Linear slow-start can be lossy. Your simulation doesn't show the dc-dc downstream load, so only mA are being supplied.
The DC-DC has to be running during slow-start, to be effective.
You need to calculate the joules, depending on total reflected DC load current plus rail capacitance dv/dt.
RL
You're supposed to be able to do anything with the FPGA (?) driver. Isn't that the whole idea behind direct digital control?
RL
We can't think of a way to start up the dc/dc inverters with digital control. The peak currents are just too high.
Smaller fets might work, at loss of efficiency. The power supply rampup should work.
Downstream of each dc/dc converter is a half-bridge PWM-programmable DC power supply. But it's off during startup. The startup energy dissipation is tiny, and the soft-start fet is under 20 mohms when things are operating.
Yes. We will turn on the square wave drive and start the DC voltage ramp simultaneously. The power supply to the inverter will ramp linearly from 0 to 48 in about 200 ms.
Did that of course. The only startup load is the isolated side caps.
fredag den 16. december 2022 kl. 17.02.10 UTC+1 skrev legg:
sure, the 50/50 and phaseshift just gets around the problem of the gate drivers swallowing short pulses, but it seems the board isn't wired to be able to do that
I don't think there is any way to use digital control to avoid gigantic peak currents at startup. And certainly not worth launching a project to try.
torsdag den 15. december 2022 kl. 17.40.46 UTC+1 skrev John Larkin:
but avoids the issue if the gatedrivers swalloving short pules
it think if you disconnect the BINPs on both drivers and a hardwire them high, you could
On a sunny day (Fri, 16 Dec 2022 07:54:44 -0800) it happened John Larkin snipped-for-privacy@highlandSNIPMEtechnology.com wrote in snipped-for-privacy@4ax.com:
Resistor?
Most things I do these days are hobby related on veroboard stuff... Even giggle Hertz stuff..
You'll need a ball park capacitive load anyhoo.
I=C dv/dt.
Different dv/dts set the fet current, or vice-versa, depending on what's being controlled during the inrush limit event.
All I noticed in the sim was ~20mA of fet current.
er . . .don't swot the fet gate at turn-off.
RL
Love the 48V+ gate enhancement.
I want one of those two terminal current generators.
RL
Try this:
Another recent discovery is that we need schottky diodes in parallel with the mosfet substate diodes in the h-bridges. Weird things were happening during the anti-shoot-through times of the LTC4444 driver chips. I've seen mosfet substrate diodes decide to act like high-power step-recovery diodes.
Dremel!
It's a TLP191 pv optocoupler, essentially a floating 7 volt 25 uA power supply.
On a sunny day (Sat, 17 Dec 2022 09:32:51 -0800) it happened John Larkin snipped-for-privacy@highlandSNIPMEtechnology.com wrote in snipped-for-privacy@4ax.com:
OK, that will work, short dissipation in the MOSFET during the ramp. When switching in a resistro during filter capacitor charge would work too. I understood from your posting that at startup there is no other load?
Yes, nice, and gold!
Interesting
With a high-z gate circuit, there may be issues when the 48V rail is first applied.
A conventional fet inrush limiter trys to use Cgd as a dv/dt limiter. In your case this would require a pmos element - the conventional circuitry sticks inrush limiting in the negative rail per telecom practise.
RL
RL
200 uF of capacitance and maybe 100 mA of stuff on the load side, power for some isolators and a bit of logic.
Yes, we could charge up the loads through a power resistor that is shorted by a fet after a second or so. That would be simpler than my cicuit. I'd still need an over-48-volt supply if I use an n-channel fet, or buy a really low Ron p-fet.
That style of construction confuses me. Dead bug confuses me too. I prefer
I'm planning a product line using the Pi Pico as the compute core. There is an amazing culture around the Pi, a zillion maker-type kids and school courses. It's sometimed dismissed as a toy but serious people are making products around Pi's. Every other uP that we have used has gone EOL anyhow (except, surprisingly, the 68332.)
Looks like a few clicks and a zip file upload will get us cheap boards in 5 days. I'll try it.
It Spices nicely. The 48V comes from a kilowatt of MeanWell switcher and it doesn't jump up in nanoseconds. There is a tiny output jump when it's enabled, which turns out to be the body diode of the depletion fet, namely the FPGA driving the load!
I'm stuck doing it on the high side, and have that 17 mohm n-fet in stock.
On a sunny day (Sun, 18 Dec 2022 08:11:19 -0800) it happened John Larkin snipped-for-privacy@highlandSNIPMEtechnology.com wrote in snipped-for-privacy@4ax.com:
Its nice, combining all those small boards! In my case many of the chip adaptors have .1 inch spacing pins so they fit nicely in the 'veroboard' holes. Wires and SMDs directly soldered between the round isles, even the edge connectors fit.
Bit wasteful with gold on that high power resistor board? Or is that not gold? The nice thing of those boards with holes I use is that no dremeling in needed. The dremel makes a lot of mess. Long time stability is good (but I use 60/40 solder), if I use the eurocard format (100 x 160 mm) standard boards then those fit exacty in these casings:
I need to fix a break in some flex PCB cable in a old eeePC I have.. Any idea of some good conductive glue ?
I have never used the Pico, I did see your post in the Raspy group for coffee and tea :-0 Yes, much is unobtainable now, seems industrial use of Raspberries is high and gets priority.
I have 5 Raspis in use now, one Pi4 8 GB just used for web browsing and downloads. A Pi4 4 GB does security recording and all sorts of measurements (radiation, weather, GPS, AIS, planes) 24/7. The advantage is the HDMI out and the GPIO connector. A 4 TB USB harddisk is connected to each PI 4. Have ported most stuff I wrote to the Pi now.
I may try a Pico one day, but most things can be done with a simple PIC at micro amp use if needed.
Let us know how it pans out!
There is no shortage of Pico boards. RasPi have enough wafers in stock for 20 million chips and more are in the pipeline. There are hundreds of thousands of RP2040 chips at the usual distributors at the moment.
Of course you can also make an over 48v supply for cheap low rdson nfets using your pv opto device or by capacitive charge pumping up from a logic level square wave.
piglet
It's ENIG on FR4. Microinches of gold. Every couple of years I have one of our PCB vendors make me a few square feet of gold-plated FR4.
It looks good, solders great, and doesn't tarnish like bare copper.
It's an art form. Fun.
Yes!
Last time I checked, Digikey had 19,000 Picos in stock at $4 each.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.