Progress on the MX5 conversion

Technical discussion on converting internal combustion to electric
User avatar
acmotor
Senior Member
Posts: 3595
Joined: Thu, 26 Apr 2007, 03:30
Real Name: Tuarn
Location: Perth,Australia

Progress on the MX5 conversion

Post by acmotor » Thu, 17 Jul 2008, 18:01

Ian,
Just like a front wheel drive on a wet road. Its all fine 'till it lets go ! Controlled power slide is not an option.

Realistically, can you state a current limit you will be content with ?

I thought so... Image

iMiEV MY12     105,131km in pure Electric and loving it !

juk
Groupie
Posts: 179
Joined: Sun, 02 Mar 2008, 02:53
Real Name: Craig
Location: Perth

Progress on the MX5 conversion

Post by juk » Thu, 17 Jul 2008, 22:03

More front wheel drive bashing. I raced a peugeot 205 gti against a series 4 rx7 on the Targa stage called 'The Sideling' during moderate rain. The stage was uphill for about 7 km, followed by downhill for the same distance. Both cars started level. By the top the 205 had a lead of about 1km.

Why? because the front wheel drive could safely apply the power much sooner than the rx7, only risking understeer. I understeered out of every corner.

You'll be pleased to know that the RX7 killed me on the whole stage, because the first seven kms going uphill in the rain with an ambient temperature of about 15 deg C i cooked the brakes on the 205 even though it's running the same brakes as the 405 and only weighs 880kg. Goes to show how quick i was actually going.

User avatar
acmotor
Senior Member
Posts: 3595
Joined: Thu, 26 Apr 2007, 03:30
Real Name: Tuarn
Location: Perth,Australia

Progress on the MX5 conversion

Post by acmotor » Fri, 18 Jul 2008, 06:00

That stirred the nest !!! Image

Granted the 205 GTI is good. But then that is no ordinary car and many front wheel drives can only dream of being that well balanced.
I drove one around the UK and know them well (I used to be in the Peugeot car club, owned two Pugslies - still have the T shirt)
Personally I did find the torque steer of the 205 took a bit of getting used to. Not complaining, just commenting.

Now, what about a 205 EV conversion ? Image
iMiEV MY12     105,131km in pure Electric and loving it !

zeva
Senior Member
Posts: 422
Joined: Sat, 15 Dec 2007, 05:09
Real Name: Ian Hooper
Location: Australia
MSN: sigmunky@hotmail.com
Contact:

Progress on the MX5 conversion

Post by zeva » Fri, 18 Jul 2008, 06:13

Yeah, go the 205 GTi conversion! That would be a great little EV!

And I'll do an RX7 conversion, and we'll see who can complete The Sideling quickest. Sounds to me like that RX7 driver just lacked the necessary cojones if he couldn't handle his RWD in wet weather Image

(I'm just being combative now, feel free to ignore me!)
Ian Hooper
--
"Never doubt that the work of a small group of thoughtful, committed citizens can change the world." - Margaret Mead
http://www.zeva.com.au

juk
Groupie
Posts: 179
Joined: Sun, 02 Mar 2008, 02:53
Real Name: Craig
Location: Perth

Progress on the MX5 conversion

Post by juk » Fri, 18 Jul 2008, 07:10

Actually, he is the better driver. The two cars whilst having totally different personalities were dead even in the dry on the same roads. (allowing for the 205's brakes shortcoming).

I've got the 205 sitting around at my old mans place getting in the way. It's tempting to convert it. It has brand new brakes, but will need to be upgraded if i convert it, because it will be raced. Autokhana would really suit it. I autokhana'd it once before, was getting 3 seconds a lap on a VT clubsport, the clubsport owner was not happy!

User avatar
MikeG
Groupie
Posts: 97
Joined: Thu, 12 Jun 2008, 14:44
Real Name: Mike Glew
Location: Canberra ACT
MSN: michaelglew@hotmail.com
Contact:

Progress on the MX5 conversion

Post by MikeG » Fri, 18 Jul 2008, 13:52

I've been thinking of converting a Pulsar SSS (N15), as I used to circuit race one of these years ago and it was a nice platform (but admittedly FWD, which is good for power but bad for power delivery & grip).

