Design ideas for a low-count part, low power isolated DC-DC power supply?

[SiliconWizard]

As the title suggest, I'm looking for ideas to implement an isolated DC-DC power supply with as few parts as possible. Targetting low input and output voltage (5V max) and low max output power (50 mA max). Efficency would not matter much as long as it's not ridiculously low (thus > 20-30% would be great).

Obviously, flyback topologies are popular, but I would like to find alternate ways, avoiding the need of a transformer if possible.

Any ideas welcome.

Note: a 1:1 converter is fine and even preferred.

[mariush]

Adum5000 : https://www.analog.com/media/en/technical-documentation/data-sheets/ADuM5000.pdf

5v to 5v 100mA digital isolator .. around 33% efficient. So you'd need 150mA at the input to get 50mA at output.

Higher dc input voltage? Any switching regulator capable of 150mA output would do, go for higher frequency to get smaller inductors, ceramic capacitors etc

for example AOZ1282 (max 36v in, max 30.6v out , 600mA, 1 mhz) .. that's around 85-95% efficient.


Want AC to DC or something like that?

there's regulators like LM5008 which can accept up 95v input voltage.  Assuming around 70% efficiency when doing 90v in, 5v @150mA out, you're looking at around 1100mW or  1100 / 90v = ~ 12mA

To make sure you don't go over 90v, you could always use a basic linear regulator like  LR8K4 for example - takes up to 440v , outputs desired voltage (maximum Vin - 12v) at up to 20mA
So you could have AC in - > diode for half wave rectification -> small capacitor to have min 100v at ~ 15mA all the time -> LR8K4 to do 100v+ to 90v 12mA -> LM5008 to do 90v-> 5v 150mA -> adm5000 to isolate.

[Marco]

Adum5000

Technically still using a transformer, just a tiny integrated one, but then that's hard to avoid.

[schmitt trigger]

Your post title includes the word "isolated".
That on its own, suggests a transformer being employed somewhere.

The reason flybacks are so extremely popular for low-power, isolated supplies is precisely because of its low parts count, and low cost.

[SiliconWizard]

Adum5000

Technically still using a transformer, just a tiny integrated one, but then that's hard to avoid.

Yes, I've already used some of those "isoPower" isolators from AD. They work fine and pretty compact.
And yes, they use integrated micro-transformers as far as I know.
From experience, they tend to have rather high inrush currents at power on and they are pretty noisy (can make EMC testing fail even when you think you have routed them properly - they need extra care). But they do work fine.

I'm looking for ideas for a discrete solution. I'd like to avoid transformers and although that may seem pretty difficult at first sight, the challenge is to find other approaches. Maybe capacitive solutions for instance.

[mikeselectricstuff]

Why not just buy an off-the-shelf DC-DC module ? That easily satisfies "as few parts as possible"

[nctnico]

I'm looking for ideas for a discrete solution. I'd like to avoid transformers and although that may seem pretty difficult at first sight, the challenge is to find other approaches. Maybe capacitive solutions for instance.

Capacitive should be doable with 'safety' capacitors (class X or Y but you have to look up which ones exactly).

[mariush]

A bit on the ridiculous side, but you could have a bunch of tiny SMD leds on the circuit board and then place a solar cell on top of it... maybe have a layer of glass between them for isolation purposes.
Horrible efficiency obviously. 

[SiliconWizard]

I'm looking for ideas for a discrete solution. I'd like to avoid transformers and although that may seem pretty difficult at first sight, the challenge is to find other approaches. Maybe capacitive solutions for instance.

Capacitive should be doable with 'safety' capacitors (class X or Y but you have to look up which ones exactly).

Right... I have something working at least on simulation now.

I guess that to get an acceptable isolation voltage this would require rather bulky capacitors. Does that sound right?

[SiliconWizard]

A bit on the ridiculous side, but you could have a bunch of tiny SMD leds on the circuit board and then place a solar cell on top of it... maybe have a layer of glass between them for isolation purposes.
Horrible efficiency obviously.

I haven't though of that actually. Inefficient as hell as is but I wasn't thinking of optronics solutions, so that gives something to investigate!

[Marco]

If you want to keep the coupling capacitance in the pF range you just can't switch fast enough with capacitive coupling, since you get picojoule range energy transfer per cycle.

[SiliconWizard]

Why not just buy an off-the-shelf DC-DC module ? That easily satisfies "as few parts as possible"

Challenge and cost, mostly. Also, most modules I have seen so far are usually for higher output power and a bit too bulky, but if you know of miniature ones and reasonably priced, that could be an option.

[SiliconWizard]

If you want to keep the coupling capacitance in the pF range you just can't switch fast enough with capacitive coupling, since you get picojoule range energy transfer per cycle.

True, I'm fine with nF-range values or even around 1 µF, although the question is the isolation voltage you can get while keeping the caps reasonably small.

[Wolfgang]

As the title suggest, I'm looking for ideas to implement an isolated DC-DC power supply with as few parts as possible. Targetting low input and output voltage (5V max) and low max output power (50 mA max). Efficency would not matter much as long as it's not ridiculously low (thus > 20-30% would be great).

Obviously, flyback topologies are popular, but I would like to find alternate ways, avoiding the need of a transformer if possible.

Any ideas welcome.

Note: a 1:1 converter is fine and even preferred.

You of course know that you can buy stuff like this in potted modules for just a few €.

I tried to make something like this because I needed a version with extra low ripple running from USB.
Commercial modules are fairly noisy.

https://electronicprojectsforfun.wordpress.com/power-supplies/a-usb-powered-15v-supply/

This has 15V output, but the circuit can be easily adapted to 5V if needed.

