Adjustable current limit for pulsed diode driver

[balage]

Hi Fellowes,

I have built basically the same circuit as this one, what a low side FET that switches the laser diode load at high frequencies: https://www.ti.com/tool/TIDA-01573#1

This can be used as a laser diode driver to make high freq short laser pulses. But my problem is a basic for a variety of pulsed driving applications: I want to limit the current pulses with the ability of adjusting digitally.

I have found this doc about the topic: https://www.microsemi.com/document-portal/doc_download/129605-sg1526-digital-current-limiting-techniques-for-switching-power-supplies
But I dont think this method is fast enough for 100-200kHz or above frequencies.

How would you do this limit?

[ajb]

What pulse lengths/rise times is your implementation doing or are you aiming to do?

That TI app note shows pulses on the order of 2ns.  That is not a trivial thing to design in any respect because of the di/dt involved, especially with a 60A pulse!

It's not the pulse repetition rate that determines the bandwidth required for active current limiting, it's the required slew rates.  There's basically no way you're going to be able to actively regulate a current pulse with 1ns rise times, at least not trivially, because the response would need to be equivalent to some GHz.  Your best option is to set pulse energy and peak power is to control the amount of energy available to supply the pulse, which you can do by trimming the supply voltage.  This will affect rise somewhat times as well, which may or may not be much of an issue for you.  Hypothetically you could also switch bypass capacitors in and out to adjust the energy storage, but it's practically impossible to do that without increasing the loop impedance to unacceptable levels.

[langwadt]

https://groups.google.com/g/sci.electronics.design/c/_M5jfScQ7CE/m/6WEy4aoPAwAJ

[balage]

Ah, sorry: the current is 3 amps maximum and the freq is 200kHz max with pulse widths down to lets say 50ns.
What I see is that the limiting element must be something that dissipates power, and there is no way to get curent limiting and dissipation in the same time.

[T3sl4co1l]

You can always make a conventional current mirror / source circuit, and just tune it to run fast.  Transistor capacitances are the enemy to settling in that time frame, and there is certainly no time to servo the current with an error amp, it needs to be right the first time.  Your input will be a square wave of controlled amplitude and duration (perhaps set by VREF --> DAC --> analog switch between DACOUT / GND --> current source).

Duty cycle will be limited by power dissipation, which is limited by transistor size and cooling.

Consider using RF transistors: they have much lower capacitances than power transistors.  Downside, you get far less Rds(on) for your buck.  But as cost hasn't been mentioned yet, I'm going to assume a solution is valid at any cost.

Tim

[David Hess]

200 kHz is roughly a 2.5 microsecond pulse width and faster, so assume a settling time of 1/10th that or 250 nanoseconds.  That is feasible with current mirrors like T3sl4co1l suggests.  A closed loop solution using a video amplifier could probably do it but attention to amplifier saturation when off would be needed.  A current switching design using transistors or diodes could absolutely do it and much faster but requires the current source to operate continuously which may be objectionable.

[Marco]

Current steering using those EPC GaN MOSFETs.

So you have one MOSFET as a current source with an opamp, with two MOSFETs to steer the current either to some low voltage supply rail (to keep static power consumption manageable) or to the laser diode. The supply rail the laser diode connects to determines the max compliance voltage.

[balage]

Consider using RF transistors: they have much lower capacitances than power transistors.  Downside, you get far less Rds(on) for your buck.  But as cost hasn't been mentioned yet, I'm going to assume a solution is valid at any cost.

That is why I have seen RF FETs in some high freq LD drivers.

Lets say it must be more professional than cheap. 

The driving signal is from an MCU.

200 kHz is roughly a 2.5 microsecond pulse width and faster, so assume a settling time of 1/10th that or 250 nanoseconds.  That is feasible with current mirrors like T3sl4co1l suggests.  A closed loop solution using a video amplifier could probably do it but attention to amplifier saturation when off would be needed.  A current switching design using transistors or diodes could absolutely do it and much faster but requires the current source to operate continuously which may be objectionable.

I don't know yet if a closed loop control is needed. It's complicated but worths a try.

Current steering using those EPC GaN MOSFETs.

So you have one MOSFET as a current source with an opamp, with two MOSFETs to steer the current either to some low voltage supply rail (to keep static power consumption manageable) or to the laser diode. The supply rail the laser diode connects to determines the max compliance voltage.

Do you mean that the current source is running constantly (so it can be closed loop controlled easily) and this current can be switched to the diode or to the ground?

I was wondering if a high current output opamp could be used direcctly somehow. https://mouser.com/ProductDetail/Apex-Microtechnology/PA165PQ?qs=TiOZkKH1s2TxoDJsZ7B3ng%3D%3D

[T3sl4co1l]

Do you mean that the current source is running constantly (so it can be closed loop controlled easily) and this current can be switched to the diode or to the ground?

