1
\$\begingroup\$

I'm looking to make a linear (LM317 based?) high current / voltage regulator (30 amps at 0 - ~35 volts).

I looked at some desings online but still have some problems with it.

Maybe you can guide me in the right direction? Will a design like this make sense?

Q:

First, why is the output not regulated while input changes? (Input is AC 10 V with a DC offset of 20 V.)

Second, why the R1? a why such a low value, almost 650 mA goes through it! that will fry it and also waste a lot of power.

Last any tips? How many transistors up top (NPNs)? Component Models? (SMD ONLY!) Does this design event make sense?

Also it will be powered by a AC 230 V, 50 Hz to lower AC voltage toroidal transformer, then rectified to DC, what would be a good output voltage to choose on the transformer secondary side? (Vrms).

. LTSpice screen

Design taken from LM317 datasheet:

screen from lm317 datasheet . Thanks!

\$\endgroup\$
4
  • 2
    \$\begingroup\$ That was the way we did things back in the 70's. Buck converters are cheap and easy these days. Do you really want a large iron transformer and massive heatsink? \$\endgroup\$
    – Kartman
    Sep 26, 2021 at 7:34
  • 3
    \$\begingroup\$ When drawing 30 A at voltages close to zero your pass transistors have to dissipate 35 V * 30 A = 1 kW in total, which isn’t trivial to design for. Why do you explicitly need a linear design? \$\endgroup\$
    – jms
    Sep 26, 2021 at 8:04
  • \$\begingroup\$ If you want absolutely a transformer (for such currents), use a rectifier and then head inductance (to limit peak current) instead of capacitance (or with, after) ... \$\endgroup\$
    – Antonio51
    Sep 26, 2021 at 10:46
  • \$\begingroup\$ @Poul Freeman - 2 years on: We are happy with this post. It represents a typical learning curve approach. You have learned much since then. Others can learn much about why and why not to take this approach from reading the Q&A. I just upvoted it. This is the sort of Q& that people come here for. Well asked. Looking for good aanwers. Gets good answers. Learns. Splendid !!! :-) \$\endgroup\$
    – Russell McMahon
    Oct 4 at 2:02

3 Answers 3

1
\$\begingroup\$

You already have a good answer but I want to add a few tips if you are really sold on a linear regulator. Related to that, one of the reasons I am responding to this is that I have used both linear and switching power supplies and a lot of the switching supplies seem to have quirks making them harder to use when compared to linear, so maybe you have a similar reason.

In general, if you want a variable 30A linear supply realize that you can't go at it "brute force", meaning, use one transformer with a fixed output voltage and just dissipate the power on a heat sink. As the other answer has pointed out, that is just unreasonable as well as impractical. I understand it is tempting to look at the spec sheet and based on the principle of "add external transistors and grow current to as high as you want" you decide that 30A is what you need and go for it. Unfortunately, there are limits to this and for such high current, you will have to add some "intelligence" to your design.

So specifically to your question about general design tips, here are a few:

  1. You will need a bigger capacitor - 10uF is grossly inadequate. That applies no matter what else.
  2. Use a transformer with multiple output taps and relays to switch the input voltage based on the required output voltage. If you keep the RMS output of the transformer no more than 10V above the required output voltage, you are dissipating 300W (plus more due to other losses etc.). Still significant, but manageable.
  3. If you don't want to use a multiple tap transformer, use multiple switching regulators and use relays (or MOSFETS) to select which one feeds your linear regulator. You can buy 90%+ efficient regulators that will receive up to 72V input and produce 36/24/12V at more than 30A. Depending on the dropout voltage of the pass transistor, you can limit dissipation that way.
  4. Variation of (3), but use a variable output switching regulator in front of the linear regulator. I haven't done the math, but if you use a double potentiometer and a few external resistors, you may be able to setup your voltage dividers so that the switching regulator always outputs a voltage that is 2 to 3V above what you want your linear regulator to produce. Dissipation in that case would be minimized to losses of the switching plus the small drop on the linear regulator. Even with a 60V RMS input and 1V DC output at the linear regulator (which is the worst case), with a 90% efficient switching regulator you may be able to keep dissipation below 300W.

