4
\$\begingroup\$

I'm being vague about part of my project, but please ask me to fill in critical details I'm missing.

I need to power a 24V @ 20A device for 5 hours (one time), within a weight budget of 700 grams, at a distance of 500ft from the ground. A 500ft tower is impractical in this instance. Nobody will die if the device fails, and since it's a hobby project I am OK with some serious liberties in terms of normal safety devices to drive the weight down.

The highest voltage I can realistically attain is 310-340 volts DC (rectified & smoothed 220), which is then raised via a pair of 26awg cables 500ft in the air, and transformed back into 24V @ 20A. My initial plan is to hack apart a COTS power supply, and strip out the AC rectification side (doing that on the ground), remove as much cooling as I can (because I will have ample airflow), and other hacks -- but I wonder if another topology can be made simpler and still handle a 15:1 reduction ratio. I'm assuming a fair amount of weight is being taken up with the full transformer inside the power supply, even though it is a high-frequency design.

I can live with high ripple and poor regulation (0.5v P-P, 10% regulation), I don't need mains isolation. Am I better off chaining 3 buck regulators? Should I go for a custom-wound transformer and a flyback topology?

\$\endgroup\$
  • \$\begingroup\$ You may be able to remove the rectifier diodes, but I expect the bulk capacitor(s) will need to stay. The inductance of your 26 AWG supply wires will be on the order of a couple hundred µH so you'll need the capacitors to keep your input current at some semblance of a steady value. \$\endgroup\$ – scanny Sep 27 '16 at 6:47
  • \$\begingroup\$ If you don't need isolation primary-secondary, you can do it with a a 100-200 gram buck converter. \$\endgroup\$ – winny Sep 27 '16 at 6:50
  • \$\begingroup\$ What is the 24V@20A device? If it's something you can pull apart and rewind as a 240V 2A device, it'll run off your 320V supply directly (given 2A * 42 ohms wiring loss) eliminating the converter problem altogether. \$\endgroup\$ – Brian Drummond Sep 27 '16 at 12:07
  • \$\begingroup\$ 'ample airflow' '20A' '500ft' 'weight limit'. I bet you are trying to make a quadrocopter hover in one place for 5 hours. \$\endgroup\$ – silverscania Sep 27 '16 at 12:46
  • \$\begingroup\$ @silverscania: yes :) the "why" is what's vague. Brian, rewinding isn't an option in this case. \$\endgroup\$ – Bryan Boettcher Sep 27 '16 at 15:42
2
\$\begingroup\$

What about the MeanWell SDR-480P-24 ? Maybe gut the PCB and chuck the chassis... the PSU feels pretty light including the metal chassis. It's fanless and seems to have a pretty good efficiency - making me wonder if the secondary side is sync-rectified or conventional schottky-based.

If you insist on DIY, I would say the forward concept needs less mass in the inductor/transformer core than flyback, and the more modern FET generations are more efficient = need less heatsink surface (and mass). The recent TrenchFET/FinFET generations have gradually less and less gate capacity and RdsON. Hence the most recent PSU generations run gradually cooler and are lighter. Synchronous rectification on the secondary side adds to that effect, although at 24 V I'm wondering if the effect is still noticeable, compared to a Schottky diode. As for ripple voltage... you can only stretch the secondary elyts a little, not 200%. They have a problem with internal heating (due to RdsON) and relatively little heat-dissipating surface, no amount of forced airflow will quite help. Solid Polymer are available up to 25-35 V, they're light and have a huge permitted ripple current (at around 100 kHz optimum switching frequency). So I don't think ripple voltage will be an issue, if you stick to the elyts' permitted ripple current.

If you need 5 hours continuous runtime, thermal capacity (inertia) alone won't allow you to undercut heatsink surfaces/masses (compared e.g. to a UPS designed to last for 10 minutes). Plenty of forced airflow is a different story though :-)

\$\endgroup\$
2
\$\begingroup\$

A 1:15 reduction should be perfectly fine with a buck converter.

I would be more curious whether the fine 26 AWG wire can withstand the strain applied when hanging 150 m from a tower and you will have descent voltage drop since the wire will have about 42 Ohms. Therefore the converter ratio will increase which is better for the efficiency of the converter.

You might have some problems to get your hands on a decent choke for such high voltages but I know that they exist and you should definitely consider that rating.

\$\endgroup\$
  • \$\begingroup\$ Re: "decent choke for such high voltages" : Reminds me of a recent troubleshooting case (I do not want to be specific) where a buck converter, nominally specified to work at 24 or 48 V (wide range input) worked relatively okay at 24 V, but would fail miserably after a few minutes at 48 V. The symptom was, that the toroidal choke was scorching hot, and on a 'scope you could see that it behaved more like a resistor :-) The choke core material was probably simply inappropriate, but your line about a "good enough choke for the voltage" does ring a bell with me :-) \$\endgroup\$ – frr Sep 27 '16 at 6:47
  • \$\begingroup\$ @frr: So you think the core material was inappropriate for the switching frequency and therefore the ckokes Q factor dropped? \$\endgroup\$ – Simon Sep 27 '16 at 7:45
  • 1
    \$\begingroup\$ yes, essentially. Possibly a choke intended for EMI suppression (having a low Q) sneaked into a carton of buck SMPS chokes or something. The fact that the problem (shutdown) was more prominent at 48 V could be a side effect of some protective feedback in the PWM controller kicking in earlier, when heat from the choke was combined with a rather high voltage at the PWM's Vcc input. Or maybe the choke was even less efficient at higher dI/dt, not sure. It was not a case of saturation: there was no "broken triangle" in the primary current. There was no slope at all :-) just a rectangle. \$\endgroup\$ – frr Sep 27 '16 at 10:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.