I thought about converting my current car, a 200SX (S15) but given the low km I do in that car it would be a waste of time.

Audis use an Aluminium chassis right? I wonder if an A3 conversion would be a good idea... Lightweight is the key...

If I could find something lightweight, FWD, 4-door with good size boot space, this would be a good option for a touring EV if you could get enough battery power aboard.

is it true that Lithium Ion batteries need to be adjacent to eachother???
Last edited by MikeG on Fri, 18 Jul 2008, 03:54, edited 1 time in total.

zeva
Senior Member
Posts: 422
Joined: Sat, 15 Dec 2007, 05:09
Real Name: Ian Hooper
Location: Australia
MSN: sigmunky@hotmail.com
Contact:

Progress on the MX5 conversion

Post by zeva » Fri, 18 Jul 2008, 16:35

Lithium batteries, no they don't have to all be adjacent to each other (I have a split pack in my MX5) though if it's possible to keep the pack in one block, it is better.

Partly the BMS can get confused if its signal wires have to travel too far between packs (in an electrically noisy environment), and partly it's just more effort installing two separate packs and connecting them with large power cables!
Ian Hooper
--
"Never doubt that the work of a small group of thoughtful, committed citizens can change the world." - Margaret Mead
http://www.zeva.com.au

zeva
Senior Member
Posts: 422
Joined: Sat, 15 Dec 2007, 05:09
Real Name: Ian Hooper
Location: Australia
MSN: sigmunky@hotmail.com
Contact:

Progress on the MX5 conversion

Post by zeva » Tue, 22 Jul 2008, 04:49

So I took apart the Curtis today to see what went wrong. For anyone interested in a bit of power electronics and controller design, here's a report on my findings:

http://zeva.com.au/tech/curtis

If anyone from Curtis is reading this.. Please don't sue me for bagging your product - direct your energy to designing us a better controller instead Image
Ian Hooper
--
"Never doubt that the work of a small group of thoughtful, committed citizens can change the world." - Margaret Mead
http://www.zeva.com.au

User avatar
Kearon
Groupie
Posts: 50
Joined: Tue, 10 Jun 2008, 21:07
Real Name: Kearon de Clouet
Location: Sydney
Contact:

Progress on the MX5 conversion

Post by Kearon » Tue, 22 Jul 2008, 04:57

Onya Hoops - Top work!
(I don't think you can really get lawyer-trouble by pointing out obvious design flaws in a 20 year old product...I reckon you're safe!!)

Kearon de Clouet: www.evcapri.com
current conversion: 1990 Ford Capri : 45 x TS90Ah : Impulse 9

zeva
Senior Member
Posts: 422
Joined: Sat, 15 Dec 2007, 05:09
Real Name: Ian Hooper
Location: Australia
MSN: sigmunky@hotmail.com
Contact:

Progress on the MX5 conversion

Post by zeva » Tue, 22 Jul 2008, 05:07

Yeah, also I'm not worth much (ahem, financially I mean) so there wouldn't be much point suing me. As they say, you can't get blood from a stone! Image
Ian Hooper
--
"Never doubt that the work of a small group of thoughtful, committed citizens can change the world." - Margaret Mead
http://www.zeva.com.au

User avatar
acmotor
Senior Member
Posts: 3595
Joined: Thu, 26 Apr 2007, 03:30
Real Name: Tuarn
Location: Perth,Australia

Progress on the MX5 conversion

Post by acmotor » Tue, 22 Jul 2008, 08:27

Hoops,

I visited your link. Well done to post this sort of info up. Image

It would be intersting to know if the Mosfets blew from over-current or voltage punch through. At the time of coming to a halt the full battery voltage would have been across the mosfet.

