Switchable Isolated Current Source

[azeeman]

Here's a simple wide input voltage range current source with opto-isolator cutoff. I intend to use this as a bootstrap current source for a very wide input voltage range switcher.
Significant startup current is required for the isolated DC to DC converter used for high side gate drive. Because of this the current source needs to be switched off once the low voltage supply has started to reduce power dissipation. If input voltage is 300V and startup current is 20mA, power dissipation will be 6W.
The current source can accept an input voltage from 6V to the MOSFET drain voltage limit. The opto-coupler LED will require current, but both the current and voltage are much lower with much lower power dissipation. If low voltage supply is 12V and LED current is 1mA, power dissipation will be 12mW in the LED and 200mW in the MOSFET gate resistor.
I tried using an opto-coupler with an accessible base terminal as the NPN transistor without success. The NPN transistor should have high gain at very low collector current to allow a large resistance value for R1 and low LED current to turn off the current source.
The large gate resistance and MOSFET input capacitance will act as a low pass filter to protect the MOSFET gate against voltage spikes.
A 2N3904 transistor gives 11mA when the input voltage is 5V and 23mA when the input is 31V. The intended input voltage range is 10V to 300V so some tweaking may be required.

[PGPG]

A 2N3904 transistor gives 11mA when the input voltage is 5V and 23mA when the input is 31V. The intended input voltage range is 10V to 300V so some tweaking may be required.

For small Q2 collector current (for 31V it should be less than 0.1mA) you should not have Vbe higher than 0.7V so current higher than 0.7/43=16mA. If you measure 23mA I would expect that may be your current source is not stable and oscillates.
Replacing R1 with current source you can ensure almost constant Q1 Ic current what should ensure working with the same Vbe for wide voltage range.

[azeeman]

I was using my power supply current meter for testing and it's not very accurate. I just completed testing on the machine where it's going to be used and the current is much more stable. The voltage source is a human powered generator on an elliptical exercise machine. The measured peak voltage that I can generate is 250V which gave a measured current of 15.6mA.
At the lowest possible speed that I could manage, the current was about 13mA and the voltage around 12V. The current does not have to be precise and only has to be sufficient to drive a DC to DC converter used for isolated MOSFET gate drivers. Any excess current will be shunted out by a protective Zener diode.
The circuit will supply current for less than a second before the main switcher takes over so MOSFET temperature rise will be negligible and the same for the Zener.
The main switcher will regulate the voltage to 12V so no current will flow into the Zener during normal operation.
I just added the generator specifications for reference.

[azeeman]

I checked on DigiKey and found the lowest current available for a current regulator is 200uA and the maximum voltage for higher current devices is 190V. The current is unnecessarily high, the voltage is too low and the cost is fairly high.
https://www.digikey.ca/en/products/filter/current-regulation-diodes-transistors/1164?s=N4IgjCBcoEwGwHYLRAYygMwIYBsDOApgDQgD2UA2uACwCsADHNSALokAOALlCAMqcAnAJYA7AOYgAviQC0MKKHSRBAV2JlKIWq2kgZATgVooq9eUhUILSbrhGhAEx4yw9CB26QQIEpwCe7AQ8WHjoNkA
 The MOSFET I'm using, SSFU6511, is major overkill with a voltage rating of 650V and a current rating of 11A, but it only costs $2.28 in quantities of 10.
https://www.digikey.ca/en/products/detail/good-ark-semiconductor/SSFU6511/18667492?s=N4IgTCBcDaIMpwGIFUBsBWAjJkBdAvkA
It should be possible to find a cheaper MOSFET, but I'll be using the same MOSFET as the main switcher and also as the elliptical brake current switcher so buying 10 at once makes economic sense. SMD type MOSFETs with suitable ratings will be even cheaper, but those are hard to work with.

[PCB.Wiz]

I checked on DigiKey and found the lowest current available for a current regulator is 200uA and the maximum voltage for higher current devices is 190V. The current is unnecessarily high, the voltage is too low and the cost is fairly high.

Yes, take a look at ARK micro DMZ6001E.
That’s a SOT23 600v depletion mode mosfet, you set the current using a single resistor.