Yes.  A current source on the bottom, then a diff pair on top.  The current is always running, so it can be set very accurately, e.g. with a slow servo loop (op-amp) -- and its bandwidth can be improved by adding resistance and inductance between the CCS and diff pair.  The diff pair should be driven differentially, so that the "tail" voltage doesn't vary a lot -- helping out the CCS that much more, but even if not, the voltage swing (compliance range) only needs to be around Vgs(on), rather than whatever the load voltage is.

Also, efficiency isn't completely terrible -- the "dump" side can be a low voltage source, while the load can be arbitrarily high.  Of course, you do need multiple sources in that case, but sometimes those are handy.

BTW, note that extra compliance range is needed for an inductive load, which, we're talking fairly low impedances here (3A and, under 150V maybe?), so, expect to draw transient voltages to get the current flowing.  Depends what your expected loads are, but something to think about.

Tim

[balage]

I am not sure if I understand current steering well. So basically it should be the schematic here:
The A and B signals are applied inverted. R1 is the dummy path of current and D1 is the laser diode path.

I am not sure this could be a final solution because of the dissipation. Or how do you mean to steer the current to a low voltage rail? The supply of R1 should be quite a low voltage, right? But in this case the current source should work on the same frequency as the diode switch.

Current steering using those EPC GaN MOSFETs.

So you have one MOSFET as a current source with an opamp, with two MOSFETs to steer the current either to some low voltage supply rail (to keep static power consumption manageable) or to the laser diode. The supply rail the laser diode connects to determines the max compliance voltage.

Also to keep the current source MOSFET's dissipation low the supply voltage should be low, but in this case the FET would work in the very low D-S resistance region.

[T3sl4co1l]

You've got C and E swapped, it has to use NPN / Nch.  And then R1 can be arbitrarily low, and Q3's supply voltage can be higher as needed for compliance range.

Tim

[balage]

Laser diodes are current driven and current sensitive semiconductors. A change in drive current equals a change in the devices’ wavelength and output power. Any instability in the drive current (noise, drift, induced transients), will affect the laser diode’s performance characteristics. Specifically, they will affect the output power and wavelength. Furthermore, the temperature of the diode junction is directly affected by current. The current instability of the source will cause junction temperature swings; the output characteristics (again power and wavelength) will change.

This stability thing is actually a reason to make a precise driver even if it is dissipating more.

You've got C and E swapped, it has to use NPN / Nch.  And then R1 can be arbitrarily low, and Q3's supply voltage can be higher as needed for compliance range.

Tim

But if the two path's supply is different then the current source opamp has to regulate every time when the paths are changed. Am I right? I mean it will not has a stable condition.

[Marco]

Current steering, even with those GaN MOSFETs will probably ring quite a bit.

I've been trying to think up a closed loop fast voltage to current circuit with a SiGe NPN, a BUK4D60-30 MOSFET and a 1 Ohm shunt connected to the positive supply rail (with the load in between the shunt and the MOSFET) but I can't really think up something elegant.

[T3sl4co1l]

This stability thing is actually a reason to make a precise driver even if it is dissipating more.

It's your regular accuracy tradeoff, you can have speed or efficiency or whatever, but some combinations are easier than others.

Example, you could maaaaybe just pull this off with high efficiency, using a GaN inverter and resonant topology, but it'll be a lot of work to achieve, and still probably not all that stable (output ripple filtering directly affects step response, so the settling and stability are at odds here).

But if the two path's supply is different then the current source opamp has to regulate every time when the paths are changed. Am I right? I mean it will not has a stable condition.

CCS only sees the differential pair's tail voltage, which due to balanced drive, is nearly constant.  What's left is due to Early effect / channel length modulation, a small effect for most types.

Current steering, even with those GaN MOSFETs will probably ring quite a bit.

I've been trying to think up a closed loop fast voltage to current circuit with a SiGe NPN, a BUK4D60-30 MOSFET and a 1 Ohm shunt connected to the positive supply rail (with the load in between the shunt and the MOSFET) but I can't really think up something elegant.

Yeah, layout will need to be tight, and expect to use series gate resistors (or ferrite beads), and probably source degeneration too, to keep them from singing.

SiGe into VDMOS, odd bedfellows...  But I suppose that's because, you can't really get anything else that's fast and moderately powerful?

LDMOS is probably worthwhile, with the downside that the packaging may be a bit awkward? -- you've got signal flowing through what's normally the ground return path, and also heatsinking.  (At least, most types are common source..)

Not that power types, with the heatsinking on the output node (drain), are any better in that regard, heheh.

Tim

[Marco]

SiGe into VDMOS, odd bedfellows

Well, you'll need some extra amplification and a fast NPN is cheaper than some exotic opamp. The MOSFET same thing, cheapest device which gets in the neighbourhood, though honestly something like EPC8004 or ST/NXP LDMOS isn't exactly expensive.

With NPN+PMOS a voltage to current transformer which uses a feedback shunt on the same side of the MOSFET as the load is a little more straightforward.