Hope that helps.

\$\endgroup\$
1
  • \$\begingroup\$ Thanks, great idea with that switch mode before linear! I think I will use different taps on the transformer and a switch mode before the linear with NPN. \$\endgroup\$
    – user296633
    Sep 26, 2021 at 10:19
4
\$\begingroup\$

Im looking to make a Linear (LM317 based?) high current / voltage regulator ( 30 Amps at 0 - ~35 Volts).

No offense but, if you are to go for a linear regulator for this particular requirement you should have very good reasons.

Also it will be powered by a AC 230v 50hz to lower ac voltage toroidal transformer, then rectified to DC, what would be a good output voltage to choose on the transformer secondary side? (Vrms).

I'd like to talk about this first. Can you imagine how a large transformer you need? If you can't, let's make a guess: It's obvious that you should go for at least two Darlington pairs for 30A (I used single 2955-3055 pair for my old 30V/5A design, for thermal reasons of course. Why not at least 2 pairs for 30A?). This means that you can expect tens of volts to drop across the regulator at 35V/30A. So your transformer could have a rating of 230V-to-60V/60A i.e. 3.5kVA minimum. It'd not be funny to expect such a transformer to have dimensions of around 25 cm x 25 cm x 25 cm, and a weight of at least 35 kilograms.

First, why is the output not regulated while input changes? (Input is AC 10 V with a DC offset of 20 V.)

10uF capacitor across the input is not enough to remove such large, low-frequency ripples. And you can't expect the linear regulator to remove these ripples at 3.1V/0.1Ohms = 30 Amps.

Second, why the R1? a why such a low value, almost 650 mA goes through it! that will fry it and also waste a lot of power.

I think you should care more about the regulator's waste of power and that of the transformer that I talked about above. For example, assuming 60VDC coming from the step-down rectifier + filter, the waste of the regulator will be (60V-35V) x 30A = 750W at 35V/30A. Even worse, it'll be 1.8kW at 1V/30A output. This level of waste is close to the heat power of this. We can't neglect the excessive heat of the transformer at these conditions.

\$\endgroup\$
1
  • \$\begingroup\$ I was thinking about that power dissipated on the transistors too, the transformer i can get even bigger if i want. I think i will go for a different design tho. Thanks \$\endgroup\$
    – user296633
    Sep 26, 2021 at 9:08
2
\$\begingroup\$

How many transistors up top (NPNs)? Component Models? (SMD ONLY!) Does this design event make sense?

No it doesn't make any sense to have 1kW dissipation and ask for SMDs.

In the typical linear benchtop power supply (I have 30V 5A ones) there is a transformer with multiple output voltage taps, with a bunch of relays to switch them, so the pass transistor only has to drop a few volts, which reduces its dissipation and makes it more friendly to SOA.

Basically if you want a fully linear variable 30A power supply get a bunch of cheap benchtop linear supplies like Korad KD3005P and wire them in parallel. It will be cheaper and with better protection. Note the schematic in the question has no protection. In case of output short it will just shoot all the power transistors into orbit.

First, why is the output not regulated while input changes? (Input is AC 10 V with a DC offset of 20 V.)

Because Q2 runs out of hFe at high current. A real transistor would also run out of SOA and burn.

Second, why the R1? a why such a low value, almost 650 mA goes through it! that will fry it and also waste a lot of power.

That's normal, you don't want a high value resistor because that would increase open loop gain too much and make it unstable. Even with this low value, stability of this circuit is very dubious.

Last any tips?

Yeah, for this current and voltage, it's the same as a motherboard CPU VRM, so same solution, use a synchronous multiphase buck DC-DC.

The whole thing will be cheaper than just the heat sink for the linear regulator.

Because basically, if you use a linear regulator, it's because you need very low noise for a sensitive low current circuit, and don't want to bother with filtering out the noise from the buck. Or it's a bench power supply: I bought a pair of linear ones because when I see some funny noise stuff on the scope or in measurements, I don't want to have to investigate to find out if it comes from the switching supply or not.

But what kind of load draws 30 amps and requires very low noise?

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.