Are there solid state relays with variable slew rates?

[vini_i]

My google foo is failing me. I'm looking for a solid-state relay that will let me gently apply and remove a load. Absolutely worst-case scenario, one second to come on and one second to turn off.

The background is as follows. We are testing a specialty power supply (30VDC 100ADC max). The supply will not tolerate rapid changes in current. To do the test we use an insanely expensive DC load that lets us ramp the load. Theoretically, this can be replaced with some fixed resistors and a solid-state relay that can ramp the load on and off. We could develop something but much rather prefer an off-the-shelf solution.

Any thoughts?

[T3sl4co1l]

Uhh...

N...well, yes, I guess, definitely?

You have precisely such a relay -- just connect your insanely expensive DC load in series with the desired load, program it to ramp from 30 to 0V in 1s, and there you have it!

As a lone component, you are expecting something to be able to dissipate up to three kilowatts for a whole second.  A quick perusal of available devices says no.  Very few semiconductor devices handle that much power at all; among them, none do it continuously, or down to an arbitrarily low "on" voltage (e.g., industrial IGBT modules).  Such a unit must use multiple devices in parallel, with a control circuit to stabilize and equalize the power between them.  (A typical e.g. TO-247 sized power transistor will be able to handle maybe 100-300W each over this time frame, so expect to need 10 or more.  Or there are other configurations that can use fewer transistors and a cheaper power dissipator (like a dumb resistor), but at expense to e.g. speed or emitted noise.)

Perhaps it doesn't need to be as precise, or fully-featured, as your load is.  Perhaps a cheaper model, of similar overall capability, would suffice?  I guess I don't get it; you're testing "a" power supply, and you already have the load -- what's the problem?  Is it rented and you'll be testing for a while?  Are there multiple supplies being tested, in production perhaps?  What's the test budget?  Etc.

You could make your own perhaps, but the design time spent for a one-off, won't be cheaper than your off-the-shelf unit, at least if you just need a few.  And it won't be as well characterized and calibrated as the commercial unit is.

Tim

[TheMG]

Why not ramp up the power supply output voltage itself? Much simpler to control the power supply itself than to stick something in series with the load.

Also why can't the load be switched on all at once? Does it have a high inrush current that trips the power supply?

[vini_i]

Uhh...

N...well, yes, I guess, definitely?

You have precisely such a relay -- just connect your insanely expensive DC load in series with the desired load, program it to ramp from 30 to 0V in 1s, and there you have it!

As a lone component, you are expecting something to be able to dissipate up to three kilowatts for a whole second.  A quick perusal of available devices says no.  Very few semiconductor devices handle that much power at all; among them, none do it continuously, or down to an arbitrarily low "on" voltage (e.g., industrial IGBT modules).  Such a unit must use multiple devices in parallel, with a control circuit to stabilize and equalize the power between them.  (A typical e.g. TO-247 sized power transistor will be able to handle maybe 100-300W each over this time frame, so expect to need 10 or more.  Or there are other configurations that can use fewer transistors and a cheaper power dissipator (like a dumb resistor), but at expense to e.g. speed or emitted noise.)

Perhaps it doesn't need to be as precise, or fully-featured, as your load is.  Perhaps a cheaper model, of similar overall capability, would suffice?  I guess I don't get it; you're testing "a" power supply, and you already have the load -- what's the problem?  Is it rented and you'll be testing for a while?  Are there multiple supplies being tested, in production perhaps?  What's the test budget?  Etc.

You could make your own perhaps, but the design time spent for a one-off, won't be cheaper than your off-the-shelf unit, at least if you just need a few.  And it won't be as well characterized and calibrated as the commercial unit is.

Tim

That is exactly how we use the DC load. The load is ramped on and then ramped off.
Perhaps where I may have lost you is we sell the equipment to test the specialty supply. The DC load is around $30k. If we can drop that cost down to $2k that would be the goal.
Also, the relay would only have to tolerate the full 3kW for a short time. The load elements would be high wattage resistors. The relay would ramp the load with the final state being fully on or fully off. The ramp time can be shorter too, one second was just an estimate for the worst case.

[vini_i]

Why not ramp up the power supply output voltage itself? Much simpler to control the power supply itself than to stick something in series with the load.

Also why can't the load be switched on all at once? Does it have a high inrush current that trips the power supply?

The supply is tested like a black box. We have no control over its operation. It gives an output of 28VDC. Its control circuits are slow and a sudden current change will cause it to over voltage or under voltage, tripping the supply.

[T3sl4co1l]

That is exactly how we use the DC load. The load is ramped on and then ramped off.
Perhaps where I may have lost you is we sell the equipment to test the specialty supply. The DC load is around $30k. If we can drop that cost down to $2k that would be the goal.
Also, the relay would only have to tolerate the full 3kW for a short time. The load elements would be high wattage resistors. The relay would ramp the load with the final state being fully on or fully off. The ramp time can be shorter too, one second was just an estimate for the worst case.

Ah!  Then that would make a case for the design option.  Are we talking like, tens, hundreds, thousands of units per year?  If it's a really low quantity thing, you might still want to stick with an off the shelf device; it would be worth taking some time to shop around.  More, the sky's the limit -- it doesn't sound like a hard problem, probably a design budget under $10k would suffice, and I would guess a module of that basic capability might cost a few hundred bucks to manufacture.  Would you need any design assistance, per chance?

