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

[Giaime]

I used these in the past:
http://www.ti.com/isolation/transformer-driver/products.html

With those transformers:
https://katalog.we-online.de/en/icref/SN6501-Rev-G_Push-Pull_4?sid=3832d0653a

[schmitt trigger]

Thanks for bringing the Royer converter.
Indeed, it cannot be any simpler, component wise, than this converter.

Although a caveat, if I recall correctly, is that the transformer core should employ a “square loop” material for optimal efficiency.
But for the low power level involved, perhaps any power ferrite may be sufficient.

[ezalys]

You might consider a resonant royer... think royer but with the primary shunted with a capacitor.

[Wolfgang]

Hi,

a square hysteresis loop material is no requirement. A normal transformer is absolutely sufficient at small power levels.
I used old ISDN transformers for my low-noise converters and some energy harversting experiments, also employing Royer
converters:

electronicprojectsforfun.wordpress.com/energy-harvesting/

[ezalys]

Also, how many layers are you targeting? You might just stick the transformer inside the PCB... then the inductor is zero parts. Then you do resonant royer or forward or something.

[blackdog]

Hi

Last year i needed a low noise low power power supply with 3 to 5 outputs.
So i designed something that was not switching and it uses a small audio power IC in an Wienbridge configuration.

This is a topic on a Duchs forum om mine, use Google translate if you wish.
https://www.circuitsonline.net/forum/view/138876/1#highlight=voeding

And this is the main schematic witch shows only two seperate outputs.


Just for some inspiration... 

Kind regards,
Bram

[SiliconWizard]

Thanks for the additional ideas and the Royer converter.

@ezalys: as nctnico suggested, using just flat coils on both sides of the PCB instead of a true transformer is much simpler and should work well enough with resonant coupling. Would not require a minimal PCB thickness.
If coils directly as PCB traces, as I further suggested, give a decent efficiency, it would really be a low-cost and low part count solution. I like the idea.

One additional benefit of using resonant coupling is that you can get waveforms close to sine waves if your Q factor is decent, which yields way better results in terms of EMI than flyback converters.
This is also what you get with blackdog's solution, but the quiescent current would certainly be much higher.

[T3sl4co1l]

The smallest/cheapest DC-DC converters are just multivibrators with a transformer hanging off it. No protection, no control, only raw AC.  They tend not to go into this aspect in much detail, in the datasheets, but it is apparent from the characteristics they must provide (e.g., output regulation is sloppy and PSRR is nil).

You'll find circuits can in fact be simpler, but even more compromises must be made.



This is the basic plan of the blocking oscillator.  The transistor turns on hard, due to positive feedback, then turns off due to rising inductor current and/or falling base current (due to the relatively small value of Cbb, which discharges through Rb); in either case, hFE rising until it comes out of saturation.  When it switches off, the voltage flies back, where you employ a clamp diode on the collector (for a boost or SEPIC converter) or secondary (for an isolated flyback supply).

Like the Royer, a feedback winding is needed, which itself may be more expensive (i.e., custom magnetics) than throwing a few more transistors at it (especially transistors with nearly zero incremental cost, i.e., on an IC).

Discrete circuits are entertaining, but do not fool yourself into thinking they are in any way size- or cost-effective except in very rare cases today.   (I'd say, the case of making a million qty run is rare enough to count.)

Tim

[schmitt trigger]

I used old ISDN transformers for my low-noise converters and some energy harversting experiments, also employing Royer
converters:

electronicprojectsforfun.wordpress.com/energy-harvesting/

Very interesting comparison! Thanks for sharing.

You comment that a low drop rectifier is mandatory, is absolutely correct to extract the last millivolt.
Do you think synchronous rectification would significantly help here? I understand that the circuit's complexity would increase, and thus this suggestion may be only a brain-fart.

[Wolfgang]

I used old ISDN transformers for my low-noise converters and some energy harversting experiments, also employing Royer
converters:

electronicprojectsforfun.wordpress.com/energy-harvesting/

Very interesting comparison! Thanks for sharing.

You comment that a low drop rectifier is mandatory, is absolutely correct to extract the last millivolt.
Do you think synchronous rectification would significantly help here? I understand that the circuit's complexity would increase, and thus this suggestion may be only a brain-fart.

Sync rectification could help, thats correct. There was some IC that did it very elegantly LT4320 but this does it to 600Hz only.
Maybe there is a faster part just as simple ?

[SiliconWizard]

Sync rectification could help, thats correct. There was some IC that did it very elegantly LT4320 but this does it to 600Hz only.
Maybe there is a faster part just as simple ?

https://www.infineon.com/cms/en/product/power/ac-dc-power-conversion/ac-dc-pwm-pfc-controller/synchronous-rectification-ics/
(Minimum working voltage will not fit all applications though, but it's a lot lower than for the LT4320.)
Some interesting stuff: https://www.monolithicpower.com/en/products/ac-dc-power-conversion/synchronous-rectifiers.html

[IDEngineer]

they are pretty noisy (can make EMC testing fail even when you think you have routed them properly - they need extra care).

