Eliminating RF interference from HID(xenon) ballasts

Greetings!

Some background:
I have installed after-market(chinese) Xenon HID High-beams in my car. These have a ballast which transforms 12VDC from the car battery to 23kV AC required to ignite the lamps. The problem is that every time i turn on the HID:s the car stutters and warning lights inside the car(ABS, EPS etc)turn on.

I work with R&D for a car company so I've conducted measurements both on the car itself and on separate HID-kits and a spare car battery. Unfortunately my area of expertise is combustion engines so i have only a rudimentary knowledge of electronics. My conclusion so far is that there is high frequency interference returned from the ballasts which propagate through the DC-cables(the error disappears when i use an external battery so airborne interference is not the cause) to the battery which in turn cause a reset of some node on the CAN.

On the battery, oscillations with and amplitude of of ~150V and a frequency of 15-50MHz occur when the HID:s are turned on. I don't know if they are common mode or differential mode, it is difficult to get repeatable measurements with the oscilloscope since it picks up the airborne noise also.

The ballasts use approximately 70A each in peak current during the first ms after the lamps are turned on but it appears(also a bit uncertain) that the disturbances occur before the large current peak.

I've tried filtering the current cables from the battery to the ballasts close to the ballasts with capacitors, 1, 10, 100 nF(which should have low impedance in this frequency range) and ferrite chokes scavenged from computer cables. These help to some extent but don't eliminate the problem completely. The error still occurs from time to time.

My next step would be to order ferrite cores with correct permeability from an electronics supplier. However, if the EMI is not common mode i will have to pass the + and - cables from different directions and the ferrite cores will become saturated(because of the 70A current) and will not be effective against the EMI, right?

Some questions:
1 Do you know if EMI from these types of transformers is usually common mode or differential?
2 How should I wind the cables around the cores?
3 Should I even use ferrites or something else entirely? Inductors?
4 If the EMI is differential mode and i cannot find ferrite cores which do not saturate, how should I go about in adressing the problem?

As I said, this is not my field of expertise so maybe I'm asking the wrong questions and maybe I'm not providing sufficient background information, in that case comment and I'll try to be more clear!

I'm in over my head here so I'm extremely grateful for any assistance!

Regards
F

Some experimentation is needed.
Inductors (ferrite cores) will certainly reduce interference because they resist sudden changes in current in the cable passing through them but that may also limit the current surge you actually need to ignite the lamps.
Capacitors across the ballast are probably a better idea. If you connect them as close to the ballast as possible, they will act like a local 'reservoir' of energy and allow the surge to be drawn directly from them instead of only loading the wiring loom back to the battery.

A combination of capacitors directly across the ballast and a ferrite ring in the wire feeding power to them might be another good option.

I would suggest you try a capacitor of say 10,000uF rated at 64V wired directly across the 12V feed at the ballast. If you want to try a ferrite as well, the cable must loop through the hole and around the outside as many times as it will fit. I'm guessing the cable will be quite thick so you may only manage a few turns. The capacitor must be fitted directly across the power and ground terminals as close to the ballast as possible and the ferrite nearby on the +12V cable feeding them both.

For eliminating common more conducted interference, you should try and deal with it at the source. Usually the source of such interference is some node with high voltage and high frequency signals, which is coupling directly to other parts of the system (including the chassis/frame of the car) via electric fields (capacitive coupling, basically). Even though it's still happening via coupled fields, it's still conducted interference, so your second battery experiment doesn't disprove it. So what you would want to do is prevent all that electric field from reaching anything other than your local circuit ground. Generally this is done with a faraday screen or shield surrounding the problematic nodes, and you attach that shield directly to your local circuit ground. This allows those currents to flow, but in a small, controlled loop which doesn't interfere with other systems.

If that problematic node is in the lamp itself, then it may be an issue of mounting the lamp differently. Hard to tell without seeing its construction in detail.

For differential mode interference, you should be able to get away with just using high quality snubbing capacitors. Look for polypropylene metalized caps, like the MKP types from Vishay or EPCOS. You will likely need several uF total, placed very close to the ballast, in order for them to be effective. Even if 100nF at 50MHz is very small, there may also be additional impedance in the form of ESR and ESL (parasitic series resistance and inductance) that completely dominates that. So use bigger caps, with very short wires. If you do things properly, it's likely you can get away with no chokes at all.

One side of the 12V ballast supply (probably negative) should be connected directly to chassis so common mode interference is unlikely. I think the current surge is the culprit rather than radiated interference and I would concentrate on keeping the impedance across the ballast 12V input as low as possible.

Brian.

The battery negative is likely connected to the chassis at one point, but not near the actual ballast. Therefore the return currents from the interference source will follow a long path, all the way back to the battery terminal. If any electronics use the chassis as a reference or ground return path (unfortunately this is often the case), then they will be fully affected by the interference.

