Trying JLCPCB for 200A 6kW LED pulser

Having finished a proto PCB output-stage for our 200A 6kW LED blaster, and with the project's power engineer, Rob Legg, off for a Christmas break, I was eager to get PCBs made. I reasoned a Chinese PCB house would work through Christmas. Inspired by J.L., I tried ALLPCB. Smooth website, but a SUBMIT button freeze-up.

JLCPCB also proclaimed fast production and fast DHL delivery, in under a week. It was a little disconcerting to get a Dear WinnieldHill order confirmation email, but hey, there's a new spelling for my name.

Haha, ordered 15 boards, total cost with shipping, only $31, that's $2 per board. It's an incomplete prototype, but still useful for playing around at full power. R.Legg burned out a few $150 LEDs, with long pulses, at my request, but now has other parts on hand for the next round. Anyone else wants to join the fun, I can send you blank RIS-796_1 prototype PCBs.

Now my order for 15 pieces from ALLPCB, has gone through, so I'll have 30 blank boards for test. 200-amp pulses anyone?

What's the duty cycle?

--

"When constituencies are small their elected representatives must concern themselves with the local interests of their constituents. When political representatives are distant and faceless, on the other hand, and represent vast numbers of unknown constituents, they represent not their constituents, but special interest groups whose lobbyists are numerous and ever present. Typically in Europe a technocrat is an ex-politician or a civil servant. He is unelected, virtually impossible to dislodge during his term of employment and has been granted extensive executive and even legislative power without popular mandate and without being directly answerable to the people whose interests he falsely purports to represent."

- Sir James Goldsmith (Member of the European Parliament) 1933 - 1997

On 23 Dec 2019 00:14:37 -0800, Winfield Hill

Nice...

but Austria is way to expensive for shipping...

Anyway - I expect some interessting measurements, the thermals in the high current path... I would have done this.. (doing stuff with 300 -

ear/eye-protection during power up will be mandantory

good luck and have fun!

- Michael Wieser

We'll pay shipping. It's just an envelope.

Whatever you want. But this test board has only a small heatsink for the TO-247 MOSFET. Ohmite WA-T247-101E, has thermal res 10C/W, with modest air blowing on it from the back. With 30V or more ( V = L di/dt ), for fast switching, can mean up to 6kW dissipation. So 0.2% duty means there'll be 12W of loss.

Our PID application will be 0.1% at 30Hz.

:-)

Trying to make the schematic look better, I made a silly editing error, forcing a modication and re-order the PCBs. The files have also been updated on DropBox.

I'm designing a fast o/e converter, roughly 1 GHz, as a sort of experiment, but if it works well we may as well call it a product. But how will I test it? If I pulse a laser, it may do all sorts of weird things, like gain switching. I guess I could check the laser pulse with some other o/e converter, maybe an unamplified one.

I'm thinking that I could use a 20 GBit SFP module and our SRS clock generator to make a variable-frequency optical square wave generator. The telecom TOSAs don't run the laser to extinction, so we wouldn't get any gain-switch spikes.

An excellent opportunity to buy a telecom modulator and run it from one of your pulse generator products. Telecom modulators are nice 50-ohm travelling wave things with a V-pi of 15V or less and some have flat responses out past 30 GHz. I used to have an Avanex SD-40, which cost $6k around 2006. Beautiful.

Cheers

Phil Hobbs

In x-Chapters 3x.19, the section put up for this thread, Paul manages to get 2ns pulsed light response out of a blue LED. It's hard to get LEDs to go much faster. But it's easy to get laser diodes to go fast. My fastest network analyzer tops off at 3GHz. Years ago I took an ordinary cheap laser from an old DVD player, and drove it with a bias-T, maybe 45mA DC and RF up to 3GHz from the analyzer. Then a small photo diode, similarly-biased, was blasted by the laser, into a bias-T to the network analyzer's response input. This combination was flat to past 2GHz, probably the PD was falling off by 3GHz. The laser diode didn't do any weird things.

A few years ago I spent a few months upgrading some of my RIS-617 low-noise PD preamp circuits, adapting them to fibre-optic detectors offered by QPhotonics. These go to 3GHz, at a cost of under $200, or 12GHz for $1.5k. That sent me scrambling for similarly-fast fibre-optic laser diodes, eBay-cheap IR communication diodes, with butterfly-connector blocks, and fibre-optic attenuators. I restricted my tests to modulation, rather than full on/off. The laser-diodes were fast and well-behaved.

One product that Jonathan is working on is a lithium niobate EOM driver. Some people buy our little picosecond pulse generators and add someone's expensive amplifier, so we figured we'd put it all in one box.

So I guess we should buy some modulators to test it realistically.

The commercial driver amps are intended for AC-coupled telecom use, not optimized for DC-coupled narrow pulses.

We have a couple of New Focus detector boxes, two different wavelength ranges. Unamplified, but super fast and wide wavelength range.

Have you played with gain switching? Lots of diode lasers do that. Starting at zero bias, apply a pretty fast current pulse. The laser sleeps for a while, wakes up, and makes a gigantic pulse that is typically 10s of ps wide. That makes an ugly leading edge on a rectangular pulse. But if you tease the drive and the pulse width, you can harvest that first spike and supress the tail.

You have to keep it short or you'll blow the facet off the diode.

Cheers

Phil Hobbs

It sounds like a case similar to our experience applying 200A pulses with theatre LEDs. You have to blow them up, to find out what you can't do.

Yup. The issue is that the last couple of atomic layers by the facet have much less restoring force on their electrons (due to the missing nearest neighbours). This effectively lowers the local band gap, making them absorb quite strongly inside the laser line.

If you dump enough heat into the facet, the shear stress due to thermal expansion just blows the facet right off the die.

The rest of the die is generally reasonably OK, but the regeneration is much reduced, and of course the optical quality of the facet is ruined.

Cheers

Phil Hobbs

We've only done that in the roughly 1 amp and below range, and the diodes seem OK. We kill the drive, or even back-bias the diode, after that first fast spike comes out.

Fun. I may have to try that.

Cheers

Phil Hobbs