5

On the PCB I'm designing, the total power consumption might be as high as 86W (including output power), with supply voltage between 18 and 75V DC (it is OK not to allow full output power bellow 22V). So I'm fine with 4A max input current.

I'm trying to find how to do the input "fuse". I would like some non destructive over-current protection, as opening the system to change a fuse would void calibration (calibration is expensive, and can't be done in the field).

I was initially thinking about using PTC fuses, but I couldn't find any the both allowed 4A continuous current, and resisted to more than 75V. The nearest thing I found was 0ZRM0375FF1A, which is max 3.75A and max 72V (and is far to big : 34mm high!)

So what kind of component/circuit might enable to get an over current protection :

  • that is non destructive (ie operational after power-off, or after some time, or ... : anything as long as it don't require opening the device)
  • can withstand 4A DC continuously, and block or limit if current exceeds this current (I don't have any hard requirements there, it can be fuse like with an It², or current based, ...)
  • can withstand 75V nominal (so at least 100V would be nice)
  • reasonably small PCB area (in particular height <15mm)
  • if possible low power dissipation when not triggered (1W max, ideally <100mW)

Do you know of any technology that might work? Or maybe there is some active solution based on current sensing+MOSFET?

In the later case, it might help that there is already some reverse current protection in place (ideal diode):ideal diode So the "fuse" might be placed after the ideal diode if avoiding the risk of reverse current makes it easier, or before if it is possible (or it might be somehow combine with the ideal diode (which does reverse current protection and current Oring)

EDIT : Answers to the various comments : First off all, the board will be the first version of a prototype, so specifications are not well defined yet : only high level specifications are set yet, for the rest it's mainly "do as good as you can", with the particularity that cost isn't very important (if you have a 50$ solution, it's still fine), while PCB space and heat dissipation are far more important.

@TimWilliams Source can be anything (it's up to the client to provide it). Strict minimum is a 24DC (with some margin). As the secondary source is PoE, which is up to 57V, we used 18-75V converters, so it is a nice plus to have everything working from 18 to 75V (a good solution with slightly lower voltage limit is still acceptable). The Load is composed of 2 DC/DC converters, and some connection for user-specific load.

For transients, I have to admit that I don't really know what to protect for exactly : there aren't any low level requirements, it's just "protect as well as you reasonably can", while keeping PCB size and power as low as possible . I added a SMF60A_R1_00001 for some transient voltage protection. But if you have any improvements to suggest, I'm interested.

Any current above 4.5A for more than "startup" is definitively a fault. Cutting at 4A is also fine (the 18V supply + max consumption is really an edge case, so I don't care if we derate the system a bit bellow 22V)

For how fast to react, again, I don't have any specific requirements. Slow enough that it isn't triggered during power up (or faster, but limiting current instead of cutting it), and "as fast as possible" otherwise. As it is the first prototype, I'd rather have it cutting a bit too slow than a bit too fast, so at least I will be able to measure startup current to design it better for next iteration.

I don't really care about the voltage drop in itself as long as it don't exceed 1V (even 4V might be acceptable), but we are having quite a hard time managing thermal dissipation (waterproof enclosure), so I would like to keep power "loss" as low as possible (1W is really the max, it would be nice to be bellow 0.1W).

Calibration matters because there is a camera in the system which needs high quality precision, that is likely to be invalidated if opening the system. Calibrating again is a complicated process.

Generating some noise is OK (the signal will just power some DC/DCs), but I don't think a buck converter can be withing the 1W dissipation limit (it would require 98.8% efficiency)

@vir : overvoltage protection was a type (I corrected it). As mentionned, I added a TVS to help a bit against overvoltage transients. If you have any suggestions to do better, I'm interested (if it is for continuous overvoltage protection, it would only make sense if resisting 320V DC or 230VDC, but those cases are quite unlikely, so I wan't planing to protect against them : but if you have a compact solution, why not).

