Author Topic: BMS short circuit interruption capability  (Read 4016 times)

0 Members and 1 Guest are viewing this topic.

Offline BiduleohmTopic starter

  • Contributor
  • Posts: 38
  • Country: fr
BMS short circuit interruption capability
« on: March 24, 2020, 04:33:38 pm »
Hi guys :)

Long story short I work on a BMS (battery management system) and I need some help on the disconnect part of it which is basically a 300 A 64 V SSR. If you want the whole story you can see it here.

The problem I have is that I want to be able too interrupt a 10 kA short-circuit current with the mosfets surviving. The interrupt part is ok but the mosfets surviving isn't because of the inductive spike.

I took the (defavorable) case of the battery connected with 2 m of 10 mm diameter wire, which is about 2.4 µH, from my calculations.

The battery can be considered as a 64 V voltage source with a short circuit current of 10 kA.

I have 2x 10 mosfets connected source to source to form a bidirectional switch on the negative side of the battery. Each mosfet can handle 1 kA for 10 µs and 1.1 kA for 5 µs, I calculated the total time from over current detection to mosfet turned off to be under 4.46 µs wort case. I use these mosfets.

I used the formula from this answer to calculate the power dissipated by a 64 V TVS and I get about 100 kW (120 J in 1.2 ms). That's a lot but do-able with multiple TVS diodes in parallel; the real problem is that 64 V TVS diodes have a clamping voltage of 103 V at best which is obviously a lot higher than the 80 V the mosfets can handle. 100 V mosfets start to be in the too high Rdson domain, but more importantly they are a lot more expensive than the 80 V ones and the project is highly cost sensitive.

But then is occured to me that the inductance would need about the same time to store the energy than it needs to unstore it (about 1.2 ms) and I'm capable of switching off the current 250 times faster than that (5 µs) so it shouldn't have the time to store a lot of energy, and so it should be a lot easier to dissipate it and not kill the mosfets.

I attached the schematic as a PDF but I can provide the KiCad files if needed. The section of concern is in the middle, the mosfets are Q15x and Q16x. I only put 4 of the 20 mosfets for now because if I modify something I don't want to have to redo all of it. B- is connected to the battery negative and P- is connected to the loads negative. I can also provide the schematics of the over current detection section if needed.

If you wonder about the 100 nF caps and the schottky diodes on the mosfets it's to prevent them to turn on in case of a high dV/dt on the drains (like when someone is connecting the battery to something for example). The two gates drivers in parallel are for redundancy and a lower turn-off time.

Is the disispated power formula and my numbers correct?

Am I correct about the fact that the energy stored will be a lot lower than 120 J?

If so, how much energy should I expect to have to dissipate in how much time?
« Last Edit: March 24, 2020, 04:37:57 pm by Biduleohm »
 

Offline strawberry

  • Super Contributor
  • ***
  • !
  • Posts: 1199
  • Country: lv
Re: BMS short circuit interruption capability
« Reply #1 on: March 24, 2020, 05:34:17 pm »
transistor amplified zener
 

Offline BiduleohmTopic starter

  • Contributor
  • Posts: 38
  • Country: fr
Re: BMS short circuit interruption capability
« Reply #2 on: March 24, 2020, 05:39:55 pm »
Yes, I tought about an active circuit but I need to know what kind of energy I'll have to dissipate to design it accordingly.
 

Offline jbb

  • Super Contributor
  • ***
  • Posts: 1265
  • Country: nz
Re: BMS short circuit interruption capability
« Reply #3 on: March 24, 2020, 11:49:44 pm »
How about some alternatives:
- Reduce MOSFET off speed to reduce dV/dt. Will increase heat dump into MOSFETs, so check Safe Operating Area (SOA)
- High rupture current DC fuse to break the dead short (delay MOSFET turn off a little so the fuse goes)
- Limit MOSFET drive voltage a little (depends on devices) to use inherent MOSFET current limit (ie FETs move from saturation to linear region) and don’t let it get to 10kA. You can also add desaturation protection for fast overcurrent detection
- maybe you could use a crowbar clamp, i.e. when the voltage starts to surge up (as the MOSFETs start to reduce the current), use an SCR to short the output rail to 0V. It’s vicious, but it’ll keep the voltage spikes out.
 

