Tritium_James wrote: I wouldn't necessarily call them errors - yes, it means you need to use less efficient 600V parts instead of 200V parts. The long gate drive traces mean you need slow switching edges which increase your switching losses. But it means you can make it low cost and relatively easy to put together. It's "good enough" for most purposes, and they've sold thousands of them. I'd say that's quite a good design.
I would think the long gate drive paths CAUSE slow switching speeds rather than requiring them.
Curtis have been building these things for years and years and it seems they still don't use a micro - and the things that go with it like temp and current sensing, etc as mentioned elsewhere here. And I wouldn't call the LS thing "low cost" - its about $US1900. The labour content alone, resulting from its really poor mechanical design, must be horrendous. Its taken me hours and hours to get it apart and I still haven't sliced off all the silicone, let alone unsoldered all the blown up bits.
I've had a good look at the Cougar design and it is decades ahead of these things! I am very tempted to chuck the LS controller board altogether and replace it with a Cougar one. But I will get the thing running again first - and then maybe stick it on eBay!
Nevilleh, I'm not sure that your 'wires' gate drive idea will actually help much. Don't forget the other half of that circuit is the 'ground' return, it's important to equalise/minimise the length of this too.
The problem here is that if you ran an equal number of ground wires back to the controller (ideally as a twisted pair with each gate drive wire) then they will probably melt and die quite quickly - ground at one end of the power stage is quite likely to be several volts different from ground at the other end, and if you're not careful this might get quite a large current flowing in your gate drive 'ground' wires. This is a difficult problem to solve without a complete redesign of the entire layout.
If I were you I'd still give your discrete wire gate drive idea a go anyway, because it might help. But don't be too surprised if it doesn't.
Yes, you're quite right about the ground path. In this case, the entire bottom of the power board is covered with a ground plane to which each and every IGBT emitter is connected. It is probably as good as it can get and its not possible to make any changes to it anyway.
When I get it repaired I'll set it up on the bench with a fairly low voltage power supply and a small motor and have a look at the gate drive signals. That will give me a comparison for when I try the wires.
Its quite interesting, actually!
I would have bought one, but they had ceased production when I ordered my controller. As it turned out, it took me so long to get the car built that they had started up again and I could have bought one then. But I already had the LS.
However, what a learning experience! I know more about dc motor controllers now than I ever really wanted to.
I think that this thing will work OK once I have rebuilt it - and hopefully improved it.
All the bits are here now and I have already replaced all the semiconductors on the controller board. I'll fire it up from a bench power supply shortly and see what it does. Watch this space
thats exactly what happened to me, Zilla stopped taking orders so i got a LS controller,now it blown up!! the time it took me to build the car Zilla was back up and running,we should have waited but thats life!
avolt wrote: thats exactly what happened to me, Zilla stopped taking orders so i got a LS controller,now it blown up!! the time it took me to build the car Zilla was back up and running,we should have waited but thats life!
Neville, if you try the wired gate idea I suggest you cut the original long track immediately after the last IGBT in each group. Having a long track going nowhere is very likely to introduce reflected waves and ringing.
Yes, thanks for pointing that out.
It looks like I can use 150mm of wire for the gate connections and if I use 1mm dia it has an inductance of around 170 nH. Using the gate capacitance (worst case) of 2500 pF, critical damping occurs with R=16.5 ohms. If I have 20 IGBTs in parallel and my driver can source 6A with a 15V supply, the min gate resistor should be (15/6) x 20 = 50 ohms. If I use 56 ohms then my RLC (56 ohms, 170 nH and 2500 pF) circuit is well over damped and behaves more like an RC than a RLC with a time constant of 140 nS.
The turn-on speed should be no faster than the Trr of the flyback diodes and they are about 60 nS worst case. Thus the min gate resistor value still gives turn-on speeds slower than the diode Trr, so all should be well.
I note that the gate resistors used in the present LS board are 200 ohms which would give a turn on time nearly 4 times slower.
Any comments?
Edit: Further reading tels me my idea of using another gate driver chip is not good for IGBTs in parallel as it introduces another variation in gate drive timing.
Last edited by Nevilleh on Wed, 07 Apr 2010, 03:18, edited 1 time in total.
Assuming those RLC values are correct I agree with your calculations. Are the existing gate resistors really 200 ohms and not 20 ohms? Did you put a meter on them?
You could go down to 47 ohms since the inductance delays the peak current for about 11 ns by which time the gate capacitance has charged to about 1 volt and so there is only 14 V across the resistor and therefore 0.3 A per gate (x 20 = 6 A total). I simulated it in a spreadsheet.
