw. Even at 10 mw idle power, you have an awful efficiency number. However, 13 mw total is pretty low power.

If you are using aluminum electrolytic caps on the output, they might have a 1 ma leakage current. They might have more. That 300 mw alone would blow the numbers above away entirely. Regardless of the circuit, parts selection (and budget) will place some limits on what you can do.

Bob

[dom0]

I agree, efficiency specifications only make sense when the power drawn is much larger than than the power requirements of the controller, switch driver etc.

1 mA is probably a bit on the high side for the output cap. A small cap of 400.. 450 V WV and 10 µF or so should have less than 100, maybe 200 µA specified leakage, even for rather ... cheap types. Iirc Nippons KXJ series (I used them for something else) had something like 5 % of C*V product leakage current.

[uncle_bob]

Hi

One of the classic ways to get the efficiency up is to increase the output cap. You go up to a few hundred uf. Then you put the controller to sleep (as in sub 10 ua drain at low voltage) for some reasonable period of time. After maybe 100 ms or so, you wake up, see how the output is doing. If needed you give it a bit of a bump. If not, you go back to sleep.  How well this works and how big a cap you need depends a lot on the current draw and how dynamic it is.

Bob

[Digibin]

I agree, efficiency specifications only make sense when the power drawn is much larger than than the power requirements of the controller, switch driver etc.

Yes agreed - I guess I mean power conservative rather than power efficient.

Exploring options I've looked at adding an optocoupler to the feedback loop of the photoflash charger circuit shown in Fig44 here. Due to the forward current requirement of the LED this would increase power consumption. So I need it to be low current. I don't know how little current you can typically get away with in optocouplers, but this one appears to work down to less than 100uA of LED current according to the datasheet. Could I use this, bias LED from 300V for 1mA, giving roughly 60mA collector current, tie collector to 5V and use 50R emitter resistor to ground to give output of 3V. This would give 0-3V at the output indicating approximately 0-300V, essentially replacing the voltage divider in the photoflash charger circuit.

[uncle_bob]

I agree, efficiency specifications only make sense when the power drawn is much larger than than the power requirements of the controller, switch driver etc.

Yes agreed - I guess I mean power conservative rather than power efficient.

Exploring options I've looked at adding an optocoupler to the feedback loop of the photoflash charger circuit shown in Fig44 here. Due to the forward current requirement of the LED this would increase power consumption. So I need it to be low current. I don't know how little current you can typically get away with in optocouplers, but this one appears to work down to less than 100uA of LED current according to the datasheet. Could I use this, bias LED from 300V for 1mA, giving roughly 60mA collector current, tie collector to 5V and use 50R emitter resistor to ground to give output of 3V. This would give 0-3V at the output indicating approximately 0-300V, essentially replacing the voltage divider in the photoflash charger circuit.

Hi

Ok, a few issues with optical feedback:

1) Low current = high noise
2) transfer function is temperature dependent
3) transfer function is not particularly linear over the whole range.
4) LED is temperature dependent (not a big deal run straight off 300V)

The normal solution in any case is to put another winding (or tap) on the transformer and generate a low voltage supply on the isolated side. You can then move an op amp over there and play various tricks with the control loop. The other solution is voltage feedback. That means you are only monitoring when you are driving the transformer. (no sleep mode).

Bob

[Digibin]

The trouble with that is it's more difficult to find a suitable part of the shelf. It starts becoming quite specific; 1:10 turns ratio or thereabouts, aux secondary winding plus other specs need to be right. And importantly it needs to be small. Most of the miniature 1:10 flyback transformers I've found just have the four pins i.e. one winding on each side, such as this or this. Some of the larger ones have additional windings but I only have a small area to play with (Arduino shield). I've had a good look but can't see anything on distributor websites.

As an alternative perhaps I could design something around the LT8301 since this regulates an isolated output based on primary side waveforms, with no third winding or optical feedback. Could this work?

[dannyf]

Oscillator, + hf transformer + diode voltage doublers (optional). Old analog TV has a lot of those things.

[edavid]

You should also look at Nixie power supply designs... probably more relevant to your low-current needs than photoflash chargers.

[uncle_bob]

Hi

The whole magnetics thing does indeed get a bit exciting. The normal way to do this is to tear apart an old transformer and re-wind it. Every time I've gotten involved with this sort of thing, we have would our own parts. Off the shelf just isn't practical.

Bob