A beefy electronic load (#P4F1)
Well, that was a long week…actually, it was a fortnight
(the alignment of images really sucks when placed near the header of the post – thanks WordPress!)
So this puppy was manufactured by my new friends over at OSH Park (ordinary order, no extras) and I think I hit the express jackpot. I registered on June 8th and submitted my files – three hours later (1 AM my time on Friday) I already got notice of the panel assignment. Maybe the small size of 54mm x 31mm (2.1″ by 1.2″ for the Yankees out there) just made a perfect fit on the remaining board size and they sent it straight to production As OSH Park charges 5 dollars per square inch of board area (for three units!), these were 12.75 USD including the free standard shipping. The boards were complete on Tueday 13th and they took another three days to Germany, including the sometimes lengthy stop at customs. Now that’s fast, I can tell ya
What does this board contain? Well, as I already rambled in the previous post, this is a dual-MOSFET electronic load based on the quite simple prototypes from Dave Jones and Martin Lorton. It uses the LP324 low-power variant of the good old LM324, as I figured this is one of the slowest op-amps on the market. You know, I want a stable load that doesn’t get all excited about the weird and wonderful regulation spikes from cheap power supplies – and so I tried using the fourth op-amp as an intermediary filtering stage before feeding the two output op-amps. I didn’t really test that beforehand, but I did some tests regarding values of the filter caps and resistors in the RC networks.
I have to say that it looks a bit jagged, have a look at the separated 15V and GND symbols and the missing pixels on the wire connections of quite a lot of the vertical parts. I’m new to Kicad, and this is my second board ever made. Maybe there’s some flaw with the ongoing nightly builds that need to be used if you want to view other projects that have modern rubbish like controlled impedance traces…?
Also, I do not know to this day how to properly tell Kicad that some traces ought to end at a connector that will run wires to some more external circuitry. I mean, without deleting those external components from the schematic itself and therefore losing it as an reference sheet when building the units. I did add the connectors in parallel and just removed the additional parts (like the MOSFETs and the precision pots) before sending the file to the PCB manufacturer. There has to be a more elegant way of doing this – can anyone help me out?
First of all, I wanted a higher voltage for driving the MOSFETs closer to their minimal RDSon. Martin has used an LM7806 to hook up the entire thing to 8V or more (he recommends 9 to 12V to be in the regulating regime, but not burning too much power in the regulator). I went for an LM7812, which I’d say should be supplied by 15V DC. The LP324 is fine with that – the TI datasheet says 3V to 32V, so we’re good in both cases. This allows higher gate voltages, which may come in handy for high currents at very low voltages, say close to depletion of NiMH or alkali cells.
Next, I set two 15k resistors in line after the ten-turn precision voltage adjustment, which also gets buffered by 33n and 10n against GND and the op-amp output – you don’t want hard feedback in that unity gain op-amp, but a smidgen of carefully placed capacitance should keep out the worst high frequency crap from the outside world. In theory, this should act as a low-pass in the order of the 1 kHz mark.
I did the same thing again after the limiting trim pot – but that turned out to be a disaster. I’m not sure if both of the input offset voltages (2mV “typical”) from these two stages add up to an unbearable amount, or if there’s something wrong with two low-order filter stages in general – but running this configuration gave me a minimal current draw of almost 3 amps. Three amps…good lord. I think Martin had something in the low 10mA range as minimum current – could be 3 or 30, I can’t quite remember. But 3A is unacceptable…which is why I ripped off some of the IC legs and skipped this second op-amp. Other than that, this low-pass filter works nicely on the first op-amp alone.
Now the rest of the circuit is again packed with filtering, but it resembles the general idea of Martin’s electronic load quite well. There’s a 15 ohm gate resistor, another 100k for feedback between gate and source (which runs to V-), and also 1k from source to GND. I’ve used the IRFP250 MOSFET as it is readily available and cheap. The 0.1R load resistor in the schematic was used for testing, but it is, for some bizarre reason, currently more than double the price of the 0.22R (or larger) at Reichelt. Both are much smaller than the 1R used by Martin, but remember: This resistor not only carries some of the thermal load (and thus, limits total current when too large in resistance), but it is also used for feedback. Each of the MOSFETs has its own source resistor and a separate op-amp dedicated to it. As long as the op-amp can pick up a reasonable voltage between MOSFET source and GND for regulating the gate voltage, we’re good. So for the final unit, I went with the 0.22R, which is small enough, especially when run in the dual MOSFET configuration that effectively halves the resistance.
One small peek onto the MOSFET assembly on a Fischer SK02-100-SA heat sink (100x115x63mm³, 1.3K/W), together with the 50W 0.22R power resistor in the back:
So, does it work?
It doesn’t, and you 7800 fanboys may already know why: For some reason, the Kicad schematic has swapped pins of the 7812 regulator. Which means that Vin is on the correct pin, but of course the regulator totally freaks out when driving Vout (the left leg in rendered image on the top) to the GND plane and having GND (the body contact) on the regulated 12V rail. Hell yeah, there was no magic smoke, but that’s only because my fixed voltage power supply doesn’t offer 15V and I went to using the current-limited lab power supply instead
Don’t get me wrong, this is entirely fixable for testing purposes, but it doesn’t look good in a “production” unit. Photo of my fully working board is attached (don’t judge me on that amount of flux residue on the LP324 footprint after the smallish voltage incident and subsequent silicon replacement!)
For that reason, I will give away the unpopulated third board free of charge, maybe even including shipping, depending on your location. Drop me a message if you want it.
Here’s more of the flux massacre
These problems led me to creating a variant of the board that uses the now removed filtering op-amp as a third output op-amp for another MOSFET. I would absolutely not recommend ordering this board again without fixing the described issues, but if you want to have a look at the Kicad files to spin your own – here you go:
The .csv holds kinda-sorta a BOM for the board… (bloody tabs not working in a code environment – what’s it good for, then?!)
amount type Reichelt_order_no
1 33nF MKP10-400 33N2
2 0.1R 25W METALL 0,1
1 10k 64Y-10K
1 LM324 LP 324 M
1 100µF MKP4-630 1,0µ
1 1µF MKP4-630 1,0µ
2 1k RND 1206 1 1,0K
2 100k RND 1206 1 100K
2 15R RND 1206 1 15
4 15k RND 1206 1 15K
1 100nF X7R 1206 BF 100N
2 470pF X7R 1206 CF 470P
2 10nF X7R-G1206 10N
2 33nF X7R-G1206 33N
I do have more information on the performance of this thing to share, and also news on the triple MOSFET board. But that’s for another follow-up post. See ya!