"Pretty noisy" doesn't begin to describe those nasty little transmitters. I wasted a couple of weeks once trying to quiet them down, both radiated and conducted. In some cases the PRIMARY is noisier than the secondary, corrupting the incoming supply. TI devotes many pages in their appnotes trying to put RFI/EMI lipstick on these pigs... I tried all of their "solutions" and none was even barely acceptable. I'd be embarrassed to have my name associated with something that noisy.

As I've noted elsewhere in these forums, also watch out for isolation capacitance. Think of it as a qualitative measurement of the coupling across the supposed "isolation" barrier. It's generally ghastly, with medical-rated modules being modestly better for horrific increases in price. Note that being expressed in terms of capacitance means the coupling is frequency dependent... anything high frequency on one side is cheerfully conveyed right through the module to the other side.

I'll never touch those modules again if I can help it. I have lots of samples from many suppliers here, in various voltages and configurations, and hope I never have to work with them again. My recommendation: Stay far, far away unless you're building something like Widlar's Hassler.

[Wolfgang]

The rant is always the same; People want chips in a microscopic package (meaning low dissipation capabilities). In the sake of energy efficiency, switching times are reduced as much as they can, creating a UWB noise level all around the circuit. Then, some cosmetic RFI measures try to amend things until it is tolerable.

If you want quiet circuits, stay away from fast components. Efficiency will suffer somewhat, but RFI problems are a lot less.

There are some Jim Williams appnotes where switching speed was deliberately reduced to keep RFI problems at bay.

[SiliconWizard]

they are pretty noisy (can make EMC testing fail even when you think you have routed them properly - they need extra care).

"Pretty noisy" doesn't begin to describe those nasty little transmitters. I wasted a couple of weeks once trying to quiet them down,
(...)
I'll never touch those modules again if I can help it. I have lots of samples from many suppliers here, in various voltages and configurations, and hope I never have to work with them again. My recommendation: Stay far, far away unless you're building something like Widlar's Hassler.

Yep, that partly explains why I'm looking for alternatives, even if they have to be made of discrete parts.
Apart from the noise issues, the startup can be a huge problem as well especially if you're powering them from switching regulators. Had a few headaches. Lastly, I also noticed that they tended to run very warm even at minimal load.

[ANTALIFE]

https://power.murata.com/data/power/ncl/kdc_nxe1.pdf

[Marco]

Why do they all use these ridiculous poor regulation, minimum load architectures instead of something nice with pulse skipping and feedback?

[T3sl4co1l]

Why do they all use these ridiculous poor regulation, minimum load architectures instead of something nice with pulse skipping and feedback?

They do.

The ones that do, cost twice as much.

Tim

[LukeW]

"isolated DC-DC power supply"

"avoiding the need of a transformer if possible"

Use a big light bulb and a solar cell.

Or a motor-generator.

[Marco]

The ones that do, cost twice as much.

Why though? Does Mean Well put a nugget of gold in each and every regulated module? Efficient market hypothesis is failing here.

[T3sl4co1l]

It's all about market.  Correct.

As is the case with proof-by-contradiction, we can examine our premises and determine which ones are failing.

A nugget of gold, no, that would be uneconomical by any stretch.

More silicon?  Likely.  A controller/regulator is needed, and cheap though they are in quantity, they're not quite as cheap as a pair of jellybean BJTs.

What does that leave?  A somewhat higher price leads to lower production numbers, which leads to much higher production costs per part.

Likely, the cost difference at equal quantity is only, say, 20 or 40%, but the quantity difference magnifies it.

Furthermore: since it's impossible for a regulated converter to compete in the same space as an unregulated one, they might as well move to an upscale market, targeting reinforced isolation for example, or reduced EMI.  RECOM's products are notoriously expensive ($20 for a 5W converter?), but equally high in quality and capability.  So, that would account for the last stage of cost difference.

Tim

[Marco]

A controller/regulator is needed

If you can't make an ASIC just use a microcontroller on both sides. The microcontroller can have a regulator, function as a controller and function as a digital isolator with a little air core transformer on the PCB. It won't cost much more than a regulator or controller, packaging dominates at that scale. So basically the extra cost is the micro on the secondary side, big whoop. Also some extra engineering cost, but look at the Murata NXE1 ... compared to the investments for embedding a toroid in the PCB stack I don't think it's all that relevant.

Their highest tech module as far as physical design goes has a stone age controller, without any real need for that given component costs. It's a crying shame.