You describe " pulling up to a stop it gave a brief surge then went dead". To me this would be typical of a single mosfet failing to a short as a result of a voltage punch through (possibly of a hot and vulnerable mosfet silicon juntion). It then passed a short burst of current, too much for its little legs to handle and moved over to the dark side !
The controller no longer worked because this little display of metal vapour took out the gate drive circuit to all mosfets.

If the case of the controller was at 80deg C then the mosfet junctions would have been at 200deg C and very prone to punch through.

I don't buy the the PWM signal propagation theory. Electrons used to travel at 3 x 10^8 m/s when I went to school.
More likely I would be guessing that the mosfet that failed is at the connections end of the bus bar where the bus bar is hottest and the impedance to the realworld connection the lowest. So that mosfet gets hottest and works the hardest. Not rocket science.
I wonder if infact the inductance of an 11" motor is more than a 9". Any data on this ? I would expect the lack of inductance more an issue as the di/dt is higher and more of an issue for the current limiting of the Curtis to handle.
Here is an interesting example of this point...
www.itee.uq.edu.au/~aupec/aupec99/walker99.pdf

It would be the very low resistance of the 11" that would cause sharp current rise in PWM, that is likely not well handled by the elderly curtis analogue circuit as you say.

Your layout ideas on the Zeva controller address much of this problem
with attention to symetry on all connections to parallel mosfets (including stray capacitances)
What current rating would you target for your controller ?
I see 1200A as a never to go near, still to be derated for temperature, rating, so maybe 600A ?
Isn't this a bit light on for the MX5 ?

We all await the product !   Image

iMiEV MY12     105,131km in pure Electric and loving it !

zeva
Senior Member
Posts: 422
Joined: Sat, 15 Dec 2007, 05:09
Real Name: Ian Hooper
Location: Australia
MSN: sigmunky@hotmail.com
Contact:

Progress on the MX5 conversion

Post by zeva » Tue, 22 Jul 2008, 08:56

Hi Tuarn,

Yep, you're right, the difference in time taken for the electrons to travel to the other end of the case would be insignificant. But I'm sure I've heard that the distant FETs switch every so slightly later than the closer FETs.. Perhaps the slight difference in resistance between driver and gate due to longer PCB tracks results in slightly lower current and hence slightly slower gate charging?

Anyway, on second thoughts (and further discussion with the EVDL crowd) it would indeed seem that lower impedance is the problem with the Curtis controllers, particularly driving NetGain motors which have fewer, thicker windings in their armature (lower resistance but also lower inductance). The Curtis would have been designed and tested with Advanced DC motors, which have thinner windings and hence higher inductance.

The surging before death does suggest punch through, and there's plenty of damage evident which suggests the PWM signal lines may have been damaged with it. Further investigation needed.

This first controller I'm working on will officially be rated to 500 battery amps, though motor side current should be OK up to about 750A though for a bit more bottom end torque. That should be enough for most vehicles as a substitute for 1231Cs, but won't keep me happy indefinitely in the MX5! So yes, I do plan to work on a 1000A version next. But it's also gonna need more powerful batteries.

(Hmm I really should go to bed..)
Ian Hooper
--
"Never doubt that the work of a small group of thoughtful, committed citizens can change the world." - Margaret Mead
http://www.zeva.com.au

zeva
Senior Member
Posts: 422
Joined: Sat, 15 Dec 2007, 05:09
Real Name: Ian Hooper
Location: Australia
MSN: sigmunky@hotmail.com
Contact:

Progress on the MX5 conversion

Post by zeva » Tue, 22 Jul 2008, 17:17

Here's the explanation on the likely cause of failure from the inimitable Lee Hart, as posted on the EV Tech list, seems to make sense:

The Curtis 1231 has enough trouble running a 9" motor; I'd say an 11" is
just too big for it. Direct drive puts even more load on the controller.
You really need a bigger controller!

I don't think the MOSFETs die from the single gate driver and relatively
slow turn-on; these were design choices Curtis made for economy and to
reduce RFI. The switching time is still fast enough so that conduction
losses dominate switching losses. What killed the MOSFETs was *heat*!