Offline BiduleohmTopic starter

  • Contributor
  • Posts: 38
  • Country: fr
Re: BMS short circuit interruption capability
« Reply #4 on: March 25, 2020, 12:10:37 am »
- Reduce MOSFET off speed to reduce dV/dt. Will increase heat dump into MOSFETs, so check Safe Operating Area (SOA)

I can't reduce the dV/dt or the mosfets will fry.

- High rupture current DC fuse to break the dead short (delay MOSFET turn off a little so the fuse goes)

Same problem, the mosfets will protect the fuse but not the other way around.

- Limit MOSFET drive voltage a little (depends on devices) to use inherent MOSFET current limit (ie FETs move from saturation to linear region) and don’t let it get to 10kA. You can also add desaturation protection for fast overcurrent detection

The mosfets are fully saturated in normal use to have the lowest Rdson possible so I guess that will not be possible.

- maybe you could use a crowbar clamp, i.e. when the voltage starts to surge up (as the MOSFETs start to reduce the current), use an SCR to short the output rail to 0V. It’s vicious, but it’ll keep the voltage spikes out.

That's likely what I'll be doing (but not to 0 V as it would maintain the short circuit, it'll be somewhere between 64 and 80 V) but where I'm stuck is calculating what power/energy I will need to dissipate with this circuit.
 

Offline jbb

  • Super Contributor
  • ***
  • Posts: 1265
  • Country: nz
Re: BMS short circuit interruption capability
« Reply #5 on: March 25, 2020, 03:23:46 am »
- Limit MOSFET drive voltage a little (depends on devices) to use inherent MOSFET current limit (ie FETs move from saturation to linear region) and don’t let it get to 10kA. You can also add desaturation protection for fast overcurrent detection

The mosfets are fully saturated in normal use to have the lowest Rdson possible so I guess that will not be possible.

Hmm.  From the data sheet, Figure 11 shows that Rdson is basically flat past 6V (pretty flat at 5V, even) at 4 mOhms (double this for real world calculations).  Under non-fault conditions, Figure 9 suggests even 200A per device won't push them out of saturation.  For 400A peak load, that's a minimum of 2 in parallel.

Under fault conditions with a 6V drive and a bit of guesswork off the top of Figures 9 and 10, we could guesstimate that the devices will self limit current to very roughly 600A each (it's worse when they're hot).

Do you want to handle 300A continuously?  That's going to be difficult... Let's also have a look at thermals.  Assuming a large heatsink of ~1 K/W, and a DT of 40 K  (45 C ambient  to 85 C heatsink), you could dissipate is 40W.  That indicates an allowed drop of 40 / 300 = 120mV.  And thus a resistance of 0.4 mOhms.  Given that the current passes through 2 switch banks, you have 0.2mOhms per bank.  That's going to be 20 in parallel.  At this point, you can't rely on the MOSFETs self limiting the fault current, unfortunately.



Hold on, I may be a dummy.  The fault current will be sourced out of the batteries, yes?

In that case, the fault current will attempt to drag the output node below ground when you turn the MOSFETs off.  You could just use a diode clamp to prevent the voltage from swinging below ground, and this would naturally clamp the MOSFETs to roughly the battery pack voltage.
 

Offline mzzj

  • Super Contributor
  • ***
  • Posts: 1298
  • Country: fi
Re: BMS short circuit interruption capability
« Reply #6 on: March 25, 2020, 07:00:53 am »
Hi guys :)

But then is occured to me that the inductance would need about the same time to store the energy than it needs to unstore it (about 1.2 ms) and I'm capable of switching off the current 250 times faster than that (5 µs) so it shouldn't have the time to store a lot of energy, and so it should be a lot easier to dissipate it and not kill the mosfets.
Overcurrent trigger at 300A and 5us delay?  You need to turn off only 430A at that time.  Muchos easier than 10kA.
And don't try to dissipate the inductive energy in TVS when you can add freewheeling diode(s)?

You also probably better to account for battery pack inductance too.
 

Offline Daixiwen

  • Frequent Contributor
  • **
  • Posts: 367
  • Country: no
Re: BMS short circuit interruption capability
« Reply #7 on: March 25, 2020, 09:19:38 am »
The voltage on the MOSFets can exceed the 80V breakdown voltage, under certain conditions. They will enter avalanche mode and as long as the avalanche energy doesn't exceed the limits they will survive. Figure 6 in the datasheet also shows you how long you can stay in avalanche depending on the current, and gives you pointers to two application notes on the avalanche mode. I think AN7514 could help you.

