proper shielding in a harness

[bry195]

if you have taken a glance at the wiring diagrams you see in places a note that says shields must be intact at least 10mm to say the coil.
in america the statement is the same but usually .5 inches.

what is a shield? have you ever cut the insulation off a multi-conductor wire? found some foil wrapped around the wires or maybe braided aluminum? this is a shield.
why a shield? a shield is just that. it keeps noise from escaping or noise from entering the conductors. it can be used both ways. the noise is dumped to ground with a good shield in place.
how do you use a shield? normally shields are grounded at the source of the signal or termination point. huh? the dme has transistors that are delivering a current to a coil. the dme is the signal source and the coil is what is being driven. the tps is a signal source and the dme is being driven. keep in mind my rule is a starting point and you will find exceptions to the rule. the shield is normally grounded at 1 end only and again the shield should be intact up to 10mm of the driven device.

one exception to this rule is if you have high frequency noise. anything above about 8khz in my experience should have the shield grounded at both ends and the connections and conductors to ground should be wide rather than dense.

http://i1060.photobucket.com/albums/t445/bry1955/1-1-15/1e50d1f08edd485afed7f288081002f5_zps278f27c8.jpg

so what I have is a 4 conductor shielded wire. it has an outer aluminum braid and an inner shield conductor. each pair of wires has its own foil shield. this is a very nice telecom cable that im applying to coil drivers.

I pulled the outer braid apart and turned it into a conductor. I took both shield conductors and folded them over the insulation and heat shrunk the whole cable. I seperated the conductor pairs and left the foil intact up to 10mm before the pin. I heatshrunk those seperately.

why did I fold the foil over the heatshrink and pulled the shield conductors under the heat shrink?

this is a trick I learned at Bosch. never saw it anywhere before they did it. the shield conductors will get from multiple assemblies will get hose clamped or zip tied (metallic) together right where the heatshrink is. strain relief and shield grounding in one. so that manages the grounding of the shield conductors but why did I leave the foil at the ends? Just in case. if I still have a noise issue I will move the clamp to the end of the paired wires. its usually not needed but I left them there and they can be cutoff once I test everything out.

[wazzu70]

Good post. When I was wiring up my system the "sensor grounds" threw me off. I didn't know what they were or how to ground them.

On a standalone ECU and on the factory ECU you will see a pin for sensor ground, which is where these ground to (along with other sensor grounds). This pin usually has better protection to help filter noise than a standard ground pin.

[bry195]

thank you sir. Im in shield hell so I figured someone else might find themselves in shield hell at some point as well.

[dburt86]

I was in a shield hell. I got caught up in it so bad, i started questioning things i knew to be correct. The thing to remember about shields (Bry195 actually told me this and it stuck) is that a shield that is NOT grounded and should be, is literally an antenna, basically working against you. So keeping that in mind, i spent a couple minutes looking at a schematic for my year, located all the shielding wires and made sure i brought them to a ground on the ECU.

Very simple concept that is easily overlooked and will bring you world of hurt.

There are only a handful of shielded components in the stock OBD1 BMW harness. Injector signal wires, coil signal wires, and crank sensor are shielded. Its vital for all of those sheilds to make it to ground, just like BMW did it. TPS, CLT and MAT all share a common sensor ground.


And like Wazzu, it took me some time to figure out he difference between a sensor ground and a ground coming from the DME/standalone to Earth ground. 2 very different things happening there, but only separated on a ETM by a small symbol or letter.

[bry195]

I was in a shield hell. I got caught up in it so bad, i started questioning things i knew to be correct. The thing to remember about shields (Bry195 actually told me this and it stuck) is that a shield that is NOT grounded and should be, is literally an antenna, basically working against you. So keeping that in mind, i spent a couple minutes looking at a schematic for my year, located all the shielding wires and made sure i brought them to a ground on the ECU. Very simple concept that is easily overlooked and will bring you world of hurt. There are only a handful of shielded components in the stock OBD1 BMW harness. Injector signal wires, coil signal wires, and crank sensor are shielded. Its vital for all of those sheilds to make it to ground, just like BMW did it. TPS, CLT and MAT all share a common sensor ground. And like Wazzu, it took me some time to figure out he difference between a sensor ground and a ground coming from the DME/standalone to Earth ground. 2 very different things happening there, but only separated on a ETM by a small symbol or letter.