In a direct drive EV, you run the motor at high currents and low duty
cycles. This generates a lot more heat in the MOSFETs. Worse, it
generates a lot more heat in the *diodes*. The Curtis controllers have
relatively weak freewheel diodes.

Some numbers: The 1231C has 18 IXTH50N20 MOSFETs (rated 200v 50a 0.045
ohm) and 18 SR4180R diodes (special; probably a selected MR2402F, rated
200v 24a 200nsec). Let's say you are pulling 500 motor amps, and the
duty cycle is 50%:

MOSFETs:
     R = 0.045ohms/18 = 0.0025 ohms
     V = IR = 500a x 0.0025ohms = 1.25v
     P(peak) = 1.25v x 500a = 625 watts peak
     P(ave) = 0.5 x 625w = 312 watts average (50% duty cycle)
diodes:
     V = 1v at 24a (spec for MR2402F)
     I = 500a/18 = 27.7a per diode
     P(peak) = 1v x 500a = 500 watts peak
     P(ave) = (1-0.5) x 500w = 250 watts average (50% duty cycle)

Note that diode losses are almost as much as MOSFET losses. And the
diodes are a smaller part!

Here is what I suggest your failure mode was: When you started from a
stop, the duty cycle was very low; say 10% on. That meant the MOSFETs
carried the 500a 10% of the time (62.5w average), and the diodes carried
it 90% of the time (450w average). The high diode power killed some
diode; it failed shorted. An instant later, the MOSFETs turned on; they
saw a short circuit, so they died, too. Now you had a dead short across
your battery, and things *really* started to happen fast! Parts
exploded, traces vaporized off the PC board. From the amount of carbon,
there was probably a fire.

(PS: Did you have a fuse that blew, or did you have to wait for the
carnage to burn itself out?)
Ian Hooper
--
"Never doubt that the work of a small group of thoughtful, committed citizens can change the world." - Margaret Mead
http://www.zeva.com.au

User avatar
Thalass
Senior Member
Posts: 741
Joined: Sun, 12 Aug 2007, 07:29
Real Name: Ben Rypstra
Location: Perth, WA, AU

Progress on the MX5 conversion

Post by Thalass » Tue, 22 Jul 2008, 18:59

Damn. And I'm going to be in that neck of the woods in a couple of hours, too! [:p ]


Oh well, good luck with the controller! It's great to see the MX-5 back together again!
I'll drive an electric vehicle one day.

User avatar
acmotor
Senior Member
Posts: 3595
Joined: Thu, 26 Apr 2007, 03:30
Real Name: Tuarn
Location: Perth,Australia

Progress on the MX5 conversion

Post by acmotor » Tue, 22 Jul 2008, 19:10

Lee is on to it. (almost)

I would comment though that the problem with power electronics is that is does not perform simply to the calculations as Lee suggests. Practical layout means that you are wise to halve all expectations of voltage current and power and de-rate for temperature.
... and the more parallel power devices you use the more you must de-rate as well.

Lee's calculation that...
MOSFETs:
     R = 0.045ohms/18 = 0.0025 ohms dosen't work out in practice
The load is never evenly divided.

BTW the back emf (catch) diodes do not conduct 90% of the time when mosfets are off so his calcs are wrong there. If the inductance of the 11" is lower then the diodes work less than with a 9".
The mosfets in a hot box with non-symetric feed are most likely to fail (the one closest the output terminal).

Hoops, check the diodes (out of circuit)





iMiEV MY12     105,131km in pure Electric and loving it !

munkeevegetable
Noobie
Posts: 1
Joined: Sun, 02 Nov 2008, 17:52
Real Name: Nathan Buck
Location: Melbourne, AU

Progress on the MX5 conversion

Post by munkeevegetable » Sun, 02 Nov 2008, 18:55

Hey Ian,

any news since completion? I've been following this closely and very impressed with the end result. How is your ZEVA controller going?