At that point I think you'd better use a simulator to see what kind of current and voltages you can expect when switching off during a short circuit. The worst case would be starting with a continuous current through the 2.4uH inductance just under your short circuit detection threshold, and a sudden short circuit. If you can find a SPICE model of your mosfets that include the avalanche mode it will be even easier to see if you are in the safe operating area.
 

Offline BiduleohmTopic starter

  • Contributor
  • Posts: 38
  • Country: fr
Re: BMS short circuit interruption capability
« Reply #8 on: March 25, 2020, 03:32:16 pm »
Hmm.  From the data sheet, Figure 11 shows that Rdson is basically flat past 6V (pretty flat at 5V, even) at 4 mOhms (double this for real world calculations).  Under non-fault conditions, Figure 9 suggests even 200A per device won't push them out of saturation.  For 400A peak load, that's a minimum of 2 in parallel.

Under fault conditions with a 6V drive and a bit of guesswork off the top of Figures 9 and 10, we could guesstimate that the devices will self limit current to very roughly 600A each (it's worse when they're hot).

Do you want to handle 300A continuously?  That's going to be difficult... Let's also have a look at thermals.  Assuming a large heatsink of ~1 K/W, and a DT of 40 K  (45 C ambient  to 85 C heatsink), you could dissipate is 40W.  That indicates an allowed drop of 40 / 300 = 120mV.  And thus a resistance of 0.4 mOhms.  Given that the current passes through 2 switch banks, you have 0.2mOhms per bank.  That's going to be 20 in parallel.  At this point, you can't rely on the MOSFETs self limiting the fault current, unfortunately.

It's actually 2.4 mOhms worst case @ Vgs = 10 V and Tj = 120 °C (also confirmed by the electrical parameters table, 1.4 worst case @ 25 °c and 3.1 worst case @ 175 °C), I will never go past Tj = 100 °C and Ta = 55 °C (I have temperature sensors to ensure that).

I have 10 mosfets in // (as said, only 2 on the schematic for now because it'll probably change and I don't want to have to change dozens of components if I can avoid it) so that's 30 A per mosfet continuous, 40 A surge for max 30 sec.

So each mosfet will dissipate at most 2.2 W continuous and 3.9 W surge for 30 sec. They will be on a 2 oz double sided PCB with lots of thermal vias, heatsink on the bottom side and a copper busbar along the drains on the top side (both for current capability and thermal equalization between mosfets). And same thing for the other 10 mosfets.

In theory 6 to 8 mosfets would be enough thermally speaking but they are limited to 1.1 kA each so for the 10 kA capability I need 10 of them.

Hold on, I may be a dummy.  The fault current will be sourced out of the batteries, yes?

In that case, the fault current will attempt to drag the output node below ground when you turn the MOSFETs off.  You could just use a diode clamp to prevent the voltage from swinging below ground, and this would naturally clamp the MOSFETs to roughly the battery pack voltage.

Yes, current will be from the battery so it'll flow from P- to B-. You can also have a charge current the other way around, but it'll be limited by the charger so no chances to ever go into the kA domain even with the worst charger fault you can have.

I can't use a diode as the switch must be bidirectional and cut current in both directions.



Overcurrent trigger at 300A and 5us delay?  You need to turn off only 430A at that time.  Muchos easier than 10kA.
And don't try to dissipate the inductive energy in TVS when you can add freewheeling diode(s)?

You also probably better to account for battery pack inductance too.

The OCP is actually at around 425 A because I want to be able to handle a 400 A surge for 30 sec, anything above that is considered as a short and should be interrupted in 4.46 µs worst case (but that's from datasheets, I haven't mesured in the real world yet). The mosfets can handle the 10 kA for 5 µs @ Tj = 100 °C with 10 % margin and for 10 µs with no margin.

I assume the worst case were the current has risen to 10 kA in less than 5 µs, so faster than I'm able to detect and interrupt. It may be unrealistic, I'm not sure about what is the worst dI/dt I can expect from a short circuit.

I can't as the switch is bidirectional, I must limit the voltage between P- and B- between 64 and 80 V and between -64 and -80 V (but once I can do it one way it shouldn't be a problem to do it the other way).

Yep, I can double the inductance and use 5 µH but it'll not make a big difference at this point.



The voltage on the MOSFets can exceed the 80V breakdown voltage, under certain conditions. They will enter avalanche mode and as long as the avalanche energy doesn't exceed the limits they will survive. Figure 6 in the datasheet also shows you how long you can stay in avalanche depending on the current, and gives you pointers to two application notes on the avalanche mode. I think AN7514 could help you.

