Super low power isolated MOSFET driver

[Klaus Kragelund]

On Tuesday, January 21, 2020 at 1:22:12 AM UTC+1, Klaus Kragelund wrote:

On Monday, January 20, 2020 at 12:34:31 AM UTC+1, Tim Williams wrote:
The bipolar 555 is piss, obviously you'd use a CMOS model, yeah.

I've done that before (happened to be a discrete circuit). But have you
thought about common mode immunity yet?

Consider using something like this,
https://www.digikey.com/product-detail/en/pulse-electronics-power/PH9185.011NLT/553-2052-1-ND/3503429
and don't simply nod at the CT windings: take full advantage of them. The
CMRR of this part is surprisingly good, up to surprisingly high frequencies.
You will have a very hard time doing it unbalanced.

Even an Ethernet transformer will do, if you don't mind the 1kV isoation
rating (in this case obviously you would mind). The flux is quite small,
but more than adequate for little pulses, and the isolation barrier is
smaller (lower capacitance). Double balanced design again, and usually
comes with CMCs included (for even more immunity).

Forget if you can find COTS Ethernet transformers with reinforced kV+
isolation, if so that may be handy. Huh, probably same price as the Pulse
part anyway...

Another approach that may be of interest, pulse or frequency modulation..
ASK through one transformer allows you to maintain high-side power, while
filtering the signal for immunity.

OOK through two transformers, you could do the same while sharpening it with
a flip-flop (in essence, detect and filter two signal paths, and route them
to your 555's inputs).

PSK would be... interesting to decode without a clock reference (two
transformers would be easy, one would be harder), but could be unambiguous
(in phase = on, etc.; or 90 deg. to the right = on, etc.).

FM, you could use a crude discriminator, like a missing-pulse detector; give
or take additional filtering before or after, you can clean up most
interference.

Could all be done at VHF too, in which case planar magnetics are feasible
(coreless or otherwise), or even at UHF where interference from power
supplies is unlikely and the resulting bandwidth (which might be say 5-10%
of Fc) is competitive with monolithic drivers.


You can do isolation with spiral coreless coils quite simple:

https://www.researchgate.net/publication/224224765_Insulating_IGBT_driver_with_PCB_integrated_capacitive_coupling_elements

The important parameter to look for is the parasitic capacitance over the isolation barrier, which creates problems with conducted emissions and susceptibility to fast dV/dt transients on the switching node

Another nice paper is this one:

https://www.semanticscholar.org/paper/Coreless-Printed-Circuit-Board-(-PCB-)-Transformers-Sain-Chen/f683ebe9da5a97ce8b9e69a7105105d78979dc41

About modulation, nothing really beats a AM. Just start an oscillator when you need to trigger the FET, transmit the pulse through the transformer, impedance match, rectify and apply to the gate. With <10MHz osc freq and a bleeder resistor you can easily get sub us propagation delays

Cheers

Klaus

Just for fun, see this paper about a coreless PCB transformer flyback converter with high efficiency:

https://www.researchgate.net/publication/263285748_A_ZVS_Flyback_DC-DC_Converter_using_Multilayered_Coreless_Printed-Circuit_Board_PCB_Step-down_Power_Transformer

For a flyback core losses can be high, so using a converter without a core is not a bad way to tackle that problem

Cheers

Klaus

 

[whit3rd]

On Monday, January 20, 2020 at 11:06:35 AM UTC-8, jla...@highlandsniptechnology.com wrote:

It might be scary to float a flip-flop up there. High voltage circuits
might make EMI that could confuse a flop.

For megavolt systems, the Faraday cage for the control room can be
located at the hot electrode. Other than bathroom break anxiety, no problems
were reported.

 

[RBlack]

On Sun, 19 Jan 2020 17:34:39 -0600, Tim Williams
<tiwill@seventransistorlabs.com> said:

The bipolar 555 is piss, obviously you'd use a CMOS model, yeah.

I've done that before (happened to be a discrete circuit). But have you
thought about common mode immunity yet?

Consider using something like this,
https://www.digikey.com/product-detail/en/pulse-electronics-power/PH9185.011NLT/553-2052-1-ND/3503429
and don't simply nod at the CT windings: take full advantage of them. The
CMRR of this part is surprisingly good, up to surprisingly high frequencies.
You will have a very hard time doing it unbalanced.

Even an Ethernet transformer will do, if you don't mind the 1kV isoation
rating (in this case obviously you would mind). The flux is quite small,
but more than adequate for little pulses, and the isolation barrier is
smaller (lower capacitance). Double balanced design again, and usually
comes with CMCs included (for even more immunity).

Forget if you can find COTS Ethernet transformers with reinforced kV+
isolation, if so that may be handy. Huh, probably same price as the Pulse
part anyway...