I've been planning a conversion of a '64 Triumph Spitfire, and your conversion served largely as an inspiration. I'm guessing, unfortunately, the prices of everything are about to (or already have) skyrocketed. So I'm hoping it doesn't hit you/other importers of EV parts too hard.

--
Nathan

zeva
Senior Member
Posts: 422
Joined: Sat, 15 Dec 2007, 05:09
Real Name: Ian Hooper
Location: Australia
MSN: sigmunky@hotmail.com
Contact:

Progress on the MX5 conversion

Post by zeva » Mon, 03 Nov 2008, 03:41

Hi Nathan et al,

Not much to speak of since original completion; sadly it's still sitting there waiting for a controller. I've had a Zilla on order for a few months now so it should be here soon, that'll be the final solution.

I had a prototype of my own design in the car a few weeks ago but it went up in smoke! (Just the controller, not the whole car.) My fault, I had the current sensor resolution out by a factor of 2 in the microprocessor code, and consequently fried the MOSFETs. It did go like a bat out of hell for a short time though! I've got a new prototype on the bench almost finished now (a better design, and the current limit is definitely correct!) so it might be driveable again soon. Or I might blow some more MOSFETs.. That's R&D for you Image

And yes, the global economic crisis and collapse of the exchange rate has really been hurting the cost of EV parts. It suddenly costs 30% more to do a conversion, e.g $20K instead of $15K - big difference! I think we're going to see a bit of a slowdown in the industry for a while.
Ian Hooper
--
"Never doubt that the work of a small group of thoughtful, committed citizens can change the world." - Margaret Mead
http://www.zeva.com.au

User avatar
mattW
Groupie
Posts: 32
Joined: Sat, 01 Mar 2008, 07:07
Real Name: Matt
Location: Sydney
Contact:

Progress on the MX5 conversion

Post by mattW » Mon, 03 Nov 2008, 22:29

Sounds like the perfect time to start exporting controllers back to the states.. Image

User avatar
Hatchet
Groupie
Posts: 44
Joined: Mon, 16 Jun 2008, 02:49
Real Name: Brett Bishop
Location: Brisbane
Contact:

Progress on the MX5 conversion

Post by Hatchet » Tue, 04 Nov 2008, 03:40


Keep going with the controller Ian. I suspect that there are more than a few out there waiting for your Australian designed and made controller. I know I am as I want more than a Curtiss but less than a Zilla.

Brett

antiscab
Senior Member
Posts: 2476
Joined: Mon, 26 Nov 2007, 05:39
Real Name: Matthew Lacey
Location: Perth, WA

Progress on the MX5 conversion

Post by antiscab » Tue, 04 Nov 2008, 04:33

lol....so something with faster current limit?
and maybe a big fat "this motor is too big u dolt" light?

loving the exporting controllers idea though, keep it up Ian

Matt
Matt
2017 Renault zoe - 25'000km
2007 vectrix - 156'000km
1998 prius - needs Batt
1999 Prius - needs batt
2000 prius - has 200 x headway 38120 cells

tassie_tiger
Groupie
Posts: 119
Joined: Mon, 16 Jun 2008, 14:58
Real Name: Andrew
Contact:

Progress on the MX5 conversion

Post by tassie_tiger » Tue, 04 Nov 2008, 15:56

Last time I looked cafe electric had no buyers of their business so there is an opening in the market for a new player.

I know if I were in Ian's position, I would be trying my hardest and spending much time on getting it to market!

bga
Senior Member
Posts: 492
Joined: Mon, 01 Sep 2008, 19:27
Real Name: Bruce Armstrong
Location: Perth WA

Progress on the MX5 conversion

Post by bga » Tue, 04 Nov 2008, 17:49

Off topic, but may be relevant:

The long Curtis bus bars may be part of the problem.
I would expect that current sharing in this design would
be poor at high currents. I haven't done the numbers,
but is likely that the transistors on the output end are
working a lot harder (twice+?) than those at the backwater end.
(low RdsOn transistors will make this worse!)