At that point I think you'd better use a simulator to see what kind of current and voltages you can expect when switching off during a short circuit. The worst case would be starting with a continuous current through the 2.4uH inductance just under your short circuit detection threshold, and a sudden short circuit. If you can find a SPICE model of your mosfets that include the avalanche mode it will be even easier to see if you are in the safe operating area.

I thought about that but they can't handle much energy in avalanche (and Figure 6 tell us they can't go past 300 A each, that's only 30 % of what I need them to handle) and I prefer to stay out of avalanche mode anyway.

I'm worried about the simulation omitting some important details. Also finding the SPICE model will probably not be easy (impossible?) but I'll look for it.
« Last Edit: March 25, 2020, 03:36:14 pm by Biduleohm »
 

Offline Siwastaja

  • Super Contributor
  • ***
  • Posts: 9336
  • Country: fi
Re: BMS short circuit interruption capability
« Reply #9 on: March 25, 2020, 04:24:59 pm »
Add inductance to limit dI/dt, use a fast OCP, then you can guarantee the current can't go above certain value. Use freewheeling diodes to discharge the inductance into DC link capacitance; large part of capacitance should be of higher ESR to prevent ringing.  A crowbar circuit of some kind may be needed, but it can be slow because the capacitance limits the voltage rise speed. Though, just using enough electrolytic capacitors (calculate for maximum voltage swing when all the inductance-stored energy discharges into the capacitors) may be cheaper.

Freewheeling diodes need to be rated for single pulse only, hence shouldn't be expensive or too big.

The faster you can make the OCP, the less inductance you need. You may want to measure the current rise rate of a shorted battery; it may have enough inductance already. Inductance limits the current rise speed, while resistance limits the maximum current itself. Using enough inductance removes the short-circuit current (defined by the total system R) from the equation.
« Last Edit: March 25, 2020, 04:29:02 pm by Siwastaja »
 

Offline BiduleohmTopic starter

  • Contributor
  • Posts: 38
  • Country: fr
Re: BMS short circuit interruption capability
« Reply #10 on: March 25, 2020, 04:57:32 pm »
Increasing the inductance can only make the problem worse as it'll stock more energy, also it's not very practical nor cheap given the current it should handle (300 A continuous...).

Use freewheeling diodes to discharge the inductance into DC link capacitance; large part of capacitance should be of higher ESR to prevent ringing.  A crowbar circuit of some kind may be needed, but it can be slow because the capacitance limits the voltage rise speed. Though, just using enough electrolytic capacitors (calculate for maximum voltage swing when all the inductance-stored energy discharges into the capacitors) may be cheaper.

Freewheeling diodes need to be rated for single pulse only, hence shouldn't be expensive or too big.

I thought about using capacitors but I'm worried about resonance because I don't have a predetermined inductance as it's dependent on the wire size, length, etc... I took the pretty defavorable case of 2 m of wire but it can be anywhere from 10 cm to a few meters.

I can't use freewheeling diodes, the switch is bidirectional.

The faster you can make the OCP, the less inductance you need. You may want to measure the current rise rate of a shorted battery; it may have enough inductance already. Inductance limits the current rise speed, while resistance limits the maximum current itself. Using enough inductance removes the short-circuit current (defined by the total system R) from the equation.

It's already as fast as I could make it. And I should already have received the battery but the corona virus decided to change that...

The problem isn't the short circuit current, the mosfets can handle it, it's the inductive voltage spike.

What I need to know is how much energy is stored in a 2.4 µH inductor fed by 64 V @ max 10 kA for a duration of 5 µs.
 

Offline mzzj

  • Super Contributor
  • ***
  • Posts: 1298
  • Country: fi
Re: BMS short circuit interruption capability
« Reply #11 on: March 25, 2020, 09:07:48 pm »

What I need to know is how much energy is stored in a 2.4 µH inductor fed by 64 V @ max 10 kA for a duration of 5 µs.
OCP set to 420 A would give you 550 A  after 5 us short and 0,35 J
 
The following users thanked this post: Biduleohm

Offline BiduleohmTopic starter

  • Contributor
  • Posts: 38
  • Country: fr
Re: BMS short circuit interruption capability
« Reply #12 on: March 25, 2020, 10:45:28 pm »
Thanks a lot ;)

What formula did you use? As I want to be able to do the calculation for other values.
 