We use this one, rated for 4 kV:

https://octopart.com/749024012-würth+elektronik-20427545

It's intended for Ethernet but we use it to run a serial ADC for a
floating measurement channel. Two sets of windings used in parallel for
the power at a few hundred mW, and one each for the Tx and Rx data at
about 100 kbps.

We buy them direct from the manufacturer, hopefully for a better price
than the distributors are quoting above! This is going in a medical
product and has to be UL listed, so it's still cheaper than paying the
certification costs for a custom transformer.

Another approach that may be of interest, pulse or frequency
modulation.
ASK through one transformer allows you to maintain high-side power, while
filtering the signal for immunity.

OOK through two transformers, you could do the same while sharpening it with
a flip-flop (in essence, detect and filter two signal paths, and route them
to your 555's inputs).

PSK would be... interesting to decode without a clock reference (two
transformers would be easy, one would be harder), but could be unambiguous
(in phase = on, etc.; or 90 deg. to the right = on, etc.).

FM, you could use a crude discriminator, like a missing-pulse detector; give
or take additional filtering before or after, you can clean up most
interference.

Could all be done at VHF too, in which case planar magnetics are feasible
(coreless or otherwise), or even at UHF where interference from power
supplies is unlikely and the resulting bandwidth (which might be say 5-10%
of Fc) is competitive with monolithic drivers.

Tim

 

On Tue, 21 Jan 2020 12:14:36 -0000, RBlack <news@rblack01.plus.com>
wrote:

On Sun, 19 Jan 2020 17:34:39 -0600, Tim Williams
tiwill@seventransistorlabs.com> said:

The bipolar 555 is piss, obviously you'd use a CMOS model, yeah.

I've done that before (happened to be a discrete circuit). But have you
thought about common mode immunity yet?

Consider using something like this,
https://www.digikey.com/product-detail/en/pulse-electronics-power/PH9185.011NLT/553-2052-1-ND/3503429
and don't simply nod at the CT windings: take full advantage of them. The
CMRR of this part is surprisingly good, up to surprisingly high frequencies.
You will have a very hard time doing it unbalanced.

Even an Ethernet transformer will do, if you don't mind the 1kV isoation
rating (in this case obviously you would mind). The flux is quite small,
but more than adequate for little pulses, and the isolation barrier is
smaller (lower capacitance). Double balanced design again, and usually
comes with CMCs included (for even more immunity).

Forget if you can find COTS Ethernet transformers with reinforced kV+
isolation, if so that may be handy. Huh, probably same price as the Pulse
part anyway...


We use this one, rated for 4 kV:

https://octopart.com/749024012-würth+elektronik-20427545

It's intended for Ethernet but we use it to run a serial ADC for a
floating measurement channel. Two sets of windings used in parallel for
the power at a few hundred mW, and one each for the Tx and Rx data at
about 100 kbps.

We buy them direct from the manufacturer, hopefully for a better price
than the distributors are quoting above! This is going in a medical
product and has to be UL listed, so it's still cheaper than paying the
certification costs for a custom transformer.

I'm using some ADUM7703s, each with a commercial dc/dc converter for
the upside power. An FPGA gets the delta-sigma data and filters it for
subsequent processing.



--

John Larkin Highland Technology, Inc

The cork popped merrily, and Lord Peter rose to his feet.
"Bunter", he said, "I give you a toast. The triumph of Instinct over Reason"

 

[Piotr Wyderski]

On 2020-01-21 01:30, Klaus Kragelund wrote:

Just for fun, see this paper about a coreless PCB transformer flyback converter with high efficiency:

https://www.researchgate.net/publication/263285748_A_ZVS_Flyback_DC-DC_Converter_using_Multilayered_Coreless_Printed-Circuit_Board_PCB_Step-down_Power_Transformer

For a flyback core losses can be high, so using a converter without a core is not a bad way to tackle that problem

Wow, that is truly crazy. I would never think that you can get more than
50% of a coreless inductive converter. Thanks a lot, Klaus!

Best regards, Piotr

 

[Klaus Kragelund]

On Wednesday, January 22, 2020 at 9:34:21 PM UTC+1, Piotr Wyderski wrote:

On 2020-01-21 01:30, Klaus Kragelund wrote:

Just for fun, see this paper about a coreless PCB transformer flyback converter with high efficiency:

https://www.researchgate.net/publication/263285748_A_ZVS_Flyback_DC-DC_Converter_using_Multilayered_Coreless_Printed-Circuit_Board_PCB_Step-down_Power_Transformer

For a flyback core losses can be high, so using a converter without a core is not a bad way to tackle that problem

Wow, that is truly crazy. I would never think that you can get more than
50% of a coreless inductive converter. Thanks a lot, Klaus!

And that is just the flyback, which is a horrible topology

If you use a halfbridge resonant inverter driving a coreless transformer you can get a lot high efficiency

Cheers

Klaus