Page 2 of 3

Re: Cal for dccd

Posted: Tue Jun 26, 2012 5:54 pm
by pavlo
There is obviously more leakage current than that, because the physical effect is very real.

The output typically floats up to around 8volts when off, I haven't put a current meter on it yet as I've only really seen this recently.

The first time on Andy.Fs car at the pod, the ignition needed to be turned off to turn the car in the pits (pushing) as the ignition was turned off the transmission windup disappeared with a clunk. Since then the same issue has been reported elsewhere on cars where the output is driven directly by a PWM or injector output on the Syvecs. CJ's setup uses a H-bridge driver from memory so the same issue will not occur I believe, and it doesn't seem to happen on the Neetronics.

Re: Cal for dccd

Posted: Tue Jun 26, 2012 6:08 pm
by TimH
Interesting - and maybe a bit odd? I would have thought that 0% duty meant no volts applied, unless there's some kind of "base duty" that's applied? 8V suggests 40% duty cycle or thereabouts?

Re: Cal for dccd

Posted: Tue Jun 26, 2012 7:37 pm
by pavlo
Not at all.

The output is either on or off, when on, it's pulled hard to ground (FET on and shorts output to ground). However when off it's not pulled hard to 12v (or Vbatt) and is allowed to float. What it floats to will be down to whatever else is on the output stage, and how the ECU itself is grounded.

It could be that there is just perhaps a 1v potential between the ECU vbatt and the chassis 12v feed used by the DCCD, it could be that the ECU power goes through a reverse protection diode too, that might drop another 0.6v or so. Suddenly we are at 1.6v or 11.4% duty, and as the DCCD has a ramp to further load the clutch, a little duty goes a long way to begin with.

Re: Cal for dccd

Posted: Tue Jun 26, 2012 8:14 pm
by TimH
If it sounds like I'm being "difficult" please forgive me - I'm not; I'm just trying to understand it :)

Yes, the output floats when off, but to turn the DCCD solenoid on I presume something has to sink some current somewhere. This ought not to be the mosfet (very low leakage current as mentioned). If there was a ground difference then that's a worry, as it means the mosfet gate drive cannot reliably control the output as needed...but there might be a clue here since I imagine the ground feeds in the ecu are spread amongst the pins, with some as high current and some low current. Maybe there is, simply, a bad earth on the car in question? That could, maybe, cause an output to not turn off properly and might lead to other issues too.

A difference in the 12V feeds ought not to make a difference, nor a flyback diode since that, also, goes to +12V, and it's a ground that will cause the DCCD to energise.

I should add that I have no practical experience of DCCD so maybe it behaves differently to how I would expect? What sort of current are we talking about for DCCD control?

Re: Cal for dccd

Posted: Wed Jun 27, 2012 10:48 am
by pavlo
You don't need a ground to energise the coil, merely a difference in potential. SO a an ECU that can only let an output float as high as 12v when the DCCD is supplied with 13v will have a 1v potential. You are not looking for a poor earth, you're looking for a poor 12v feed.

There is more to an ouput stage than a regular mosfet alone, so I wouldn't be so quick to assume what is going on here.

Re: Cal for dccd

Posted: Wed Jun 27, 2012 10:53 am
by pavlo
Also, the DCCD is not controlled by a solenoid, it is a simple coil so small currents will have an effect.

Like I have said, I have not measured the current in question, but this effect has been observed on more than one car and is not present with a Neetronics controller on a car where the DCCD is switeachable between Syvecs and Neetronics.

Re: Cal for dccd

Posted: Wed Jun 27, 2012 12:28 pm
by TimH
Thanks Paul - as I said, not had hands-on experience with DCCD so had assumed it was a PWM-controlled solenoid like VVT, that would be on or off (albeit at 300Hz or whatever). So the +12V difference seems to be the culprit.

Most likely is the flyback diode - maybe the Neetronics has no flyback diode?

You said that CJ's DCCD, driven by an H-Bridge driver, does not exhibit this problem? It could suffer the same if the H-bridge +12V is lower than the DCCD coil 12V, since they have flyback diodes built in, so H-bridge itself is not the magic answer.

I note that the OEM DCCD system has both sides of the coil connected to the controller - is this the case with the problem vehicles, or is one side of the coil taken to a 12V feed elsewhere? If so, the solution might be to wire the coil back to the Syvecs power pins?

All interesting stuff :)

Re: Cal for dccd

Posted: Wed Jun 27, 2012 1:13 pm
by pavlo
The AVCS solenoid doesn't operate as on/off, the physical end result is a variable position.

The Neetronics should have a flyback diode otherwise the induction load of the coil will result in the operation all happening over the 75-100% range instead of 0-100%. Using a diode is the correct way to connect the DCCD.

I believe CJs install uses a H-bridge driver and so I believe he should have the same effect, I don't know for sure.

The OEM controller doesn't control both sides of the coil, the +ve side of the coil goes via a relay which is not energised in error states, and may also deactivate in handbrake operation.

Re: Cal for dccd

Posted: Wed Jun 27, 2012 6:30 pm
by pat
The fact that we fit external Schottky diodes for flyback clamping should be sufficient to convince anyone that there is nothing in the ECU that will limit flyback voltage to VBAT. On an injector or PWM output there is a 60V clamp on flyback voltage, this is to allow the magnetic field in reciprocating actuators to decay rapidly. If it still isn't fast enough then the ignition outputs on the S8 are clamped around 400V if memory serves.

The DCCD coil is just that, a coil. When energised it causes plates within the diff to move closer to each other increasing viscous drag between them. The second lot of plates operate a ramp assembly with ball bearings that provide the load on the main slip limiting plates. It is a linear device and as such should have a Schottky flyback diode fitted, and said diode should be able to handle 10A. The ones we normally use on the AVCS OCVs would be marginal.

There is some biasing circuitry on the output stage which can be used for fault detection, so it is possible that there may be a very small leakage current through the DCCD coil however it should be insufficient to cause the diff to lock up to any tangible degree. It is possible that the effect observed is due to remancence since the flux decay is being actively slugged when the driver is turned off. Whilst the H Bridge driver should not be allowing reverse current it is possible that a small amount of jitter in the no-drive state is causing current dithering thus fully demagnetising the coil. It is also possible, albeit unlikely, that turning the ignition off might allow VBAT to rise briefly (there is normally no shunt across that rail other than things like the Tranzorb on the way into the ECU, but if memory serves it is after the PI filter) and thus allowing full demagnetisation. The higher flux decay slew rate with a floating VBAT may reduce remanence also.

In the first instance it would be sensible to measure the quiescent current through the coil. It may also be interesting to try setting the DCCD drive duty to 0% everywhere and see if the problem persists... should certainly help point in the right direction. Might be worth me bearing this in mind when finalising the design of the new "DBW controller"!

Cheers,

Pat.

Re: Cal for dccd

Posted: Thu Jun 28, 2012 12:57 am
by TimH
pavlo wrote:The OEM controller doesn't control both sides of the coil, the +ve side of the coil goes via a relay which is not energised in error states, and may also deactivate in handbrake operation.
In my MY05, both sides of the coil go to the DCCD control unit; the relay also goes to the control unit, as separate connections. This arrangement ought to minimise any potential difference between the driver and the 12V supply - if, indeed, it is the flyback diode and potential differences that's to blame.