Offline mzzj

  • Super Contributor
  • ***
  • Posts: 1298
  • Country: fi
Re: BMS short circuit interruption capability
« Reply #13 on: March 26, 2020, 09:44:45 am »
Thanks a lot ;)

What formula did you use? As I want to be able to do the calculation for other values.
dI=dt*V/L
dI = 5us*64v/2.4uH = 133A
 
The following users thanked this post: Biduleohm

Offline Siwastaja

  • Super Contributor
  • ***
  • Posts: 9336
  • Country: fi
Re: BMS short circuit interruption capability
« Reply #14 on: March 26, 2020, 11:19:08 am »
2.4uH of inductance stores way too much energy at 10000A, but the idea is that you won't need to ever put 10000A through that 2.4uH! Note the current has a second exponent in the energy stored, so you need to really prevent the current from reaching the resistance-limited short circuit current; act well before that happens.

Electrolytic caps are cheap and small, it's trivial to store the energy stored in the inductance in the capacitors, unless someone adds a MASSIVE ferrite core in the system, but this wouldn't be expected.

E = 0.5 * L * I^2 = 0.5 * 2.4uH * (550A)^2 = 0.36J
E = 0.5 * C * U^2, solve for C:
C = E/(0.5*U^2). Allow for 20V voltage rise on the capacitor voltage,
C = 0.36J/(0.5*(20V)^2) = 1800uF

You can clamp the voltage to a capacitor bank with a diode even if the switch is bidirectional. The advantage is you may be able to find fast diodes better than extremely fast TVS diodes (I don't know if I'm correct here; you may able to find a directly suitable TVS, as well, in which case the diode-capacitor solution does not have such advantage), and you can discharge the capacitors as slowly as you want.

Note, 2.4uH, whatever this is coming from, is already quite a lot. Using https://www.eeweb.com/tools/loop-inductance , this would be a circular loop 75cm in diameter. In a properly designed battery pack and wiring, the current and the return current run closer to each other.

The idea is, do not design your active efuse circuit around the short circuit current of the battery, because it sees this current only if you have failed the design somehow, i.e., it won't be able to detect the overcurrent in time, and switch off the transistors in time - in which case it doesn't matter, it will likely blow up anyway. Instead, design your efuse circuit to prevent the current from ever rising much over the expected maximum operating current. Then, protect (the wire insulation and the battery pack, and other things thermal) against a design failure using a traditional passive fuse, and for that, make sure it handles the 10kA current.
« Last Edit: March 26, 2020, 11:51:16 am by Siwastaja »
 
The following users thanked this post: Biduleohm

Offline BiduleohmTopic starter

  • Contributor
  • Posts: 38
  • Country: fr
Re: BMS short circuit interruption capability
« Reply #15 on: March 26, 2020, 03:42:51 pm »
dI=dt*V/L
dI = 5us*64v/2.4uH = 133A

I didn't thought about using this formula this way, thanks ;)



2.4uH of inductance stores way too much energy at 10000A, but the idea is that you won't need to ever put 10000A through that 2.4uH! Note the current has a second exponent in the energy stored, so you need to really prevent the current from reaching the resistance-limited short circuit current; act well before that happens.

Yep, I had the feeling it would work that way but I needed to be sure.

Electrolytic caps are cheap and small, it's trivial to store the energy stored in the inductance in the capacitors, unless someone adds a MASSIVE ferrite core in the system, but this wouldn't be expected.

E = 0.5 * L * I^2 = 0.5 * 2.4uH * (550A)^2 = 0.36J
E = 0.5 * C * U^2, solve for C:
C = E/(0.5*U^2). Allow for 20V voltage rise on the capacitor voltage,
C = 0.36J/(0.5*(20V)^2) = 1800uF

You can clamp the voltage to a capacitor bank with a diode even if the switch is bidirectional. The advantage is you may be able to find fast diodes better than extremely fast TVS diodes (I don't know if I'm correct here; you may able to find a directly suitable TVS, as well, in which case the diode-capacitor solution does not have such advantage), and you can discharge the capacitors as slowly as you want.

I don't want to use electrolytics for longevity purposes. But now I know the energy level is much more reasonable I have a lot of other possible solutions so that shouldn't be a problem.

Note, 2.4uH, whatever this is coming from, is already quite a lot. Using https://www.eeweb.com/tools/loop-inductance , this would be a circular loop 75cm in diameter. In a properly designed battery pack and wiring, the current and the return current run closer to each other.