Tim

[vini_i]

Ah!  Then that would make a case for the design option.  Are we talking like, tens, hundreds, thousands of units per year?  If it's a really low quantity thing, you might still want to stick with an off the shelf device; it would be worth taking some time to shop around.  More, the sky's the limit -- it doesn't sound like a hard problem, probably a design budget under $10k would suffice, and I would guess a module of that basic capability might cost a few hundred bucks to manufacture.  Would you need any design assistance, per chance?

Tim

That's really the catch. I can sell the bean counters on an off-the-shelf change at $2k. Selling them on a development project would be hard. Especially since this is going to be quantity 2.

[vini_i]

What is the load step that the power supply can tolerate?  Just divide the rated current by that amount and it will tell you how many load resistors you need to switch on in sequence with normal SSRs (or plain power FETs).  Unless the power supply is really bad, it will be much cheaper and simpler than any ramping approach.

As part of the testing, the supply requires a sudden load application to full power (but not too sudden...  ). There are 5 different load steps that are used during testing. In any combination, they still have to be brought up with a ramp.

[TheMG]

How about a DC-DC buck regulator? 28V in, whatever you want out from 0 to almost 28V. That way you can have the buck regulator ramp the load, without having to deal with the large power dissipation problems associated with linear pass solutions (even half a second at 3kW is an eternity when it comes to power dissipation in Silicon devices).

A 3kW rated buck regulator is also a heck of a lot cheaper than a programmable load.

[Siwastaja]

So let me reiterate, the actual problem is you want a programmable load and the existing one at $30000 is just way too expensive?

Failing to find a cheap enough programmable load, you consider designing your own?

In your case, you could make some compromises compared to general-purpose programmable loads, because it would be tailored just for this test case, but under the surface, it's still a programmable DC load.

Compared to designing a general-purpose programmable load, you can skimp on heatsinking (1 second ramps), but not much on semiconductor sizing (1 second is eternity from semiconductor viewpoint). As you only need a few fixed ramps, you can also save on software, controls and UI, which actually are the biggest cost factors in a general purpose instrument.

It can be a classical burn-it-in-transistors linear device, it can be switched resistor solution, it can be a buck regulator into resistors... Many ways to build a programmable DC load.

But we must recognize it's still a programmable DC load.

Did I understand you right you need to build just two?

[Kleinstein]

If small steps are OK, maybe just switch in multiple resistors one at a time.  At the high power level sharung the load to some 50 resistors / FETs may not be so bad. Switching DC can be reasonable simple.

[Siwastaja]

I agree building say 5-bit power resistor "DAC" with low-side NMOS switches would be utterly trivial.

There are going to be some transients when switching, for example when transitioning from 01111 to 10000. I'd hazard an educated guess that low-inductance resistors, fairly fast switching, and some common inductance in the supply helps filter this.

[T3sl4co1l]

Or fairly slow switching.  Switching over 10ms is much easier to manage than 1000ms, and switching 1/32 or whatever of that load is a lot better than the whole thing.

Although if you're doing a binary series, you need that switching to overlap very carefully indeed.

I have a design I like to use for this sort of thing; or, liked to, considering LM3914 is officially obsolete now...  LM3914 is, more technically, better described as a unary 10-bit ADC, so it works nicely to drive an array of equal value resistors -- particularly useful for dissipating power over a modest operating range, albeit not over a huge range of voltage or current.

Tim

[vini_i]

So let me reiterate, the actual problem is you want a programmable load and the existing one at $30000 is just way too expensive?

Failing to find a cheap enough programmable load, you consider designing your own?

In your case, you could make some compromises compared to general-purpose programmable loads, because it would be tailored just for this test case, but under the surface, it's still a programmable DC load.

Compared to designing a general-purpose programmable load, you can skimp on heatsinking (1 second ramps), but not much on semiconductor sizing (1 second is eternity from semiconductor viewpoint). As you only need a few fixed ramps, you can also save on software, controls and UI, which actually are the biggest cost factors in a general purpose instrument.

It can be a classical burn-it-in-transistors linear device, it can be switched resistor solution, it can be a buck regulator into resistors... Many ways to build a programmable DC load.

But we must recognize it's still a programmable DC load.

Did I understand you right you need to build just two?

I see where the confusion is. I don't want a programmable DC load. I have two fixed load steps, 50A and 100A. For 90% of the operation, the power supply sits at 50A. For one test the supply is taken from 0A to 100A, then it must sit there for about 20 minutes. For either load step, the application and removal needs to happen with a ramp. One second is generous. 20 to 50 milliseconds is far more realistic.

The idea is that the loads would consist of large resistors. Ceramic cores wound with nichrome wire, that kind of stuff. All I'm looking for is something that will bring those resistors online and offline with a ramp. Assuming the ramp is perfectly linear the most wattage the element would dissipate is 750w and then only for a few milliseconds. The end conditions are fully on or fully off. The element does not actually carry the load.

[JohanH]

Don't know if this would work for you, but there are thermistors with 80A rating (for limiting inrush current). You could use a relay to bypass it.