Cheap HID ballast teardown and repair

       My dad's newly purchased second hand car came with a cheap drop in HID kit for the headlamps and foglamps. He complained that one side started to flicker and blink so I suspected either a dying lamp or faulty ballast. When it starts to blink, the ballast is making an arcing noise inside so I swapped in a spare ballast and it was fixed.

       I had nothing better to do that day so I decided to see what's up. These things are normally fully potted in silicone so I wasn't expecting much but this one had partially cured potting compound which was still sticky but easily came off so I dug in deeper.

       To my surprise while peeling the silicone off, there was fine white sand coming out of the ballast and thought there would only be a small amount so I poured some of it into the trash bin. A while later, more was still coming out so I decided to investigate further. I thought this looked interesting and went to document the teardown so here are the pictures that I took.

Peeled off the top potting and found a layer of sand
(Some of the collected white sand in the container):

A second layer of potting underneath:

Here's the label side of the ballast:

After desoldering the connector and ignitor block, the PCB came out:

Yes, that PCB looked the same even before I desoldered it.
They didn't bother cleaning the flux residue off:

Top side with a little more potting compound left:

Now there's your problem. The only blown component to be found.
The film capacitor in the middle of the pic is cracked:

Main switching mosfet and diode for the step up converter:

While I was picking away at the potting compound, R2 just pulled out of it's hole!

There it is, some leaded parts desoldered and all remains of the potting compound removed:

LM2902 is a quad op amp. I suspect this controls the constant power regulation for the HID lamp.
No, I did not touch the pot to see what it does:

H-bridge to convert HV DC to square wave AC needed by the HID lamp.
Mosfets are Samwin SW830 (500V 5.5A), TO-92 devices are 13003 parts which are used as level shifters for the FETs.
Blue device is a 470V MOV, 14pin IC is an HEF4011 which is probably wired as an oscillator for the H bridge:

Controller for the step up converter - UC3843,
blue thingy is a bodge mod - an MOV across the main rectifier diode:

Original capacitor on the left, replacement capacitor found in junkbox on right.
I know it has a lower voltage rating but how can a part that small be rated at 630V?
I don't trust these chinese part ratings anyways :

Popped in the replacement cap and powered it up.
I was prepared for fireworks when I flipped the switch but luckily, it worked.
A couple on/off cycles (including several hot restrikes) and it still worked like new:

Power input after a few seconds of warming up.
It initially peaked at over 90W but slowly dropped as the bulb got to operating temp.
It stabilized at about 42W input power - typical of 35W HID ballasts:

Interesting spring clip to hold the TO220 parts to the case:

Spring pushes against the main transformer to push the devices to the aluminum case:

Top side of PCB:

Bottom side of PCB (flux residue cleaned off using IPA):

       I was surprised that this ballast had very simple circuitry. I was expecting a custom IC controller of some sort that has the HID ballast algorithm in it which controls the power required by the lamp during boosted power (for faster warm up), constant power regulation, hot restrike, open and short protection, etc. The kit probably costs around US$40 for the entire set (2 ballasts, 2 bulbs, harness) so I could see why they kept the circuit to the bare minimum. As for the sand, the only reason I could think of is that they used sand as a filler to save on potting compound. Not the best idea as sand can become conductive when it absorbs moisture. By the way, the board has a website in the silkscreen which is popnow.cn which appears to be a generic Chinese HID manufacturer.


       29 Jun 2014:

       I opened another cheapie ballast and the potting compound seems to peel off easily so I tore it apart to see what's inside even if it's a working unit.

       This one appeared to be better built as the circuitry is a bit more complex.

Ballast torn apart with all the potting compound cleaned off:

Top view of the main board:

Power devices used. The four that are fully insulated types are used for the H bridge output:

Sub board front side.
This board contains all control circuitry for the main DC-DC converter, oscillator and MOSFET drivers for the H bridge:

Back side:

Main board contains all power circuitry:

Other side view of the main board:

Not much parts on the bottom side:

Reassembled everything and it still worked:


       09 Aug 2016:

       So, my car's fog lams suddenly died, fuse blown. one ballast was a dead short across the input terminal so I replaced it and time to tear the broken one down.

This one is a slim type.

Specs are your typical generic HID ballast.

Pretty recent manufacture date.

I cheated a little here. I already removed the soft silicone potting compound.

Back of the PCB

Outer shell. I believe this is made of cast zinc.

Bastards erased the chip number. This chip appears to be an ASIC designed for HID ballasts as it both controls the main DC-DC converter and the output H-bridge.

Two of the output transistors and driving circuitry.

Back of the main switcher MOSFET and a polarity protection MOSFET.

The bump in the middle is supposed to thermally connect to the switcher MOSFET for heatsinking but this one does not touch. Although I'm sure it wasn't overheating that caused the failure as I have used this ballast for a 9hr trip. And it had just been turned on when it failed.

There is a melted spot on the insulating plastic.

It lines up to one of the output H-bridge fet.

This looked like it overheated too. PCB underneath shows signs of delaminating.

Putting hot parts on both sides of the same area of PCB is not a good design practice.

The PCB is held in by one screw to the case. And even when fully tightened, it still does not make a good mechanical support.

Page created and copyright R.Quan © 18 May 2014.