This might be detectable if the PCB has been cooked, as well as
blackened, providing the escaping smoke hasn't erased the
evidence. (Thinks discolouration and delamination from
poorly selected power resistors)

I've been investigating IGBT drivers for a while now.
These makers of these devices usually recommend using voltages of
-8V to +15V, the -8V is used to prevent inductive/capacitive effects
tripping the junction during switching.
Similar issues with FETS?
[effect name needed]

A lot of current may be needed on the gate driver.
I would not be surprised to see 15 Amps needed for that many
MOSFETs. Slow turn on/off may be a problem.

Also, most IGBT drivers contain desaturation logic that detects
the 'on' junction voltage and generates a 'trip' if it exceeds
approximately 4 volts a few microseconds after turn on.
This might be a good feature on a MOSFET controller.

Cheers

BGA
Last edited by bga on Tue, 04 Nov 2008, 06:52, edited 1 time in total.

pazs13
Noobie
Posts: 1
Joined: Thu, 20 Nov 2008, 19:32
Real Name: Paris
Location: Sydney

Progress on the MX5 conversion

Post by pazs13 » Thu, 20 Nov 2008, 19:40

Hello all

new member, this is really great what you are doing. I now have a place where i can seriously start my conversion and ask some questions along the way good luck with the conversion

Paz
Why drink and drive when you can drift and slide

User avatar
Electrocycle
Senior Member
Posts: 985
Joined: Sun, 19 Oct 2008, 20:23
Real Name: Andrew
Location: Sydney
MSN: dumhed@dumhed.com
Contact:

Progress on the MX5 conversion

Post by Electrocycle » Thu, 20 Nov 2008, 20:12

having also looked at a dead Curtis controller, I'd definitely be blaming the diodes.
Most of the failures seem to happen at low speed, as did the one I have.
This puts the blame on the diodes rather than the FETs, as mentioned in a post above, and in this case all the FETs were intact, but one of the diodes had blown parts and tracks off the "brain" board nearby!

Compared to smaller motor controllers (for radio control etc), the Curtis is very basic in its FET drive setup, so the turn on / off times will be fairly slow - which at least prevents the ringing problems you can get when you try to switch the FETs really fast!

The Engine Whisperer - fixer of things

drowe67
Groupie
Posts: 147
Joined: Fri, 18 Apr 2008, 20:14
Real Name: David Rowe
Location: Adelaide
Contact:

Progress on the MX5 conversion

Post by drowe67 » Fri, 21 Nov 2008, 21:10

zeva wrote: Yep, you're right, the difference in time taken for the electrons to travel to the other end of the case would be insignificant. But I'm sure I've heard that the distant FETs switch every so slightly later than the closer FETs.. Perhaps the slight difference in resistance between driver and gate due to longer PCB tracks results in slightly lower current and hence slightly slower gate charging?
The inductance of the signal path to the FET gates will affect switching time, and may even cause other problems like ringing of the gate signal. IIRC even a straight piece of wire has 10nH per cm. This will block rapid changes in gate current over long distances, causing FETs nearer to the driver to switch first.

Under the right conditions you can get problems on high speed digital hardware over just 10-20cm (e.g. fast rise times, daisy chain connection to multiple gates, high impedance paths for the currents).

While the clock frequencies are low for a controller, the physical dimensions, number of gates being driven, and currents are all large. I imagine the gate rise times are fast and it must be tricky to maintain good low impedance paths for currents (both to the FET gates and return via ground) given the physical size and shape of the FETs.

Ian - your approach of keeping the FETs close to their drivers, and multiple drivers sounds good. Minimising the loop area the current must flow to the FET gates and back is also useful; for example a specific ground track from each FET back to it's driver, laid right next to the gate signal on the PCB, or in a twisted pair if using wires.

However clearly the Curtis works just fine for many people (including me), so maybe it's gate drive design is adequate for the job. It look forward to having an AU built and serviced controller at a lower price, but I have no complaints about my Curtis. Also a 20 year production life is something to be proud of for any electronic product.

Cheers,

David
Driving an EV every day
http://www.rowetel.com/ev.html

Post Reply