[SiliconWizard]

I should definitely have added that I was preferably not looking for an off-the-shelf solution.

Anyway, as far as transformerless goes, the capacitive approach seems the best fit so far, although it's not an ideal means of isolation. (Reminds me of Dave's 'Does current flow through a capacitor' topic )
I'm currently simulating a simple solution which bears similarities with this: https://www.maximintegrated.com/en/app-notes/index.mvp/id/4553 , only I just use 2 diodes instead of a full bridge, which is fine for what I need (and has lower loss), and no specific h-bridge IC. This should do the trick although again it's not ideal isolation-wise, but my main goal here is not about safety, it's mainly for getting a floating ground.

This will certainly fit my requirements but obviously the safer approach still seems to use some kind of flyback DC/DC converter. The ADuM isoPower series is nice, although it's not perfect either.

In fact it sort of all comes down to transmitting power wirelessly, and there aren't that many options for this. Inductive, capacitive, photovoltaic... sure I could also consider a purely mechanical solution but I don't think that would fit my volume, cost and efficiency requirements - not to mention being mostly silent.

[Marco]

I'm currently simulating a simple solution which bears similarities with this: https://www.maximintegrated.com/en/app-notes/index.mvp/id/4553 , only I just use 2 diodes instead of a full bridge, which is fine for what I need (and has lower loss), and no specific h-bridge IC.

Simulate what happens if you ground both sides, I think it goes wonky.

[SiliconWizard]

I'm currently simulating a simple solution which bears similarities with this: https://www.maximintegrated.com/en/app-notes/index.mvp/id/4553 , only I just use 2 diodes instead of a full bridge, which is fine for what I need (and has lower loss), and no specific h-bridge IC.

Simulate what happens if you ground both sides, I think it goes wonky.

I just tried. It changes efficiency a bit, and I get a slightly higher output voltage (which I would have expected) but still works fine.
Will post a schematic shortly.

Note: I'm not using an h-bridge topology. If I did, obviously grounding the "isolated ground" would short-circuit one of the H-bridge outputs. It would not like that.

[Marco]

I think it essentially shorts out one of the capacitors, there's still no DC path, but the fact that it changes the output voltage is a bit annoying.

[SiliconWizard]

I think it essentially shorts out one of the capacitors, there's still no DC path, but the fact that it changes the output voltage is a bit annoying.

I don't mind. And this should be a degenerate case, as you'd obviously lose the isolation - I only expect the system not to be destroyed by it.

You nailed an important factor though, as if using an H-bridge as suggested in Maxim's note, this would lead to a short-circuit of one of its outputs. If it's protected against short-circuits, fine, but if not...

[Marco]

It doesn't really short it ... it just makes the MOSFET burn all the energy put into one of the coupling the capacitor each cycle, which creates a limited amount of power, but still too high for comfort.

[nctnico]

Another wild idea: how about using coils for wireless charging? These are small and self adhesive. Stick one on the top and one on the bottom of your board and you'll have your transformer with a hefty isolation barrier. Using resonant mode for energy transfer is probably a good idea.

[SiliconWizard]

Another wild idea: how about using coils for wireless charging? These are small and self adhesive. Stick one on the top and one on the bottom of your board and you'll have your transformer with a hefty isolation barrier. Using resonant mode for energy transfer is probably a good idea.

Obviously a bit larger than I would like as far as area goes, although for a few tens of mW we could probably get away with pretty small coils. Regarding height, it's hard to beat though!
I like the idea. Will also work on that.

Edit: given the very short distance (a board's thickness) between the coupled coils and the resonant mode, we may even get enough power with coils routed on the PCB directly with a limited number of turns. That gets interesting! Noise coupled to surrounding parts of the PCB may be a concern, though.

[SiliconWizard]

It doesn't really short it ... it just makes the MOSFET burn all the energy put into one of the coupling the capacitor each cycle, which creates a limited amount of power, but still too high for comfort.

Right... they use a full diode bridge so that wouldn't be a short. I was thinking of my approach which uses an half bridge basically, and thus you'd get a short-circuit. With an H-bridge driver, a full diode bridge prevents that problem, so it's preferable. The H-bridge approach gets you a x2 factor (minus the Vf and losses), so I took an half-bridge approach which was simpler and would do the trick in my case.

[not1xor1]

As the title suggest, I'm looking for ideas to implement an isolated DC-DC power supply with as few parts as possible. Targetting low input and output voltage (5V max) and low max output power (50 mA max). Efficency would not matter much as long as it's not ridiculously low (thus > 20-30% would be great).

Obviously, flyback topologies are popular, but I would like to find alternate ways, avoiding the need of a transformer if possible.

Any ideas welcome.

Note: a 1:1 converter is fine and even preferred.

what about just an high frequency oscillator and a transformator with small toroid core?
a  555 or even a buffered cmos logic oscillator or (I do not know if there is a low voltage part) just a self oscillating mosfet driver
as far as I remember there are various ANs on the subject on the net

[capt bullshot]

Lowest part count on such DC/DC usually is kind of a self-oscillating Royer converter.
Many of the small (1W) DC/DC modules are built internally this way. One needs a transformer with a center tapped primary, another feedback primary and a secondary winding, two transistors and two resistors on the primary side, a rectifier and a capacitor on the secondary. With a bit more effort, the converter can also be done with only one center tapped primary.

Often, if there's a buck regulator in your circuit anyway, one can couple a small transformer to that regulator and have an isolated output with very few components involved. Anyway, any half-decent solution with low part count involves a transformer.

Edit: There are smallish driver chips (e.g. the TI SN6505B) that ease the job of building such converters. Beware, this chip might cost more than an optimized converter built from discrete components.