Yes, it's a defavorable case but I prefer to design for the worst case (and add some margin on top...) because I'm tired of the reverse in today's products (designed to just barely work, and so breaks at the first occasion), I want something that work, no matter what you do (as long as you stay in the limits of course), and which do it for years.

The idea is, do not design your active efuse circuit around the short circuit current of the battery, because it sees this current only if you have failed the design somehow, i.e., it won't be able to detect the overcurrent in time, and switch off the transistors in time - in which case it doesn't matter, it will likely blow up anyway. Instead, design your efuse circuit to prevent the current from ever rising much over the expected maximum operating current. Then, protect (the wire insulation and the battery pack, and other things thermal) against a design failure using a traditional passive fuse, and for that, make sure it handles the 10kA current.

Yep, obviously there will be a classic fuse too, but it would not have saved the mosfets (and at more than 3 € each that would get expensive quickly...) as it's way too slow.



@all Thanks everyone for the help ;)
 

Offline Siwastaja

  • Super Contributor
  • ***
  • Posts: 9336
  • Country: fi
Re: BMS short circuit interruption capability
« Reply #16 on: March 26, 2020, 05:13:51 pm »
Yes, a classical fuse cannot protect the MOSFETs, hence your MOSFETs need to protect themselves, in other words, you need good control for them. Good that you have a fuse; even the most perfect designer isn't perfect, and this problem is not trivial. The fuse cannot save the device, but it saves from a larger catastrophe. It must be non-user replaceable, so that once it's blown, the user must replace the device, which has failed short.

Nothing wrong in electrolytics if you don't require the stablest ESR in the world, and design for the ripple current spec (and derate it a bit). Actually, limiting yourself outside of electrolytics may lead to much more serious omissions on your part; large-value ceramics especially came with mechanical reliability problems and voltage spiking due to ESR too low, and actually combining smaller values of ceramics with electrolytics tends to solve these problems. Whatever you are doing likely requires some DC link capacitance, and with pure ceramic, you are likely seeing spikes 2x or even exceeding 2x the supply voltage. TVS damping is one option to limit the overshoot to somewhere about 50%, but electrolytics easily bring it down to near zero. After all, if you require low-loss low-ESR capacitance on your link, then you need preferably at least 2-3x the C with higher ESR. Doing that with another set of ceramics, with explicit series resistors, increase the BOM cost and area quite a bit more.

The fact you see electrolytics die early in SMPS power supplies only mean they are pushed over the limits there.
« Last Edit: March 26, 2020, 05:21:52 pm by Siwastaja »
 

Offline BiduleohmTopic starter

  • Contributor
  • Posts: 38
  • Country: fr
Re: BMS short circuit interruption capability
« Reply #17 on: March 26, 2020, 07:55:20 pm »
To detail a bit more: initially it's a personal project but I share it because I think others might be interested. It's not a project for beginners so anyone making/using it should be competent enough to do that. The fuse will not be integrated to this project, it's not its responsability nor mine, it's a "if you don't want to fuse your battery it's your problem" even if, of course, I'll put a clear warning and recommendation about fuse specs, etc...

I don't think I'll use electrolytic caps as my design must work reliably from -40 °C to +60 °C continuous for 15-20 years minimum, they are not space efficient, they are costly, they would add problems to solve (i.e. big current spike when the mosfets closes while the caps are charged). Also I'm curious about what electrolytic cap would handle a peak of hundred of amps?

Most likely I'll use some TVS or some active circuit based around zeners and transitors (didn't look about MOVs either, but I will as I'm curious if they can be helpful here). I may add a RC snubber network to help spread the energy over a longer duration and to help with EMI too, but I need to learn more about RC snubbers before.
 

Offline mzzj

  • Super Contributor
  • ***
  • Posts: 1298
  • Country: fi
Re: BMS short circuit interruption capability
« Reply #18 on: March 26, 2020, 08:55:45 pm »

I don't think I'll use electrolytic caps as my design must work reliably from -40 °C to +60 °C continuous for 15-20 years minimum, they are not space efficient, they are costly, they would add problems to solve (i.e. big current spike when the mosfets closes while the caps are charged). Also I'm curious about what electrolytic cap would handle a peak of hundred of amps?

Most likely I'll use some TVS or some active circuit based around zeners and transitors (didn't look about MOVs either, but I will as I'm curious if they can be helpful here). I may add a RC snubber network to help spread the energy over a longer duration and to help with EMI too, but I need to learn more about RC snubbers before.
You have already written it in stone that you don't want flywheel diodes and electrolytic caps but yet have no idea of even basics like RC snubbers and current rise over inductance?  ???

