HID ballast startup pulse locking up microprocessors

Thread Starter

ITFlyer

Joined Feb 22, 2012
3
I have a problem with the aftermarket HID ballasts on my GL1500 Goldwing motorcycle. They work fine - but the instant draw they impart on the 12 volt bus the instant they start up is causing lockups on microprocessor-run devices also on the bike.

When first powered up, they present almost a dead short for an instant (I assume this is when they are producing the high-voltage pulse to strike the arc). The ballasts are wired directly back to the battery, and yet everywhere else on the bike I am still seeing this pulse. This is a measurement taken from the 12 volt power at one of my microprocessor driven devices. Scale is 0.1v/division, so it's about a 0.4v instantaneous drop:



I have tried everything I can think of to solve this. I've added a ferrite choke to the ballasts to dampen the spike (damping both positive and negative feeds), as well as another choke at the devices being affected. I've tried tons of different sized electrolytic capacitors at the affected devices, including running power through a diode first to prevent draining of the caps back into the bus. Still the pulse is getting through and locking up the devices.

I thought it might be EMI or RFI, so I physically relocated the HID ballast components far away from the devices. The ballasts are already encased in metal cases. I tried enclosing the affected cases in metal cases, with no difference. Now that I write this, I haven't 100% ruled this out by powering the affected devices from a different power source while starting up the HIDs, but my testing thus far (including hooking a coil up to my scope and moving the coil near the affected devices to measure EMI as the HIDs started up) show that there is nothing that should be capable of causing this.

The devices being affected are a custom Arduino device that I designed, and an off-the-shelf dashcam-type DVR writing to a micro-SD card. The Arduino device continues to work, but the LCD processor that runs its display locks up when the spike occurs. The DVR also continue to write to the card, but the video frame processor locks up when the spike occurs, so it just writes the same frame (from the frame buffer, no doubt) out - i.e. the video freezes (although the overlaid clock continues to count).

I'm looking for suggestions or ideas as to how to mitigate this pulse and fix my problem, as I'm out of ideas at this point.
 

Attachments

crutschow

Joined Mar 14, 2008
34,280
Have you tried a large capacitor (1000μF or more) directly across the ballast connections?

How much current does the ballast take in normal operation?
You might try adding a properly sized inrush limiter, such as one of these, in series with the ballast to limit the startup current (perhaps with a large capacitor directly across the ballast if needed to help provide the large startup current of the ballast).
 

strantor

Joined Oct 3, 2010
6,782
I don't think your 0.4V drop is telling the whole story. That doesn't seem like enough to affect anything in and of itself; these devices, especially the dashcam, should have their own capacitors inside - I know my dashcam will continue to operate for a few seconds after I unplug it. However, if your bike has an alternator on it, it might be overcompensating for the dip with a (delayed reaction) wildly high spike. See if you can trigger your scope on a high spike instead of a low spike.
 

Thread Starter

ITFlyer

Joined Feb 22, 2012
3
Have you tried a large capacitor (1000μF or more) directly across the ballast connections?

How much current does the ballast take in normal operation?
You might try adding a properly sized inrush limiter, such as one of these, in series with the ballast to limit the startup current (perhaps with a large capacitor directly across the ballast if needed to help provide the large startup current of the ballast).
Each ballast (there are two, one for each bulb) is supposed to pull 6 amps at startup and then decrease to 3 amps running, but I think the instantaneous draw is much higher than that.

I have tried large capacitors (as big as 10,000 uF) across the ballast connections, but it actually made it worse - they added current draw when the headlight circuit was energized, due to the empty capacitors sucking up power in addition to the ballasts.

I don't think an inrush limiter would help in this situation as the ballasts actually *require* a large inrush of current upon power-up in order to strike their arcs. Limiting current will cause the bulbs not to light off at power up.
 

Thread Starter

ITFlyer

Joined Feb 22, 2012
3
I don't think your 0.4V drop is telling the whole story. That doesn't seem like enough to affect anything in and of itself; these devices, especially the dashcam, should have their own capacitors inside - I know my dashcam will continue to operate for a few seconds after I unplug it. However, if your bike has an alternator on it, it might be overcompensating for the dip with a (delayed reaction) wildly high spike. See if you can trigger your scope on a high spike instead of a low spike.
My inclination is to agree, because a 0.4v drop shouldn't be doing much of anything - particularly knowing the power supply module inside the Arduino-based unit that I designed and built. It consists of a choke, a 7805 and associated caps, a 20 volt zener to protect against spikes, and quite a few large electrolytics in there for filtering, enough that the unit actually runs for a couple of seconds after power-off. The camera DVR has a 12v to 5v regulator, and uses this to run as well as charge an internal LiPo battery, which runs the camera after loss of power in order to finish writing the last file.

I originally tried looking for a positive spike as well, but there was none to be seen. The issue happens regardless of whether or not the engine is running, so that should rule out the alternator's regulator.
 

strantor

Joined Oct 3, 2010
6,782
I have tried large capacitors (as big as 10,000 uF) across the ballast connections, but it actually made it worse - they added current draw when the headlight circuit was energized, due to the empty capacitors sucking up power in addition to the ballasts.
.
I wouldn't connect the caps like that. I would connect them at the switch. That way instead of adding to the instantaneous drain, they supply the instantaneous drain.
Caps go upstream of the switch and stay charged at all times. They are in parallel with the battery.
As soon as the switch is flipped, current from the caps (followed by current from the battery) rushes through the switch and strikes your arc.
 

crutschow

Joined Mar 14, 2008
34,280
..................
I don't think an inrush limiter would help in this situation as the ballasts actually *require* a large inrush of current upon power-up in order to strike their arcs. Limiting current will cause the bulbs not to light off at power up.
Okay.
So how about an inrush limiter followed by a large capacitor in front of the switch to the ballast?
That way the capacitor could provide the large current the ballast requires and the limiter would limit the current spike to the rest of the circuits.
For this to work well you need to run the negative (return) lead from the ballast directly to the negative lead of the capacitor and not have it run through the chassis ground.
 
Last edited:

ian field

Joined Oct 27, 2012
6,536
Have you tried a large capacitor (1000μF or more) directly across the ballast connections?

Maybe go one further than that - isolate the HID lead from the battery with a hefty choke.

Maybe a dozen turns of enamelled wire 2 or 3mm diameter round a ferrite slug. Something ready made could probably be liberated from a scrap ATX PSU, the 5 & 3.3V rails probably have current ratings comparable to the HID striking peak.

The 1000uF might needs some help from ultra low ESR non electrolytics, probably across the battery as well as on the lead to the HID.
 

MikeML

Joined Oct 2, 2009
5,444
Why not fix the problem at the microprocessor? If a drop off only 0.4/12 causes it to reset, then there is inadequate bypassing there...

Likely, your problem in not caused by the voltage sag; it is likely a ground loop along the frame of the machine, and the micro does not have a single-point ground.
 

ian field

Joined Oct 27, 2012
6,536
Why not fix the problem at the microprocessor? If a drop off only 0.4/12 causes it to reset, then there is inadequate bypassing there...

Likely, your problem in not caused by the voltage sag; it is likely a ground loop along the frame of the machine, and the micro does not have a single-point ground.
Could be a crosstalk problem in various wiring in the loom.

Supply decoupling local to the CPU probably won't fix that if its getting into signal pins.
 
Top