-
ideas for a "terrible" DC/DC buck converter that never really needs to operate?
Posted by max_torque on 16 May, 2024 12:55 -
Hi All,
I am looking at overvoltage protection for a 24vnominal system that must continue to operate normally with up to 200v being applied to it for 500ms!
Previsouly we have used clamps/crowbars and disconnects, but the regs have just changed and we can no longer either simply absorb / clamp the overvoltage pulse and we must now continue to operate normally during the pulse.
Now, this overvoltage event actually never really happens, so most of the time (99.9999% or more actually) i need a device that just sits their, passes through 50A continuous whilst dropping the least amount of power and making the least amount of heat. But, for the test itself, and perhaps once every 25 years in service, it may actually need to step in and limit outputvoltage to under say 35ish volts whilst still passing 50A to the load and whilst up to 200V is present at the input.
This is one of those really stupid things that is more a sideeffect of out-of-date regs than something that has much use in the real world, but it's a test we have to "pass" ie to demonstrate that our system can handle this event.
During the event, i'm really not to concerned about the buck efficiency, the current ripple or to a large degree, the voltage ripple on either side of the converter either,all things that normally you'd want to be minimised, but here, it's simply not very useful to minimise those things. As a result i'm wondering about how little inductance i can use, how far i can push that inductance into saturation during the event, and what sort of control architecture to implement.
50A continuous is a reasonable level of current to have to deal with on a continous basis, and the application demands automotive environmental level design (ie -45 to 105degC ambient). A quick estimate suggests we'd want something around 10mOhm total resistance, and with the Switching elements having to be 250 V rated that means a fair number of parallel FETs. It also means a fairly bulky inductor, hence the need to run it into saturation as much as we dare?
So, i'd like any suggestions as to how to achieve this. I don't really want to have complex "Bypassing" eg a contactor or relay that helps reduce the normal series resistance when the system is not bucking around as it were, an dideally we'd get continuous reverse polarity protection included "for free" in the architecture, ie when connected backwards zero current is allowed to flow.
I've used Wurth WE-HCF inductors previous for an automotive rated DC/DC previously, but that actually operated allthe time, whereas here we basically never operate
https://www.we-online.com/en/components/products/WE-HCF
I've considered some form of linear dropper ie pass transistor, but the powers are really quite large (200 to 35v is a 165V drop, which at 50A is 8.25kW, so even for 250ms that's around 2kJ of energy) Not sure if even some massive series cascade of To-3 cased monsters can deal with this level of energy?
-
Is it actually 200V at the terminals? From what I remember, load dump specs specify both the voltage and internal resistance of the pulse source (I assume that's what you're testing)
-
https://www.analog.com/en/resources/technical-articles/loaddump-protection-for-24v-automotive-applications.html
We used a MAX16127. Wonderful IC.
The input is protected by a LARGE TVS, such as this: https://www.mouser.com/ProductDetail/Bourns/15KPA040C-SD-Q?qs=BJlw7L4Cy78s85IvP68mjg%3D%3D&_gl=1*fskbe9*_ga*dW5kZWZpbmVk*_ga_15W4STQT4T*dW5kZWZpbmVk*_ga_1KQLCYKRX3*dW5kZWZpbmVk -
As i mentioned we can not clamp nor disconnect. We are not doing Auto cert we are doing MIL cert, which now means unlimited energy and a requirement to stay fully operational during the pulse!
Hence the DC/DC buck approach..... -
Is it actually 200V at the terminals? From what I remember, load dump specs specify both the voltage and internal resistance of the pulse source (I assume that's what you're testing)
The regs just changed and we are awaiting clarification as the new regs actually look to have a mistake in them that does not show the 500mOhm output impedance for the surge generator and the energy limitation has been removed. Other regs, that could become applicable to us have never included these factors, so yes, worst case is 200V at the input terminals to our device! -
Here's an idea:
Quickly sense the surge voltage and switch in a high-energy pulse-rated resistance to limit the energy to a subsequent clamp. I'm sure that is more difficult in practice than in concept but maybe cheaper and more reliable than trying to have such a wide range buck regulator. -
Previsouly we have used clamps/crowbars and disconnects, but the regs have just changed and we can no longer either simply absorb / clamp the overvoltage pulse and we must now continue to operate normally during the pulse.
What is the difference between absorbing the extra energy (using, theoretically, a beefy 35V zener or linear reg) and normal operation, if in both cases the device subjected to such a pulse will continue to operate as if nothing happened, i.e. perform its normal functions? -
Here's an idea:
Quickly sense the surge voltage and switch in a high-energy pulse-rated resistance to limit the energy to a subsequent clamp. I'm sure that is more difficult in practice than in concept but maybe cheaper and more reliable than trying to have such a wide range buck regulator.
I actually thought about this one and have simulated it, and the problem is that out EUT has a very wide range of possible operational currents, from as little as 3A min to about 50A max depending on what mode it is in at the moment of the surge. As such, picking the resistance of the "limiting R" is fairly tricky, as it's a balance between still providing enough energy for the system to work and not exploding the clamping devices!
I think it would actually be do-able, my concern is to get it to work across all operational conditions may end up being a large amount of work and testing and changing things around. With a Buck converter, i think it could actually be easier to get it all working, esp if i use some dgital control logic to drive it, so it becomes in some respect a s/w calibration problem......
-
Previsouly we have used clamps/crowbars and disconnects, but the regs have just changed and we can no longer either simply absorb / clamp the overvoltage pulse and we must now continue to operate normally during the pulse.