I doubt it's the surge current, at least the low frequency component. Auto batteries can typically provide >100A for brief periods (a second or so) before the voltage dips enough to affect other systems. However, if the ballasts are daisy chained of a common supply bus, and the wire isn't thick enough, then the ohmic losses from the wire could cause other systems on that bus to brown out. Try routing the supply wires directly from the battery terminals.

I was referred to this forum: https://forums.swedespeed.com/showthread.php?127030-Reduce-HID-ballast-radio-interference

Where a person had similar issues. However I'm a bit sceptical to the filter they have proposed. The person who suggested the filter in also suggests 2200uF Aluminium Electrolyte capacitors. From what I understand electrolytic capacitors are not suitable for eliminating RF, they are too slow.

Will this filter really be effective for EMI in the 10-50MHz range?

Finding ceramic or other 'fast' capacitors with the capacitance he has suggested for this lowpass filter is not a trivial task.

Suggestions?

You certainly won't find ceramics with those values, they would be bigger than the whole car! (I'm excluding battery backup caps because they wouldn't be suitable anyway)
What you can do is connect capacitors in parallel, one or more 'big' value ones and one or more ceramic ones across them. This give the benefits of both types, the electrolytic has the storage capability but is poor at high frequencies and the ceramic is better at high frequencies. If you do that, I suggest two 100nF ceramic capacitors rated at 64V or more directly across the big capacitors.

Incidentally, the article mentions 16V rated capacitors, I would go significantly higher than that because as loads turn on and off on the loom, the voltage can go well into the 20V+ range or even higher, not for long but long enough to pop a capacitor.

Two lights each drawing 70A at start-up is too much for most car batteries unless you have a Diesel engine. Diesels take more current to start them so the batteries are usually larger capacity than in gasoline cars.

Brian.

Brian, thanks for answering. Yes, this thing with the capacitors made me doubt the filter shown in the other forum. I don't have the theoretical knowledge myself to start modifying the filter with good confidence. I mean shouldn't the inductor and capacitors be matched with regards to resonance?

( 140A (70A*2) is actually nothing special for a car battery, gasoline or diesel. During cranking the starter motor uses 4-500A. During a coldcold-start the starter-motor is almost short-circuited so the current can be even greater. The ECU still needs to function. A reset should not be triggered as long as Ubat is above 7-ish Volts. The 1 ms current peak only causes the battery voltage to drop by 1V; from 14.4 to 13.4 which is nothing. I havent solved the issue yet so obviously I can't say for 100% sure but right now I don't think it's the starting current and the corresponding 1V voltage drop that causes the problem. )

The components are not intentionally resonant although there will be some frequencies at which they perform differently to others. They are there to make a low pass filter, a circuit that prevents the passage of higher frequencies. Ideally, DC should pass through unhindered and any signals should be supressed. No filter is perfect though and one designed to carry very high current and also kill high frequencies is difficult to build and usualy needs bulky components.

There is a danger that if you 'over' filter the supply so it can't pass the switch-on surge, the ballast will fail to ignite the lamp, that's why I suggest a big fat capacitor across the ballast input. It will charge at whatever rate the battery/wiring/ferrite will allow but also be able to discharge instantly into the ballast when needed. The current demand at the ballast will be very irregular until the plasma reaches a steady state and the capacitor will work as a reservoir of energy right where it's needed.

Regarding ECU 'brown out' voltage, you are partly right but internally it will have it's own reservoir capacitors and power filter which allow it to keep running for a relatively long time (tens of mS at least) even if the power is completely removed.

Brian.

If the HID current return is chassis sheet metal, that's
a bad setup. If it were me I'd be hardwiring direct to
battery or the first posts downstream, both + and -.
I'd do power switching with a relay if there's none
provided, running its coil off the present light power.

A weak battery may not be able to make up the
startup current, and that current may persist longer
than a 10,000uF cap can hang if you're talking
70 amps. Filters can only do so much against a
too-weak source. I make it 3mS to slump 2V at
this load, 10V is a sorta-normal automotive
undervoltage limit I've seen on inverters and such.
3mS may be short of the needed light-off time.

Some HV power converters will work in a "hiccup"
mode during the startup phase, kicking an AUX
winding / supply until the controller's close-in supply
has received enough enough pulses to reach the
control ckty's undervoltage lockout threshold.

Your big cap (for surge slump) and your EMI
suppression (if needed?) will be done by different
things. I might recommend a fat-@ss common mode
filter, like you find in "big iron" computer power
supplies' line input and output. You can use less
than the 70A value since that's just startup and
ought to be tolerated thermally. These are made
just for the EMI problem. They may impose more
series resistance than you'd like, TBD.