Addit: This small FET would replace R1 in your circuit, and can reduce your off-state from > 1mA at 470V to some few uA

[azeeman]

The problem is not setting the current, it's stopping the current when it's no longer needed. If the startup current is 50mA, then the dissipation will be 50mA * 250V = 12.5W. A depletion mode P type MOSFET also requires a voltage to turn it off. This may be an acceptable tradeoff as one way or another, the current will need to be turned off. A photovoltaic voltage source could be used, to drive the gate but these are somewhat expensive. The threshold voltage is also not as predictable as with a BJT which is not that critical, but may result in excess current that has to be shunted with a Zener or shunt regulator.
I also looked at using a normally closed solid state relay which would simplify the problem, but they are rare and expensive.
I prefer to use an N type enhancement mode MOSFETs because they are far more readily available, cheaper and can handle higher currents.
Also, this is one off project and by buying multiple of the same MOSFETs, I can use one for the startup circuit, another for the control power supply switching and a third for controlling the elliptical brake current. All of them will work off the same high voltage, but with different amounts of current.
A previous iteration used a relaxation oscillator where a capacitor was charged using a tiny amount of current and then the charged capacitor would discharge into the LED of a triac optocoupler. The optocoupler triac would then fire larger triacs to keep everything powered up. This worked well, but requires AC input to commute the triacs and is not a general purpose solution. The charge current was a few microamps so the dissipation was well under 1W.
Once the main power was working, the power triacs were controlled to set a fixed voltage for the brake and to power the control circuit. A low voltage power supply was still required for the control circuit and a commercially available product was used for that.
Using all switchmode power supplies is a more elegant solution, but high side isolated drivers are required and these need a DC to DC converter that will work over a voltage range of 10V to 300V. One power supply is required for the 100ma low voltage control circuit and another to control the brake current over a range of 0 to 2A over the full input voltage range.
A human can typically output 100W for short periods. If 5W is expended on control circuitry, this is a significant part of the power budget.
This project is being done as an intellectual challenge and it's interesting to see the ideas generated when I post about it. For that I thank you.

[PCB.Wiz]

The problem is not setting the current, it's stopping the current when it's no longer needed. If the startup current is 50mA, then the dissipation will be 50mA * 250V = 12.5W. A depletion mode P type MOSFET also requires a voltage to turn it off. This may be an acceptable tradeoff as one way or another, the current will need to be turned off.

I was maybe unclear, the SOT23 DMZ6001E only replaces your 470k feed resistor, it is too small to manage the full current.
It can greatly reduce your 'off' state current, which is quite high in your original circuit.

[azeeman]

Thanks, I looked up the part and it wasn't available from DigiKey, there might be something similar so I'll do some more research. This is a secondary problem, what I'm more interested in is single bit pulse density control of the switcher. The current concept requires a high side gate driver with an isolated power supply. That power supply requires significant current to start and run hence the need for such a high startup current and the proposed circuit.
I just realized that if a flyback transformer is used, low side switching can be used which reduces, but not eliminates the startup current, this opens the possibility for other solutions.
Single bit PDM control works by turning on the MOSFET and the turning it off when the inductor current reaches a maximum or the regulated voltage goes too high. The on time can be variable while the off time will be fixed by a high frequency clock. The logic is very simple and the control works surprisingly well. I breadboarded a simple version and was able to control the output voltage from 10mV to over 5V. If the output voltage is fixed, this is equivalent to an input voltage range of 1:100.
It's ideal for extremely wide range input because at low input voltages, the switch can stay on for many clock periods while at high input voltages, the switch will be turned off as soon as inductor current comes close to saturation. The available on time is fixed, but the switch can be prematurely turned off during the on time period. The off time remains unchanged. Because the clock frequency can be very high and the inductor current controlled, the required inductance can be very low compared to PWM current control.
The current sensing comparator has to be very fast. This is not a problem as op-amps with 27MHz gain-bandwidth products are cheap and low side current sensing is trivially easy to implement.
A comparator also senses the output voltage and if either the current or voltage is high, the switch will be turned off.
The brake uses an electromagnet which induces eddy currents in a flywheel causing braking action. Only the brake current needs to be controlled so only low side current sensing is required.

[PCB.Wiz]

Thanks, I looked up the part and it wasn't available from DigiKey, there might be something similar so I'll do some more research.

The BSS126 is broadly similar, 600V, SOT23, depletion mode, tho slightly higher Vth numbers, needing a slight increase in RSET.

[temperance]

The MOSFET requires a zener from gate to source if the input voltage can be higher than the maximum gate voltage rating of the MOSFET.

[azeeman]

If the MOSFET gate voltage goes up, it conducts more current into the base of the BJT which increases its collector current and pulls the gate voltage back down.
The major risk is voltage spikes which can drive the gate voltage high before the BJT can react. The large value of the gate bias resistor and input capacitance of the MOSFET will act as a low pass filter and attenuate the spike. For safety, a Zener would improve circuit reliability.

[Kleinstein]

Most modern MOSFETs include a zener from gate to source. With the rather large resistor this should be sufficient. Normally the BJT and rising source voltage would provide the protection and the zener would not get activated.

A depl. MOSFET as current source instead of the 470 K could allow the opto-coupler to get away with less current. Some of the depl. mode FETs give SPMS startup as there intended use case (however without the extra turn off). It depends on the use case if it's worth it.