Can you tape the gate wires against the ground-plane?
One of the fathers of MeXy the electric MX-5, along with Coulomb and Newton (Jeff Owen).
Yeah, I put a meter on them. I was surprised o find them so large, so I double checked. I can't read resistor colour codes any more anyway. a) my eyesight isn't good enough and b) they put so many bands on the things now I can't figure out which end is which!
The wires have to cross over the Al busbars, so they need to sit up a bit.
OK, 47 ohms is better than 56. I wonder how to calculate the power rating?
My spreadsheet says the 47 ohm resistor generates about 300 nJ of heat on every edge (rising or falling). As a sanity-check: That's equivalent to pulling 0.3 A through it for 70 ns. So multiply this by twice the maximum switching frequency and you'll have its dissipation in watts. e.g. if it switched at 10 kHz this would be 10 kHz * 2 * 300 nJ = 6 mW
so a 1/4 watt will be fine.
One of the fathers of MeXy the electric MX-5, along with Coulomb and Newton (Jeff Owen).
The average might be 6mW, but peak is almost 5W (15V²/47). You should use a resistor with a specified pulse rating. Most el-cheapo 1/4W types won't provide this type of info in their datasheet.
An excellent point James. One that I was not aware of (except purely on the grounds of voltage rating). Neville, there's a good description and sample datasheet here. http://www.cs.cmu.edu/~ram/art/essay/components/
Scroll down to the section entitled "Missing Specifications".
It sounds like a 1/4 watt would still be OK, but only barely in the case of metal-film. And if you can't buy a brand with a pulse spec then you'd be safer with a carbon composition type rather than carbon film or metal film.
One of the fathers of MeXy the electric MX-5, along with Coulomb and Newton (Jeff Owen).
Yes, a very good point indeed, I shall investigate further!
I have just powered up the controller board and - Lo! - 15V pulses emerge from the fet driver chip. And I can vary them with the throttle pot. Progress indeed.
But it is going to take a wee while to rebuild the power board now, what with all the blasted silicone rubber and general hardness-to-get-at of everything. I want to put it back together as it was and check out the gate drive signals before modifying it in any way, so I can see the effect of my mods.
Every piece of Si on the controller board except a TL431 was blown, even some 1N4148s used as simple logic gates!
I couldn't find an LM317L which provides a 4.7V reference for the throttle pot circuitry, so I have shoved in a 78L05 temporarily and it works OK. I'll change it later.
Internet research indicates that carbon composition resistors are what is needed as they typically quote overload (pulse) ratings of at least 10 times. I have ordered some 1/2 watt ones from Farnell (58 cents each, mind you) and they should be able to handle 5 watt peak pulses. They are also non-inductive. The worst case being say 15V across 47 ohms is only 4.8w. Should be good!
By the way, the controller pulse frequency is 13.66 kHz. Way above my hearing range!
I've had a good look at the circuitry and I can't see any form of current sensing at all. The only thing there is, is a thermistor which is connected to an op-amp and shown on the Curtis circuit as "Main Current Limit". I surmise that it only limits current if it starts to get too hot. But on the LS controller, the thermistor is replaced by a 15K resistor, so it doesn't even have that. Is it surprising that they blow up?
I wonder if a4x4kiwi has delved into this at all?
Last edited by Nevilleh on Thu, 08 Apr 2010, 06:28, edited 1 time in total.
a4x4kiwi wrote: Here ... Reverse polarity protection all round!
That's pretty impressive. Any ideas on how it's achieved?
Surely it's not a 1000 A diode in series with one of the the battery terminals. That diode would get seriously hot and would limit performance.
And surely also not a 1000 A fuse with a 2000+ A diode after it protecting against reverse connection.
Surely also not a 1000 A IGBT or MOSFET in series with the positive terminal, that isn't turned on if the polarity is wrong.
Perhaps it's part of an inbuilt pre-charge circuit. So there is a 1000 A contactor in there that isn't turned on to bypass the precharge resistor unless the polarity is correct, and the precharge current isn't enough to fry the main silicon. Logic circuits would have other protections, including a series diode.
MG ZS EV 2021 April 2021. Nissan Leaf 2012 with new battery May 2019.
5650 W solar, 2xPIP-4048MS inverters, 16 kWh battery.
Patching PIP-4048/5048 inverter-chargers.
If you appreciate my work, you can buy me a coffee.