For fuses/breakers outside the enclosure, it's sadly not an option, as the device has to be waterproof

Sandro
  • 5,519
  • 6
  • 24
  • 1
    What is the source and load? What about transient (startup, surge, fault) conditions? How fast should it operate? How much voltage drop (other than PCB area / thermal limitations)? Why does calibration matter? Could it generate noise (e.g. current mode buck converter) or does it need to be quiet too? – Tim Williams Mar 06 '23 at 18:38
  • You mention over-voltage protection; is this a typo or do you also need overvoltage protection? A common way of doing this would be to have the input power pass through either a fuse in a panel-mount holder or a circuit breaker before making it to the board. Neither of these solutions would require opening the chassis to replace/reset the element after an over-current situation. Is there a reason your system couldn't use one of them? – vir Mar 06 '23 at 18:38
  • Whats that? How could changing a fuse render the calibration invalid? – datenheim Mar 06 '23 at 18:53
  • 2
    If you use a *panel mount fuse holder*, you can change the fuse without opening the system. – Andrew Morton Mar 06 '23 at 19:14
  • @TimWilliams @ vir @ datenheim @ andrewmorton : I answered your comments at the bottom of my question, as the answers are too long for comments – Sandro Mar 06 '23 at 20:08
  • Are you sure wrt startup? 86W at the 18V minimum is 4.77A at 100% efficiency, leaving no room to charge capacitors at any rate. Or 3.9A at 22V if UVLO is set there, leaving merely 0.6A headroom. A fuse won't blink at this, because it takes time to blow (~10ms), and the 10s (or 100s, or 1000s even) of amperes of inrush current might not bother it. But a semiconductor design will be critically dependent on how much is needed or allowed, because transistors die in ~100s µs! As for fuses, IP6x holders are available, would that suffice? – Tim Williams Mar 06 '23 at 20:42
  • @TimWilliams : I might not have been clear enough : excepted durring startup, the 4A limit is fine. For startup, there will be a bigger transient current to charge capacitors, which should not shut down the system (either a short delay should be allowed as in a fuse, or the system should just clamp current, which would simply result in slower power-up, which is fine). The issue is, I have no way to really estimate the current inrush for this first prototype. – Sandro Mar 06 '23 at 20:52
  • For the max current, the 86W use case is quite rare : usually there will be no custom output at all, or if there is it will draw little current. For those cases 4A@18V are more than enough (startup inrush excluded). What's more, the user will usually provide either 24V or 48V, so 18V is of little interest for most clients. So I have no problem specifying the product as input voltage 18-75V without custom output, and 22-75V with custom output. – Sandro Mar 06 '23 at 20:56
  • If you can get estimates of (or specify) the maximum load capacitance, that'll give you some idea of how much startup current and energy will be needed. – Tim Williams Mar 06 '23 at 20:59
  • I'm no longer at work, so I don't have exact specifications data, but from memory, the total capacitors "explicitly" we added on the input rail sum up to something in between 13 and 30µF. In addition, there are probably some capacitors inside the DC/DC converters (for the 5V converter, I'm not sure about the reference, I will have to check tomorrow ; for the 3.3V converter (starting a bit later), input capacitance is not specified, only that startup time is 30ms typical and 40ms max)) – Sandro Mar 06 '23 at 21:16

1 Answers1

9

It sounds like you need a "hot swap controller" or "surge suppression IC". For example, the Analog Devices LTC4364:

enter image description here

enter image description here

The LTC4364 gives you over current, over voltage, and under voltage protection, all of which can be set using resistors. With a second mosfet (shown tied to DGATE) you can have reverse voltage (positive and negative swapped) protection. This particular IC has a low power draw (less than 1 mA), with the mosfets typically being the tallest components. They make versions that shut down after a single event (say an over voltage event), and them make versions that will try turning on after an event. The IC can be reset by pulling SHDN low and then high again. Before you ask, the "Vout clamped at 27 V means that during a surge the output is clamped at 27 V. During non-surge events Vout = Vin.

Analog Devices is not the only maker of hot swap controller and surge suppression ICs, I just used it as an example. There are surge suppression ICs that can survive up to 500 V, some that have digital interfaces that allow for more precise feed back, and some that can handle multiple channels.

SamGibson
  • 17,231
  • 5
  • 37
  • 58
C. Dunn
  • 961
  • 1
  • 5
  • 7
  • Thanks a lot. It sounds like the right family of components. I will look into them in more detail. And given that some seem to also do the reverse current protection, I will probably be able to remove the corresponding circuit, making the aditionnal PCB space quite small – Sandro Mar 06 '23 at 20:19
  • This is a good option for overvoltage limiting; the next best is probably a crowbar circuit and fuse (assuming the fuse can be replaceable/resettable of course). – Tim Williams Mar 06 '23 at 20:44
  • Note that MOSFET based protection circuits have a habit of failing in short-circuit when exposed to overvoltage. I would additionally have a traditional one-time-use fuse to avoid a fire in that situation. – jpa Mar 07 '23 at 08:12