Cap bank to handle 550A current and some odd half joule energy is not much. Deciding factor is most likely ESR.  80v mosfets for 64v system don't leave too much margin but lets allow 3 volts for ESR @550A, that would require <5.5mOhm ESR.  10x 1000uF 100v electrolytics should cover that.

Or just keep your fingers crossed and let the mosfets avalanche.  :-/O :scared:
One is rated for 300A 10us avalance, 10x in parallel can probably hand 5x that ie 1500A for 10us.
 

Offline jbb

  • Super Contributor
  • ***
  • Posts: 1265
  • Country: nz
Re: BMS short circuit interruption capability
« Reply #19 on: March 26, 2020, 09:31:26 pm »
On electrolytics: it tends to be the heat which ages them.  But heat will wreck your battery pack much faster, so you should be able to get durable capacitors.

If there’s a chance of a vibrating environment (especially a vehicle!) you’ll need to pay careful attention to capacitor mounting. Otherwise they’ll vibrate, fatigue and fall off. When you provide this support, make sure you don’t block the emergency relief vents (typically a scribed pattern on the end of the capacitor tin) so you don’t make a little bomb.
 

Offline BiduleohmTopic starter

  • Contributor
  • Posts: 38
  • Country: fr
Re: BMS short circuit interruption capability
« Reply #20 on: March 26, 2020, 10:31:01 pm »
You have already written it in stone that you don't want flywheel diodes and electrolytic caps but yet have no idea of even basics like RC snubbers and current rise over inductance?  ???

Cap bank to handle 550A current and some odd half joule energy is not much. Deciding factor is most likely ESR.  80v mosfets for 64v system don't leave too much margin but lets allow 3 volts for ESR @550A, that would require <5.5mOhm ESR.  10x 1000uF 100v electrolytics should cover that.

Or just keep your fingers crossed and let the mosfets avalanche.  :-/O :scared:
One is rated for 300A 10us avalance, 10x in parallel can probably hand 5x that ie 1500A for 10us.

It's not that I don't want flywheel diodes, I just don't see how you can make that work with a bidirectional switch; if you have some schematic of how you make it work I'm all for it (it would be an almost perfect solution actually...). For the electrolytics caps I gave my reasons, if you think there's some that can fit the constraints, again, I'm open for suggestions.

I don't like maths too much and it's been a long time I used the inductance formulas so I forgot a good part of what I knew. For snubbers I know how they work, etc. I just need to refresh about how to calculate the values, that's all. You know, we can't be good in every part of every domain. I really try to not bother others with questions if I can learn by myself (on the whole project, which is 5 different boards BTW, I only asked for help 3 times and it was only after hours of research) but here it was a bit too much and it was on something I can't screw-up so I wanted to be sure.

The only 1000 µF 100 V cap I can find whith a low enough ESR (< 55 mOhms) is this one: https://www.mouser.fr/ProductDetail/Nichicon/UHW2A102MHD?qs=KdTYp8tcPgnQ6r7UMjz7nA%3D%3D for ten of them that would be 32 € (without tax which are 20 % here) which is more than 1/3 of the budget of the whole board, that's too expensive. Also they are 18*40 mm so 10 of them would need a lot of PCB real estate.

Yeah I'm not that savage ;D even if they should handle it actually (Eas = 0.85 J and there's 10 of them, so well over 0.36 J), I prefer to keep them out of avalanche mode. It's more like a last resort if the external protection fail to protect them.

On electrolytics: it tends to be the heat which ages them.  But heat will wreck your battery pack much faster, so you should be able to get durable capacitors.

If there’s a chance of a vibrating environment (especially a vehicle!) you’ll need to pay careful attention to capacitor mounting. Otherwise they’ll vibrate, fatigue and fall off. When you provide this support, make sure you don’t block the emergency relief vents (typically a scribed pattern on the end of the capacitor tin) so you don’t make a little bomb.

I didn't think about vibrations, but yeah, that's another constraint.