What is the difference between absorbing the extra energy (using, theoretically, a beefy 35V zener or linear reg) and normal operation, if in both cases the device subjected to such a pulse will continue to operate as if nothing happened, i.e. perform its normal functions?
If yo look in my OP you'll see i estimate how much energy needs to be absorbed by any kind of clamp or linear system, and personally i just don't think this is practically do-able...... -
There are practical aspects to this. Is the circuit for the test equipment available? Are they really tossing out the prior test equipment because of an error in a standard or will they use the equipment they've always used? I suppose they can charge up some low ESR capacitors and develop a huge current but it won't be infinite given test lead inductance, switch resistance etc. Can you publish an excerpt of the relevant standard or paraphrase the part of the standard you are worried about? Is there a rise-time/fall-time specified?
-
One think i think i'm stuck with will be "too much" switch as i have to protect for continuous 50A current carrying in a 105degC ambient, so probably my potential short duration pulse current capability is going to be massive, as such, i'm not sure that i care if i push the limiting inductor into quite a lot of saturation. If i have a fast current measurement / limitation methodolgy (not sure what this looks like yet tbh) then i might get away with a control strategy that effective has a "ALL NORMAL" control mode, where the switches are ON 100% of the time, and then sense input voltage climbing above a threshold (could be an analoug measurement or digital via a comparitor with suitable threshold) and so swap into "ABNORMAL" mode, where i just trigger short pulses of ON time to let through chunks of energy. Duty cycle and/or frequency can be modulated here based on readings of input and output voltage and average current through the device.
With a small series inductance,potentially heavily into saturation, the rate of rise of current through the inductor is going to be fast, so i will want a pretty quick response and control feedback loop, however, if my switches can handle short high current pulses, i don't honestly care too much about the input and output ripple and can use some capacitance to help smooth the current ripple to limit voltage ripple. And as i don't care about EMI i can drive those switches really hard ON/OFF and the event only lasts 500ms, so i can tollerate a significant additional switching loss i think?
-
There are practical aspects to this. Is the circuit for the test equipment available? Are they really tossing out the prior test equipment because of an error in a standard or will they use the equipment they've always used? I suppose they can charge up some low ESR capacitors and develop a huge current but it won't be infinite given test lead inductance, switch resistance etc. Can you publish an excerpt of the relevant standard or paraphrase the part of the standard you are worried about? Is there a rise-time/fall-time specified?
Absolutely right!
As it stands we are awaiting an official response from the technical authority legally responsible for issuing the standards (sorry, not at liberty to disclose any more here) but as the changes from the previous standards are far reaching and as it stands, the test houses we normally use for these tests (to the old standard) cannot actually do the new standard test because there equipment is not powerful enough as the energy limitation has been removed from the standard!
For one of the tests we are building our own spike/surge generator to meet the requirements and for one test we will use a 22kW 440V 3ph mains supplied high speed digital PSU to be able to accurately synthesize the test waveform now called for.
Min and Max Rise and fall times are indeed specified and for the 200V surge it's nominally a relatively low 40kV/s
-
If yo look in my OP you'll see i estimate how much energy needs to be absorbed by any kind of clamp or linear system, and personally i just don't think this is practically do-able......
Makes sense. Well, might be not that terrible if the spike lasts only 500 ms, but yeah.
Anyway, AC/DC switching PSUs can easily support a wide input voltage range (say 80-240V, and I guess it's not that hard to design one that has an extended lower end of the range), and any AC/DC SMPS is essentially a DC/DC SMPS after the input diode bridge and the filter cap. Is there any reason why using just a regular SMPS, modified, if necessary, for a wider input voltage range, wouldn't work? -
For some reason, this forum SW eats responses sometimes. They just don't show up. Anyway, this may be a redundant post but:
I need to leave in a few minutes on a vacation but will check back when I can. Sorry to leave the conversation abruptly. It is an interesting problem.
Without having thought it through at all, a tapped buck inductor maybe? Not sure if your regulator output is isolated or not... -
a control strategy that effective has a "ALL NORMAL" control mode, where the switches are ON 100% of the time, and then sense input voltage climbing above a threshold (could be an analoug measurement or digital via a comparitor with suitable threshold) and so swap into "ABNORMAL" mode, where i just trigger short pulses of ON time to let through chunks of energy. Duty cycle and/or frequency can be modulated here based on readings of input and output voltage and average current through the device.
If your load can tolerate the L and C under normal operation, that certainly seems like the simplest approach. I'm not sure you'd need to look at the input voltage for the modulation, seems like you ought to be able to operate like a normal voltage-regulating buck converter and just look at the output voltage and/or current with whatever control mode. A constant off-time mode triggered by the output voltage rising above some threshold seems like the simplest solution, since you could operate in that mode continuously, rather than needing to switch control modes between normal and abnormal conditions, and you could add current limiting the same way. You could use the input voltage to estimate switching losses for thermal protection, but unless you need to maximize your operating area it's probably simpler to just have a pulse count vs time limit. -
…..but as the changes from the previous standards are far reaching and as it stands, the test houses we normally use for these tests (to the old standard) cannot actually do the new standard test because there equipment is not powerful enough as the energy limitation has been removed from the standard!
If they have created an untestable nonsense, I would relax and wait.
….
Min and Max Rise and fall times are indeed specified and for the 200V surge it's nominally a relatively low 40kV/s
Until anyone can provide a real tester, it is all pie in the sky stuff.
With no one able to test, they will be under serious pressure to be less stupid.