great point. to get a feel for what the DME or EMS is doing as it relates to sensor grounds I'll give a little example.

analog inputs have a resolution. you've see it before. 8, 10, 12, 14, 16, 32 bit. what does this mean?

computers work with 1s and 0s so how does it see 5 volts? there is an analog to digital convertor. it has a resolution of 16 bits.0000000000000000 or 1111111111111111 or anything in between for example 1010101010101010. notice how each one is 16 wide? reading from right to left each bit represents a different value added together. 1st=1, 2nd=2, 3rd=4, 4th =8, 5th=16.......if the first 3 least significant bits are on or 1 the value is 7. so in the end you have a range of 0 to something like 32,767 iirc and im too lazy to calculate it right now.

so, 16 bit register is where the values from the 0-5volt input are stored so the dme can process. if the input is 5 then the register should show all 1's and the dme sees a value of 32,767. so divide 5 by 32,767 and you will see the maximum resolution the input is capable of.

so the analog to digital convertor is comparing 5volts to its reference voltage. usually 0 or its internal ground. if your sensor is using chassis ground and it does not have 0 resistance to the reference ground the dme is using then you are essentially leaving a couple bits always on. why? if you measure chassis ground to sensor ground and it shows .75 volts, then the bits that represent .75 volts in the memory register will always be stuck on. lets do the math.

lets say its a 5 volt 3 bit analog channel. 111=1+2+4=7. 5/7= lets say .75. so for every .75 change in voltage another bit changes value in the register. if you measured .75 volts from sensor ground to the chassis ground at the sensor than the least significant bit is always going to be on. 001. you lost a bit now the effective register resolution is 2 bit.

now consider a 16 bit wide register and .75 volts. a couple of your bits do nothing anymore. in some cases you may want this.

for example they say the oem e36 maf is too sensitive in push through configuration. it would be great if the dme could filter the noise so you could run the oem maf. a simple electronic filter is just removing some of the resolution from the register. if you had the source code you could create a bit mask that holds the first 3 LSF bits high.0000000000000111. now the dme wont see those minute changes. if you dont have the source code you can create a dead band by leaving a little voltage on the input. but you will have to account for this bit loss in the MAF transfer table.

[bry195]

I guess these explanations are very close to complete as it relates to analog inputs. I guess maybe the last piece of the puzzle should be explained. too bad its titled shielding. Maybe it should have been called shielding, analog/digital inputs, and transfer tables or calibration.

so we know analog voltages get converted to digital registers. in our case I believe they are 10 and 12 bit wide. so 1111111111=1+2+4+8+16+32+64+128+256+512=1000ish

5/1000=we will say .001 volts per bit. I know my math sucks but im lazy tonight.

but these registers are converted again to say flow. the transfer table does this or what is commonly called a calibration or sometimes a lookup table.

if 5 volts = 1000=600 lb/hr 2.5volts may or may not = 500 or 300 kg/hr for something like a maf.

mafs are not linear so chances are it will need a transfer table or lookup table so that volts=register values=flow. it can be a math table to do the conversion or it can be a cross reference like a lookup table.

500=200kg/hr, 250=50kg/hr.

digital inputs can be analog as well. a frequency input takes a couple more steps. a single input pulses 62 times per second or 62hz (ignition cycle). the 62 hz is processed by the cpu and converted to an integer register. 1111111111111111. so all on may = 10,000 rpm/60 to =rps/ignition cycles per revolution and now you see how rpm, frequency and so on work together

these have all been voltage and analog style inputs. there are at least 2 more common types. current or resistance and mv/v. current is a simple resistance measurement. these are superior to voltage based in that they are not skewed by noise. lastly mv/v is like a transducer or egt or rtd temperature sensor. these are just like voltage based other than they generate millivolts instead of volts. very prone to noise. usually there is an amplifier as close as possible to the sensor to limit the chance of interference from transmitting over lengths of cable.

[bry195]

here is a tip that can save you about an hour per cable:

if you have a cable that has a braided shield and you want to retain the shields or de-braid to drain the noise to ground. Rather than taking all of the strands apart and then braiding them back up you can strip the insulation back to the length you want. go down to the base of the braiding where its close to where you cut the insulation to strip it. seperate or poke a hole in the braiding large enough to where you can pull the inner conductors out from inside the braid through the hole. no damage to the braid will occur and you will have a nice tight braid that you can ground.