So now the worst case isn't when the inductance is maximum, but minimum. With 10 cm of wire that's 73 nH, so the current would be almost 4.7 kA for an energy of 0.81 J. Ok, that's still reasonable :phew:
« Last Edit: March 26, 2020, 10:33:20 pm by Biduleohm »
 

Offline Siwastaja

  • Super Contributor
  • ***
  • Posts: 9336
  • Country: fi
Re: BMS short circuit interruption capability
« Reply #21 on: March 27, 2020, 06:08:13 pm »
The idea with the electrolytic caps is that they do have ESR. Use them for damping the ceramic pack you use for low ESR. If you use 1*C of ceramic plus 2*C of ceramic with series R, the cost is likely 10x more than 1*C of ceramic and 5*C of electrolytic. If the low-ESR ceramic take most of the ripple current in the system, the electrolytics last forever (unless the environment is exceptionally hot, like over 100 deg C).

In any case, what is the device doing? Or is it just a stand-alone efuse module? Turns out in most complete applications (thinking about motor controller, most likely) the DC link capacitance is there already, and it's often a combination of low-ESR ceramic and high-ESR electrolytic, for damping.

But a beefy TVS (or a few in parallel) will likely do just fine catching the voltage spike, we got sidetracked into the capacitors.

Yes, precharging causes a current spike, but you can't just ignore the problem; if this is a module, it will likely supply a capacitor bank, and it's a PITA if the user needs to create another precharge circuit. Precharging should definitely be a feature in your thing, IMO. Otherwise it's just nuisance tripping on capacitive load, preventing operation.

You may indeed need to add a certain amount of inductance so you have a known minimum to work with so you can guarantee a current limit value (depending on your reaction speed). Note, from the MOSFET point of view only, the faster you switch, the higher the voltage spike would be (less snubbing action in MOSFET itself) - but OTOH, the slower you switch, the higher the current has time to rise to, hence you store more energy, and create a higher spike after all.

So switch quickly. Unless you have some strange requirement such as very low Iq comparator, you can easily detect the overcurrent in, say, 500ns, then switch the MOSFETs in 1us.

I actually designed such efuse circuit two years ago, and the extra challange there was the requirement for extremely low quiescent current - it doubled as a power switch - in order not to overdischarge the battery pack, even in the most demanding case - pack is ran flat, then the product is put in storage without recharging. It was a nightmarish or extremely fun design challenge (whichever way you look at it) involving precision resistor networks and extremely low Iq comparators - which are not that fast - but in the end it worked out. I did some heavy Excel magic with combining all the component tolerances.
« Last Edit: March 27, 2020, 06:23:59 pm by Siwastaja »
 
The following users thanked this post: Biduleohm

Offline BiduleohmTopic starter

  • Contributor
  • Posts: 38
  • Country: fr
Re: BMS short circuit interruption capability
« Reply #22 on: March 27, 2020, 06:50:06 pm »
Ok, I'll see for the ceramic/electrolytic mix, it makes sense.

No, there's no external cap bank, it's for residential and mobile solar installations.

There's actually a precharge feature (look just below the main disconnect on the schematic) but that's to precharge external capacitive loads like the large caps in the inverters for example. Here we're talking about caps placed across the mosfets which would be naturally charged if there's any load connected while the mosfets are open, and when the mosfets close they would short-circuit the caps. Edit: wait a second... now I see, the precharge would also discharge the caps between P- and B-, that's nice, one less problem :)

Yep, but instead of adding inductance (not practical) I can specify a minimum wire length under which the mosfets aren't guaranted to be protected. 10 cm is a very defavorable case and not accounting for internal battery inductance (inductance of the cells and of the busbars between the cells), I just wanted to see what to expect in the very worst case possible and I now know that the energy stay pretty constant and under a Joule, but the current rises quickly when the inductance is reduced.

Yes, it would not make sense here to switch slower as the gain in dI/dt will be far outweighed by the energy I need to dissipate. So yep, I'll keep the switch quickly design.

The mosfets should switch under 300 ns, the mosfet driver add 160 ns of propagation delay, the comparator (TSX3704IDT) to detect the over-current condition adds 2 µs and the amplifier (OPA4990IDR configured as a diff amp with a gain of 41) for the current shunt adds 2 µs too. That's worst-case for all the values, typically it should be quicker than that.

And yes, Iq is a concern as the BMS is connected to the battery 24/7, I tried my best to keep everything as low power consumption as possible. Which, as you experienced too, is a bit of a PITA when selecting components (you know the "I found a very nice component for this" followed by "ah, it would need 10 mA, hmm yeah, I'll find something else..." ::)), add to this a lot of BoM re-use to keep the costs as low as possible and you have even more compromises to make...
« Last Edit: March 27, 2020, 07:32:43 pm by Biduleohm »
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf