Weber and Coulomb's MX-5

Post up a thread for your EV. Progress pics, description and assorted alliteration
User avatar
Johny
Senior Member
Posts: 3749
Joined: Mon, 23 Jun 2008, 16:26
Real Name: John Wright
Location: Melbourne
Contact:

Weber and Coulomb's MX-5

Post by Johny »

Maybe you could mount a cheap thermistor on the link and use your spare analogue input to monitor link temperature. No voltage or high impedance worries then. The link temperature would soar if there were a real fault.
User avatar
weber
Site Admin
Posts: 2936
Joined: Fri, 23 Jan 2009, 17:27
Real Name: Dave Keenan
Location: Brisbane
Contact:

Weber and Coulomb's MX-5

Post by weber »

Thanks Woody and Johny. We already have temperature monitoring of the link, at the negative terminal of every cell. The MSP430 microcontroller has a built in temperature sensor and a thermal-conduction pad. So we put it as close as we could to the negative terminal and put as much copper between the two as we could. That's why I'm not too worried about losing this feature.

We could possibly eliminate the noise problem by using shielded cable. But the other reason I'm not worried about losing the feature is that I don't think it can ever actually detect a high-R terminal-to-link joint.

If you look at the following sketch which shows how the link sense is connected, you will see that because the BMUs sit on top of the link, while the cell terminal is under the link, any voltage drop caused by a high-R terminal-to-link joint will in fact appear as a change in one or other cell voltage, and will not be measured by the link sense circuitry at all. The only thing it can measure is the resistance of the link itself, and being a continuous piece of copper, it is highly unlikely to change. Pity I didn't realise this earlier. But hey, that's 7 resistors we can leave off the boards in future. Hoorah.

Image
Last edited by weber on Wed, 15 May 2013, 14:34, edited 1 time in total.
One of the fathers of MeXy the electric MX-5, along with Coulomb and Newton (Jeff Owen).
User avatar
coulomb
Site Admin
Posts: 6357
Joined: Thu, 22 Jan 2009, 20:32
Real Name: Mike Van Emmerik
Location: Brisbane
Contact:

Weber and Coulomb's MX-5

Post by coulomb »

Interesting. But I think if the strap goes HR to the cell terminal, there is a reasonable chance that it will also be HR to the bolt, so the BMU will read the correct link voltage. But that's just a wild, unsubstantiated guess.

The bolt is relatively high resistance, but compared to a higher-than-normal strap to terminal connection, I don't know. So even if the strap is HR to the terminal but still a good connection to the bolt, I think the BMU will still see a substantial increase in link voltage. Also unsubstantiated.
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.
User avatar
weber
Site Admin
Posts: 2936
Joined: Fri, 23 Jan 2009, 17:27
Real Name: Dave Keenan
Location: Brisbane
Contact:

Weber and Coulomb's MX-5

Post by weber »

The strap is not even in contact with the bolt, except it may be accidentally resting against the thread a bit. Other than that accidental contact, any current between strap and bolt must pass through vias in the BMU. So the strap will be HR to the bolt at all times, not only when it is HR to the cell terminal. By "HR" here we mean anything more than about 3 mR. And I note that stainless steel has about 40 times the resistivity of copper.

Whether the BMS measures a strap-to-terminal voltage drop as a cell voltage change or a link voltage change (or a little of both) depends on the relative values of two resistances you didn't mention -- BMU-to-strap and BMU-to-cell-terminal. BMU-to-strap is direct while BMU-to-cell-terminal goes via two stainless washers and the stainless bolt (four contacts). Yes it is possible that the 4-contact stainless path from BMU to cell-terminal could be lower resistance than the direct path from BMU to copper strap, if say the copper was badly tarnished there despite the grey goo. So I shouldn't have been so definite about the outcome. But it remains quite possible to have a high strap to terminal voltage drop that doesn't register as "link voltage". In the event of a loose bolt, gravity favours this scenario (better contact between BMU and strap than BMU and bolt).
One of the fathers of MeXy the electric MX-5, along with Coulomb and Newton (Jeff Owen).
User avatar
coulomb
Site Admin
Posts: 6357
Joined: Thu, 22 Jan 2009, 20:32
Real Name: Mike Van Emmerik
Location: Brisbane
Contact:

Weber and Coulomb's MX-5

Post by coulomb »

I was thinking that the BMU will likely have a good contact with the cell terminal through the bolt; "good" in the sense of very low voltage drop at less than the micro-amp of measurement current, plus the milliamps of current for the BMU's circuitry.

But as you point out, I was neglecting how this very low voltage can be "pushed around" by the current through the strap, which can be hundreds of thousands of milliamps. The current will divide according to the ratio of strap-to-BMU resistance and the BMU to terminal resistance, I think.

3mR would be a pretty bad strap resistance, doubling the effective internal resistance of that cell. It's hard to say what the bolt resistance would be. Comparable, I guess. If the same, the link voltage measurement would be half the actual, which would still be useful, I think.
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.
TooQik
Groupie
Posts: 179
Joined: Mon, 18 Jun 2012, 00:53
Real Name: Chris
Location: Mount Gambier

Weber and Coulomb's MX-5

Post by TooQik »

Just thinking out loud here, and apologies if this has already been asked or highlights a lack of knowledge on my behalf, but is there any reason that the BMU can't go under the strap rather than on top of it?
User avatar
weber
Site Admin
Posts: 2936
Joined: Fri, 23 Jan 2009, 17:27
Real Name: Dave Keenan
Location: Brisbane
Contact:

Weber and Coulomb's MX-5

Post by weber »

coulomb wrote: I was thinking that the BMU will likely have a good contact with the cell terminal through the bolt; "good" in the sense of very low voltage drop at less than the micro-amp of measurement current, plus the milliamps of current for the BMU's circuitry.
Milliamp singular, yes. So compared to the contact resistances, we can consider the BMU (including its remote sensing connection at the positive terminal of the previous BMU) to be a pair of high-impedance voltmeters which share a connection at the BMU's negative terminal.
But as you point out, I was neglecting how this very low voltage can be "pushed around" by the current through the strap, which can be hundreds of thousands of milliamps. The current will divide according to the ratio of strap-to-BMU resistance and the BMU to terminal resistance, I think.
I'm not sure what current you're talking about being divided here. Your most recent reference is to the hundreds of amps of traction current, but the division you describe only makes any kind of sense to me if you're referring to the earlier milliamp or microamps. I suspect we're actually in agreement, but what I would have said is that the hundreds of millivolts (caused by the hundreds of amps of traction current flowing through the few milliohms of poor strap-to-cell-terminal resistance) will be divided between the aforementioned two voltmeters according to the ratio of strap-to-BMU resistance and BMU-to-cell-terminal resistance.

So we have a triangle or delta of resistances. Even if it was in fact a star (which I don't think it is) then we know every star has an equivalent delta.
3mR would be a pretty bad strap resistance, doubling the effective internal resistance of that cell. It's hard to say what the bolt resistance would be. Comparable, I guess. If the same, the link voltage measurement would be half the actual, which would still be useful, I think.

I wasn't suggesting 3 mR of strap resistance. The 3 mR was an example of a bad strap-to-cell-terminal resistance. I was assuming the resistance of the strap was negligible. I now confirm this by calculating it to be about 13 uR (micro-ohms) (3 x 20 mm area, 46 mm long annealed copper). I calculate the conducting part of the bolt to be about 250 uR (5 mm dia x 7 mm long stainless steel). 100 A thru 13 uR gives 1.3 mV. 100 A thru 250 uA would give 25 mV. So we can even ignore the resistance of the bolt because all 3 resistances are dominated by contact resistance.

To confirm this, according to
http://viewmold.com/Products/Raymond-SP ... ashers.pdf
the loading due to our 0.6 mm thick M6 stainless belleville washer should be about 600 N. And according to
http://www.copperinfo.co.uk/busbars/pub ... s/sec7.htm
that should give us a contact resistance from copper strap to copper cell terminal of about 50 uR. I note that the area has very little to do with it. The resistance of a copper-copper joint is pretty much inversely proportional to the force, at about 3 milliohm newtons. But in case you want to know, the area of the strap-to-cell-terminal contact is about 85 mm^2.

The kind of fault we were hoping to detect with our so-called link voltage sensing would almost certainly be caused by a loss of force compressing the joint, due to either the bolt coming loose due to vibration or more likely plastic creep of the PCB.

It would be reasonable to assume that all contact resistances would increase proportionally as the force was reduced, so it really all comes down to the ratio of one copper strap to tinned-copper BMU contact versus the sum of the four contacts: tinned-copper BMU to stainless flat washer to stainless belleville washer to stainless bolt to copper or aluminium cell terminal.

You're probably aware that, like aluminium, the resistance to corrosion of stainless steel is due to a tenacious oxide layer. In the case of stainless steel it is a chromium oxide layer a few tens of nanometers thick. It turns out that it takes about 300 mV to break down this layer, and until you get to 300 mV the contact resistance is about 7,000 times that of a tinned copper contact with the same force. See this article
http://www.te.com/documentation/whitepa ... 3jot_9.pdf

And since we have multiple stainless and half-stainless contacts in series we will probably need more than 900 mV to break them down and so we would have found the bad joint by its heat before then. 900 mV * 100 A = 90 watts.

So with at least a 20,000 to 1 ratio for this voltage divider, we really will only be reading the voltage across the strap, whose resistance will not change with joint clamping force.

TooQik, if the BMU terminals were solid copper covering the full area of the cell terminal, then you might get away with clamping them between the strap and the cell terminal, but you would be doubling the contact resistance. Then we would expect to measure about half the voltage of a bad joint with our link voltage sensing.

But our BMU terminals, like most, are just the usual tinned copper on both sides of fibre-glass/epoxy printed circuit board material, with some plated thru vias connecting the two sides, and you really don't want to try pulling 100 amps through a few PCB vias, or through a bunch of stainless-washer-and-bolt contacts, as I've shown above.
Last edited by weber on Thu, 16 May 2013, 11:53, edited 1 time in total.
One of the fathers of MeXy the electric MX-5, along with Coulomb and Newton (Jeff Owen).
User avatar
weber
Site Admin
Posts: 2936
Joined: Fri, 23 Jan 2009, 17:27
Real Name: Dave Keenan
Location: Brisbane
Contact:

Weber and Coulomb's MX-5

Post by weber »

We've been mostly writing software on recent EV days. Unfortunately software is not very photogenic.

Today we sucessfully tested the "limp mode" or "turtle mode" that Coulomb and I wrote yesterday, for when things go wrong. While I was driving, my passenger (Mark Aylott) disconnected a BMS optic fibre and the power limit (battery current limit) ramped smoothly down from 100% to 5% in a little over 10 seconds. This should be enough time to realise something has gone wrong and to get off the road. Although if you're only using 30% power at the time, you won't notice anything happening until the last 3 seconds. The same will happen if a cell goes below 2.5 volts and stays there, or above 50 degC and stays there.

You can continue to use the 5% power indefinitely, assuming it is still there to be had, but of course you would only use it to move further out of danger since it may be doing damage to cells. We settled on a minimum power level that would still allow us to climb a 10% grade at 15 km/h.

P = mgh/t = 1500 kg * 9.8 N/kg * 15 km * 10% / 3600 s = 6.1 kW

It equates to about a 0.3C discharge rate from the cells.
One of the fathers of MeXy the electric MX-5, along with Coulomb and Newton (Jeff Owen).
User avatar
coulomb
Site Admin
Posts: 6357
Joined: Thu, 22 Jan 2009, 20:32
Real Name: Mike Van Emmerik
Location: Brisbane
Contact:

Weber and Coulomb's MX-5

Post by coulomb »

weber wrote: Although if you're only using 30% power at the time, you won't notice anything happening until the last 3 seconds.

Good point. We should flash a dashboard icon, perhaps the battery icon, when starting limp mode. Or I suppose we could somehow add our own turtle icon somewhere. I think it needs flashing, to get the driver's attention.

I wonder if we should also turn on the indicator when stress limits the power to a certain level, say 60% or less. That way, if a cell is touching the low voltage limit or high temperature limit, it will flash on briefly (longer for over temperature), and you are alert to the fact that power is limited, so don't overtake and consider moving to the slow lane.

So maybe this: turn on the indicator steady at say 60% power limit or lower, and start flashing it at say 25% power or less (including limp mode). This is similar to many ICE cars, which turn on a fuel icon when the tank is say 10%, and start flashing it when lower, maybe 5%.

There may be a legal requirement to flash an indicator slowly; my Subaru actually PWMs it to smoothly cycle the intensity over about a 5 second cycle. Though blinker indicators flash abruptly and fairly quickly, so that may not be needed. Maybe ICE cars consider low fuel to be a relatively low priority indication, and reserve quickly flashing icons for higher priority situations.
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.
User avatar
weber
Site Admin
Posts: 2936
Joined: Fri, 23 Jan 2009, 17:27
Real Name: Dave Keenan
Location: Brisbane
Contact:

Weber and Coulomb's MX-5

Post by weber »

coulomb wrote:We should flash a dashboard icon, perhaps the battery icon, when starting limp mode. Or I suppose we could somehow add our own turtle icon somewhere. I think it needs flashing, to get the driver's attention.
An excellent suggestion Dr Coulomb. I wouldn't bother with a turtle icon. I think the existing "CHARGE" light with its battery icon will serve admirably. And the authors of NCOP14 agree that if the warning is not audible it must be by means of a flashing lamp.
NCOP14 wrote:2.14 Battery Management

For series strings of batteries, some form of charge or balance management should be implemented. The necessity of this requirement will be dependent on the battery chemistry and technology used in the vehicle.

This is especially critical with lithium chemistry batteries which must be maintained within strict upper and lower voltage limits and upper temperature limits. Some form of device to monitor these limits on each individual cell or group of parallel cells should be present.

If a monitoring device is fitted, the monitoring device must be capable, of either audibly or visually by means of a flashing lamp, warning the driver of an impending disconnect with sufficient time for the driver to safely park the vehicle before disconnection occurs.
coulomb wrote:So maybe this: turn on the indicator steady at say 60% power limit or lower, and start flashing it at say 25% power or less (including limp mode).
Another excellent suggestion. I note that we also have a redundant oil pressure gauge which migh be made to display battery stress level by means of one of the four "Gauge" outputs from the Tritium DCU.
There may be a legal requirement to flash an indicator slowly;
In a quick search, I found only that a range of 1 to 4 Hz is specified for flashing warning lamps in SAE J595. And certainly no requirement for sinusoidal flashing.
One of the fathers of MeXy the electric MX-5, along with Coulomb and Newton (Jeff Owen).
VRLAme
Noobie
Posts: 3
Joined: Tue, 15 Sep 2009, 22:16
Real Name: Brendan Lee
Location: Brisbane

Weber and Coulomb's MX-5

Post by VRLAme »

Could you use the temperature or oil gauge in the instrument cluster to show a max temperature? This will give you a pretty good idea straight away of the condition of the pack. You could also pulse the fuel gauge between a representation of the voltage of the pack, and a representation of the voltage of the minimum cell multiplied by the total number of cells. If the needle is holding pretty steady, it's a good indication that all cells are healthy. If the needle is pulsing from one value to another, it shows that somewhere there is a cell with a lower voltage.
Sorry guys, I don't often comment, but I am avidly following your build. I hope you can take your BMS further than just your own system.
User avatar
weber
Site Admin
Posts: 2936
Joined: Fri, 23 Jan 2009, 17:27
Real Name: Dave Keenan
Location: Brisbane
Contact:

Weber and Coulomb's MX-5

Post by weber »

VRLAme wrote: Could you use the temperature or oil gauge in the instrument cluster to show a max temperature? This will give you a pretty good idea straight away of the condition of the pack. You could also pulse the fuel gauge between a representation of the voltage of the pack, and a representation of the voltage of the minimum cell multiplied by the total number of cells. If the needle is holding pretty steady, it's a good indication that all cells are healthy. If the needle is pulsing from one value to another, it shows that somewhere there is a cell with a lower voltage.
Sorry guys, I don't often comment, but I am avidly following your build. I hope you can take your BMS further than just your own system.

Thanks VRLAme. All good suggestions. I particularly like the pulsing needle idea, although we would want to use the fuel gauge to represent state-of-charge rather than voltage. However we do have a temperature gauge as well as an oil pressure gauge.

But the way our BMS works is that everything is optimised (within the low cost requirement) to get the information out as quickly as possible, about any kind of stress to any cell, so something can be done about it as quickly as possible. To that end, we transmit status 15 times a second, as a single byte, through the 9600 bit/s chain of 109 BMUs, with this structure:

Bit 7 Always 1 -- to indicate a status byte, not command or response text.
Bit 6 Comms error -- 1 if this status byte did not originate from cell 1.
Bit 5 All_near_bypass -- 1 means charging is complete.
Bit 4 Stress check bit -- inverse of bit 3.
Bits 0-3 Stress level 0..15 -- >7 is distress, >10 is alarm.

As this byte passes down the line, each cell will substitute its own stress if this happens to be worse than what's already there.

With only 4 bits for the stress level we don't bother distinguishing between stress due to temperature or stress due to voltage at this point. Every cell logs its own worst stress level (since this was last cleared) and the reason for that stress level -- i.e. whether it was due to voltage or temperature and what that voltage or temperature actually was. So after the drive, we can issue say an "8q" command, which means "query all cells with worst stress of 8 or more", to get their cell numbers and their worst voltage or temperature.

At one stage, while driving, we also had the BMS master interrogate every cell in turn, for its actual voltage, and display the highest and lowest voltages and the cell numbers they belonged to. But this operates about 100 times slower than status byte sending and can delay the transmission of status bytes.

So we figure that displaying stress level is as good as anything. The rate of change of stress with respect to time and power level will be a good indicator of whether it is due to temperature or voltage. Voltage changes instantly, temperature only slowly. We will probably use the temperature gauge to display the worst of motor and controller temperatures. So that leaves the oil pressure gauge for cell stress.

I'll take this opportunity to link to a post in another thread where I discuss a foolish thing I did with 16 of the MX-5's cells a couple of years ago. Trying to look on the bright side -- at least it gave us some cells that were easily stressed, so we could test our control loop. Image
viewtopic.php?title=some-donts-for-ev-b ... 542#p42906
One of the fathers of MeXy the electric MX-5, along with Coulomb and Newton (Jeff Owen).
User avatar
weber
Site Admin
Posts: 2936
Joined: Fri, 23 Jan 2009, 17:27
Real Name: Dave Keenan
Location: Brisbane
Contact:

Weber and Coulomb's MX-5

Post by weber »

When using the oil pressure gauge to display cell stress, it's kinda cute that, in engineering, the units for stress are the same as the units for pressure. But it turns out that our cell stress scale has more in common with the SUD scale in psychology. Multiplying SUD numbers (0 to 10) in the above-linked Wikipedia article by 1.5 to obtain our 0 to 15 scale, and adding a little poetic license, gives the following.

0.0 = Peace, serenity, total relief. No more anxiety of any kind about any temperature or voltage.
1.5 = No acute distress and feeling basically good. If you took special effort you might feel something unpleasant but not much.
3.0 = A little bit upset, but not noticeable unless you took care to pay attention to your voltage and then realize, "yes" there is something bothering me.
4.5 = Mildly upset. Worried, bothered to the point that you notice it.
6.0 = Somewhat upset to the point that you cannot easily ignore an unpleasant temperature. You can handle it OK but don't feel good.
7.5 = Moderately upset, uncomfortable. Unpleasant voltages are still manageable with some effort.
9.0 = Feeling bad to the point that you begin to think something ought to be done about the way you feel.
10.5 = Starting to freak out, on the edge of some definitely bad voltages. You can maintain control with difficulty.
12.0 = Freaking out. The beginning of alienation.
13.5 = Feeling desperate. What most cells call a 15 is actually a 13.5. Feeling extremely freaked out to the point that it almost feels unbearable and you are getting scared of what you might do. Feeling very, very bad, losing control of your pressure relief valve.
15.0 = Feels unbearably bad, beside yourself, out of control as in an electrolyte breakdown, overwhelmed, at the end of your cable. You may feel so upset that you don't want to transmit because you can't imagine how anyone could possibly understand your agitation.


So now you understand how your cells feel. Image
One of the fathers of MeXy the electric MX-5, along with Coulomb and Newton (Jeff Owen).
User avatar
weber
Site Admin
Posts: 2936
Joined: Fri, 23 Jan 2009, 17:27
Real Name: Dave Keenan
Location: Brisbane
Contact:

Weber and Coulomb's MX-5

Post by weber »

Here's another piece of mysterious bracketry for the MX-5 from the mind of Jeff Owen. He and his wife dropped by on Saturday afternoon. Jeff came up with the elegant design in the time it took me to make them a cup of tea. And I spent the next several hours fabricating it. I just baked the last coat of paint in the oven.

Image

Tomorrow I'll install it and take another photo so you can see what it's for.
One of the fathers of MeXy the electric MX-5, along with Coulomb and Newton (Jeff Owen).
User avatar
weber
Site Admin
Posts: 2936
Joined: Fri, 23 Jan 2009, 17:27
Real Name: Dave Keenan
Location: Brisbane
Contact:

Weber and Coulomb's MX-5

Post by weber »

Here's what the bracket is for. Mounting the twin DC-DC converters in the absence of the large under-bonnet battery box they used to be mounted to, facing the other way. The cavernous expanse in this photo can be compared to the "wife's suitcase" photo in this post, where that battery box is in place. viewtopic.php?title=weber-and-coulombs- ... 062#p42062

Image

As well as showing the back of the DC-DC bracket, the following photo shows another bracket, designed and fabricated by Newton (Jeff Owen), for mounting the grey ABS 12 volt junction box that is currently in progress and being held out of the way.

Image

We agreed to trailer the MX-ϟ to the Logan Eco Action Festival this Sunday and will be leaving out the bonnet box so folks can see the motor. But more generally, we have the option of more nimble cornering, at the cost of having only half the power and half the range, by leaving out the entire "B" half-pack, which consists of the bonnet box (54 cells), rollbar box (19 cells) and under-boot box (36 cells).
One of the fathers of MeXy the electric MX-5, along with Coulomb and Newton (Jeff Owen).
User avatar
weber
Site Admin
Posts: 2936
Joined: Fri, 23 Jan 2009, 17:27
Real Name: Dave Keenan
Location: Brisbane
Contact:

Weber and Coulomb's MX-5

Post by weber »

Someone sent us this grainy night-time photo alleged to be of us driving the MX-ϟ, in an attempt to blackmail us.

Image

So I'm getting in first to deny that that's the MX-ϟ, and if it is, it's not us doing it, and anyway we can just adapt. (Oh sorry, got confused with the Liberal party position on global warming there).

It must be our evil twins Maxwell and Faraday.

Here's the note that was sent with the photo.

Image
One of the fathers of MeXy the electric MX-5, along with Coulomb and Newton (Jeff Owen).
User avatar
weber
Site Admin
Posts: 2936
Joined: Fri, 23 Jan 2009, 17:27
Real Name: Dave Keenan
Location: Brisbane
Contact:

Weber and Coulomb's MX-5

Post by weber »

Here are some photos from the Logan Eco Action Festival last Sunday. Many thanks to Coulomb, Jeff Owen and Mark Aylott for helping get the MX-5 ship-shape for the festival. And thanks to Graeme Manietta for getting it there and back on his car trailer.

Image

The littlest driver of the day.

Image
Last edited by weber on Fri, 07 Jun 2013, 19:52, edited 1 time in total.
One of the fathers of MeXy the electric MX-5, along with Coulomb and Newton (Jeff Owen).
User avatar
coulomb
Site Admin
Posts: 6357
Joined: Thu, 22 Jan 2009, 20:32
Real Name: Mike Van Emmerik
Location: Brisbane
Contact:

Weber and Coulomb's MX-5

Post by coulomb »

Last EV day we decided it was important to sort out some battery box issues. So as usual, we got distracted by something else, and it didn't get done. Well, the car was on the ground for once, and it needed to go up on the stands again for the battery work, and we needed to test the power steering pump controller with the wheels on the ground. It does too make sense!   Image

The DC motor controller for the power steering pump arrived a week ago:

Image

Weber specifically picked out this one from Ebay, because it said it runs at 30 kHz, and the power steering pump is noisy enough as it is. 30 amps should be adequate to run the steering pump, we hoped, and it stated a current limit of 35 A.

Imagine our delight when it was current limiting at 5 amps (battery side), and just stopped with any turn of the wheel. It also PWM'd at 11 kHz. Nice.

So we do what usually has to be done when you buy something cheap off Ebay - we traced out the circuit and fixed it:

Image

Right click and "view image" (or similar) for a clearer circuit diagram.

The thing with IC3 and R2 is so cheap and nasty. But we looked up the specs of the 78L05 and found that its quiescent current is typically 3 mA, which gives 11 V output (who knows what is intended), and the quiescent current is supposed to vary only slightly with temperature, load, and so on. So we left that part as it was. There is a non-populated resistor, R1, from the common to the input (not output, as expected).

While three MOSFETs are populated, there is provision for five. So the current limit could be extended further if required. You can see below that the current sense resistors (R5 and R17) are just short pieces of thickish wire. I guess it works, and it is definitely cheap. It's a genuine fibreglass PCB; single sided, with one link. The power stage layout is not great, and we saw moderate oscillations when the MOSFETs turn off. The amplitude is unlikely to cause problems.

The first two changes we made were to double the PWM frequency to 22 kHz and improve the rise time, and the last was to increase the current limit such that we get up to 40 A of motor current. (The current limit appears to be set by the ratio of R23 to R14; pin 8 seems to be an op-amp output.) Impressively, Weber had the three needed parts on his shelf. They are common parts, which you could buy at Jaycar for well under 50 cents.

The MOSFETs are 75 V and 75 A rated, and don't seem to get too hot. In fact, the power steering pump motor gets hotter than the controller, so we'll have to ensure reasonable air flow to the PS motor.

Here is a view inside the unit:

Image

Weber could not find a PWM chip that has an output at pin one, so we don't know what the PWM chip actually is. Are there any hardware types out there that can guess at what it is? Just for curiosity.

In summary, at US$43 plus shipping, this is a reasonable buy, after some simple modifications. It's a shame that the modifications are needed, of course. Hopefully, this may help others to get power steering working in their conversions, or there may be other applications.

[ Edit: I forgot to mention that the mechanical mounting leaves something to be desired. The two capacitors are so tall that they rub against the lid; we placed two small pieces of tape to prevent shorts. The bar of metal that the MOSFETs and diodes are mounted on ends up in a rounded part of the extrusion, so it doesn't make good contact with the case. Weber had to file the metal bar to fit better. He also needed to apply thermal compound, since there wasn't any as supplied. Image ]
Last edited by coulomb on Sun, 09 Jun 2013, 04:44, edited 1 time in total.
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.
User avatar
jonescg
Senior Member
Posts: 4721
Joined: Thu, 21 Jan 2010, 23:05
Real Name: Chris Jones
Location: Perth, WA.
Contact:

Weber and Coulomb's MX-5

Post by jonescg »

To save me reading through the thread again, how are you connecting the pot of the DC motor controller to the steering such that it powers up when you turn the wheel?

AEVA National President, WA branch director.
neilg
Groupie
Posts: 38
Joined: Sat, 07 Apr 2012, 06:24
Real Name: Neil Gibbs
Location: Perth

Weber and Coulomb's MX-5

Post by neilg »

Coulomb

Had a look at the Motor speed controller and found the following issues.

1. There is no ground connection on the chip.
2. The cct diagram could be confusing by using the ^ to mean +~11v and it being upside down to mean GND - suggest using +11 instead of ^.
3. Are you sure the +~11 connects to pin3 and not pin 4?

For this type of cheap controller I would expect a simple quad opamp would be used eg LM324/LM348 (or even a quad comparator ie LM339). These have outputs on pins 1,7,8 & 14 with +ve on 4 and GND on 11. (LM339 has open collector outputs hence the need for R4)
Sometimes you can use a 10* magnifier and look at the chip surface at different angles and makeout the device type.

Can you take other photos of the pcb.
1. With the underside illuminated so we can better see the tracks below.
2. The underside.

Neil
User avatar
weber
Site Admin
Posts: 2936
Joined: Fri, 23 Jan 2009, 17:27
Real Name: Dave Keenan
Location: Brisbane
Contact:

Weber and Coulomb's MX-5

Post by weber »

jonescg wrote:To save me reading through the thread again, how are you connecting the pot of the DC motor controller to the steering such that it powers up when you turn the wheel?
We aren't. We just want to run the hydraulic pump motor (that replaces the original belt-driven pump) at a lower constant voltage than 12 V, to get the noise down, and we want it to current-limit so it doesn't blow a fuse every time we hit full lock. It is on whenever the traction motor controller is enabled, so it's doing double duty as a fake idle noise (along with the vacuum pump and the water pump for the WaveSculptor). The hydraulic pump motor is working fine now at 7.5 V, pulling 10 A with hands off the steering wheel. When turning the steering wheel it pulls more amps at 7.5 V and when held against the stops it stalls and pulls 40 A at 4 V.
neilg wrote:Had a look at the Motor speed controller and found the following issues.

1. There is no ground connection on the chip.
Sorry Neil. Pins 10 and 12 are both connected to "triangle ground" (the one referenced to the MOSFET sources, not the external one on the green wire).
2. The cct diagram could be confusing by using the ^ to mean +~11v and it being upside down to mean GND - suggest using +11 instead of ^.
Are you posting from the international space station? Image I would expect most people to easily distinguish up from down.
3. Are you sure the +~11 connects to pin3 and not pin 4?
Quite sure. Here's a track-side photo. Sorry I can't provide a bottom-lit top-side photo. It's all back together and bolted into the car now.

Image Image
For this type of cheap controller I would expect a simple quad opamp would be used eg LM324/LM348 (or even a quad comparator ie LM339). These have outputs on pins 1,7,8 & 14 with +ve on 4 and GND on 11. (LM339 has open collector outputs hence the need for R4)
Sometimes you can use a 10* magnifier and look at the chip surface at different angles and makeout the device type.
The part number is actually ground off, not merely blacked out as Coulomb wrote. But I too had figured out that it must be an LM339 quad comparator. Well done. They have Vcc on 3, GND on 12 and outputs are 1, 2, 13 & 14. (Not the same pinout as the quad op amps). But then I had the advantage of knowing which pins were grounded and being sure pin 3 was Vcc.

Image

[Edit: The manufacturer went to the trouble of grinding the part number off, but now a complete schematic of the device is up on the web, purely because they didn't supply the item as advertised. Specifically it was only 11 kHz PWM, not 30 kHz. The original current limit may have been 35 A into a short circuit. We didn't test that. But output voltage began to droop way before the advertised 30 amps.]
Last edited by weber on Sat, 08 Jun 2013, 22:10, edited 1 time in total.
One of the fathers of MeXy the electric MX-5, along with Coulomb and Newton (Jeff Owen).
User avatar
coulomb
Site Admin
Posts: 6357
Joined: Thu, 22 Jan 2009, 20:32
Real Name: Mike Van Emmerik
Location: Brisbane
Contact:

Weber and Coulomb's MX-5

Post by coulomb »

So pin 8 is an input then. So I think there must be an error in the circuit diagram. Our enthusiasm to get it right evaporated about the same time we had it working as we wanted.

I hereby grant world-wide, non-exclusive and royalty-free permission for anyone interested to modify my diagram to show the comparators, and correct any errors. Image
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.
User avatar
coulomb
Site Admin
Posts: 6357
Joined: Thu, 22 Jan 2009, 20:32
Real Name: Mike Van Emmerik
Location: Brisbane
Contact:

Weber and Coulomb's MX-5

Post by coulomb »

Where does the time go in an EV conversion? Often times, in places and in quantities that you would not believe when starting off.

We've had all the battery boxes finished for months now, and we've had most of the "A" half-pack installed for months, and the car has been drivable like that. Most of the fiddly little things, with the exception of the heater and chargers, are actually done.

Lucifer must have noticed this, and decided to send down some lightning bolts to throw us off target. We had an "incident" with a routine software download, which has been working for years, and which we've even been patting ourselves on the back about how good it's been. (Maybe that was our mistake Image). The nature of this incident is such that we had to take the battery boxes out, remove the lids, and individually JTAG program each microprocessor. As luck would have it(or poor design, but bad luck sounds better), the JTAG connections are at the end of the BMU boards, so one end of each string of cells (one battery box has 3 strings for just 15 cells) is under a strip of runner that is siliconed in place, so that had to be laboriously removed and replaced. Some of the battery boxes take about an hour to remove and replace. The whole idea is that this should never happen, so when it did, we had to figure out why it happened, and how to really prevent it from happening again.

That's taken about 3 weeks to write, test, and implement. The solution, which we may talk about in more detail in another post, involves two separate bootstrap loaders. That's for now; one of those should really go away, but it's simpler and maaaybe more robust to leave it that way for now. But at least, we told ourselves, it could have been worse: we could have had to take out the two worst battery boxes (for removing and re-installing); these are in the fuel tank space right in the middle of the car. By some miracle, these weren't involved in the original incident, and their LEDs were still flashing away happily.

We developed a technique for doing the repair and upgrade, and were quite confident that when we came to upgrade the software in the fuel tank battery boxes (not because they had seized like the others, but because they needed to use the new software scheme) that there would be no problems.

Indeed, the front box (which isn't too hard to get at, if as now the rollbar battery box is removed), upgraded without a hitch. But old Lucifer had one last lightning bolt for us. When we went to upgrade the rearmost battery box, the hardest one to get to, there was no echo from the front of the string of BMUs to the end. There was status information being transmitted ~ 15 times per second, but characters typed at the first BMU didn't make it to the last one. At first, we hoped that it was just that we were using the wrong pair of optic fibres, but no. When we sent a command that should turn on all the blue lights, we could tell with a mirror and looking at the right angle from inside the boot that the lights came on up to a certain point, then stopped. Shining a torch in there and manipulating the mirror, we could see that there was a break between squiggle-joined strings of cells; these are bridged with three wires that are soldered (formerly we used connectors, but these proved unreliable), and of course the wires were siliconed to strain relieve them. So you'd expect that once connection was made (these BMUs have been working for weeks at least), that they'd never come apart.

Since taking that battery box out is such a lot of work, we tried taking off the lid in-situ, and using the ~ 10 mm of clearance above the battery box to try to do something with the wires. It was a desperate move, and of course it failed. There was no hope of re-soldering the wires in-situ, and besides Weber could not believe that it could possibly be the wires that was at fault.

So even these last two battery boxes had to come out!

Image

Things are a bit cluttered in that photo, so I'll take a moment to point out a few things. This is the back of the car; at the bottom right you can see the boot lid raised. That's Weber's hand in the top left, resting on the top of the passenger's seat, inclined forward. The white sheet is protecting the convertible roof, in retracted position. Weber assures me that as horrible as it is to remove these two battery boxes, removing the soft top roof is infinitely worse. So it stayed in place and we worked around it carefully. The two battery boxes are stacked on top of each other, resting on some planks of wood, resting on the centre console area that covers the gearbox. I don't know whether it had ever had to hold up that much weight, and I'm pretty sure it was never designed to do that. But it held for us. The infamous "16 mm" spanner can be seen in the middle at the right end of the photo. The various bits of wood are all part of the voodoo that we need to get the battery boxes out and back in again, with the soft top, two contactor boxes, and a variety of electrical wires and optic fibres in the way. The netbook computer is the one we use for communicating with the BMUs; in the top left of the screen is the Windows application for sending the binary images with the password characters at the start and the checksum at the end, and appropriate timing of the bytes in between. Not visible in this shot is the engine lifter that we use to lift the heavy battery boxes (just under 50 kg each, those two).

You can see that we have the lid off the top box. It was immediately obvious that one of the wires was indeed not connecting properly, and somehow must only have been connecting via a strand of copper. (My personal theory is that Lucifer got his finder under there, but I suppose that's just blaming the unseen for our sloppy workmanship.) It was a two minute job to repair the wire, test, and re-silicone the wires.

We must have been doing this way too many times, because it seems like we spent "only" about an hour pulling those boxes out and putting them back in. Elapsed time was much greater, because we were distracted by something else, which will no doubt be the subject of another post in a week or so.

[ Edit: added "the hand of Weber" ]
Last edited by coulomb on Sun, 21 Jul 2013, 15:39, edited 1 time in total.
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.
User avatar
weber
Site Admin
Posts: 2936
Joined: Fri, 23 Jan 2009, 17:27
Real Name: Dave Keenan
Location: Brisbane
Contact:

Weber and Coulomb's MX-5

Post by weber »

The "something else" that distracted us, that Coulomb mentioned in the previous post, was "Mrs Weber"'s new Nissan Leaf and its charging arrangements, shown here.

While Coulomb and I were working on reinstalling battery boxes, Jeff Owen was heat-forming the beautiful clear-polycarbonate mud shield you can see on the end of this front-mudguard battery box, and a mirror-image one for the other side.

It is held on with a combination of the M8 lid bolt and black automotive exterior-grade double-sided tape.

Image
One of the fathers of MeXy the electric MX-5, along with Coulomb and Newton (Jeff Owen).
User avatar
Johny
Senior Member
Posts: 3749
Joined: Mon, 23 Jun 2008, 16:26
Real Name: John Wright
Location: Melbourne
Contact:

Weber and Coulomb's MX-5

Post by Johny »

weber wrote:It is held on with a combination of the M8 lid bolt and black automotive exterior-grade double-sided tape.
Where does one purchase "black automotive exterior-grade double-sided tape"?
Post Reply