Correcting the GP UB3 cluster's speedo and tach

[RobertISaar]

should be possible with either a variable resistor or variable capacitor, right?

 

for now, the GP's odometer seems correct, but the speedo and tach read somewhere between 2 2/3 and 3 times too fast, what i was thinking is to insert one of the above components between the speedo and tach driver and adjust until correct, which should leave the odometer correct as well....

 

thoughts? discrete components are cheapa and easily available to me, clusters are not.

[Galaxie500XL]

Depends on whether it's an analog device, or digital. In many cases, modern cars use digital pulses for those devices, then simply run them through a D/A converter at the device, converting pulses to a voltage. If it's digital, a simple decade counter might do the trick, by reducing the number of pulses going to the gauge.

 

I don't honestly know whether they are analog or digital, but I suspect they may actually be digital...do you have access to an oscilliscope?

Edited October 3, 2010 by Galaxie500XL

[RobertISaar]

a sillyscope would be nice, but no, does not haz. i know the 4000 pulse per mile signal the ECM sends to the cluster is a square-wave(tach should be as well), but i'm not entirely sure what happens after that, i didn't dig too deeply into the PCB the last time i had the cluster out.

 

here is what alldata says on it, but it doesn't make a whole lot of sense to me.

 

 

 

 

now, if they truly are variable voltage, i should be able to stick a variable resistor after the portion of the circuit that drives the odometer and cut down the gain rate by 1/3 to get the guages accurate..... i think? the tach should be easier though since i don't have to mess with anything other than the tach drive itself.

[Galaxie500XL]

Actually, you gave me just what I was looking for..."resistance through the sending unit determines current flow through it's coil", and so forth.

 

You say both the tach and speedometer are reading high...yet the odometer appears correct. Keep in mind, in this sort of circuit, a BALANCE of the two gounded coils will determine where the needle sits. If it's reading too high, based on the description you're giving here, one of those grounds is either poor or missing. (I'm betting poor).

 

The description of the "square wave" tells me it's digital until it gets to the dash circuitry, and the correct odometer tells me you're getting a good input... That leaves either a bad "converter", or a poor ground. Not likely that two devices would exhibit the same failure of components, which leads us back to wiring.

 

Find that, and I'd bet your gauge will work properly.

[mra32]

Actually, you gave me just what I was looking for..."resistance through the sending unit determines current flow through it's coil", and so forth.

 

You say both the tach and speedometer are reading high...yet the odometer appears correct. Keep in mind, in this sort of circuit, a BALANCE of the two gounded coils will determine where the needle sits. If it's reading too high, based on the description you're giving here, one of those grounds is either poor or missing. (I'm betting poor).

 

The description of the "square wave" tells me it's digital until it gets to the dash circuitry, and the correct odometer tells me you're getting a good input... That leaves either a bad "converter", or a poor ground. Not likely that two devices would exhibit the same failure of components, which leads us back to wiring.

 

Find that, and I'd bet your gauge will work properly.

 

I agree.

[GnatGoSplat]

There used to be a web page on the old site where I covered this very issue with pics. The original HTML files are still there, but I can't link to it because I don't remember what the filename was. Maybe a mod knows?

 

Anyway, the GP tach and speedo are both voltmeters. There are driver circuits to convert the pulses to voltage. They both have a calibration resistor that's etched onto the back of a white ceramic DIP IC looking thing. GM calibrated this by cutting a line of appropriate length into the resistor that changed its resistance. When speed reads too fast, that means there is not enough resistance across the calibration resistor and you can add another resistor in parallel. I used a 1megohm (I think?) fixed resistor across mine and that made it really close to perfect, but a variable would give you more flexibility.

[RobertISaar]

1 megaohm? damn, i wasn't thinking that much additional resistance would be necessary, then again, when in parallel, the average of the resistance values is what determines it, yes? oh well, 1M ohm potentiometers are still only .40 each.

 

this thread has been very informative... maybe a sticky or FAQ will get made out of it if the page specific about it doesn't show up in a while?

[GnatGoSplat]

With resistance, more is less! The higher the number, the less current is allowed to pass. Parallel resistance for 2 resistors would be R=(R1*R2)/(R1+R2) so it's not an average. The combined resistance will be less than both individual resistances, but much higher than the average.

 

Yeah, I hope they can get some of those old pages back up. I had a lot of electronics-based info, even teardown pics of things most people wouldn't normally tear down like the inside of an ignition module.

[RobertISaar]

wow, i must have been thinking of something else.... oh well, 1/2watt 1Mohm potentiometer(along with a bunch of other discretes) going to be en route very soon and i'll have to tear the GP dash apart again to deal with it when i know it won't be used for a while.... i'm not going to see much of a difference between a linear and logarithmic in this kind of application, am i?

[GnatGoSplat]

Nah, linear vs logarithmic shouldn't matter because you won't be adjusting it after you get it set right. Although I wonder if bigger than 1Mohm would give you more room for error. For example, if your calibration resistor is off less than mine, you'd need a higher resistance. With a 1M pot, your available range will be 0-1M so one with a higher max might not be a bad idea. The worse off your cluster is, the lower the resistance you will need as a lower resistance more drastically changes the overall resistance when in parallel.

[RobertISaar]

err..... add a high-ohm fixed resistor in series with the variable to up resistance if necessary to make the pot have finer resolution? or am i thinking of something else again?

 

"The worse off your cluster is, the lower the resistance you will need"

 

well, both are roughly multiplied by 3 as-is, if that's any kind of metric as to what you dealt with.

[GnatGoSplat]

Yeah, actually adding another resistor in series with your pot would work if you need more resistance. Multiplied by 3? You mean your speed reads 90 when you're going 30? That would be WAY worse than mine was, for sure! I think mine was about 20-25% off on the speedo and I never did anything to verify the tach.

[RobertISaar]

yeah, the speedo maxes out when the ECM says ~40MPH via the datastream and a hot idle in park is ~2500 RPM (~800 via ECM) on the tach.

[GnatGoSplat]

Damn. I didn't catch that in the OP. Have you tried resoldering some stuff? It seems like you could have a bad ground. I can't imagine a calibration resistor being THAT far off. These clusters are notorious for bad solder joints, especially where they've used a lacquer to protect the solder. It seems the lacquer crap designed to protect the solder actually makes things worse.

[RobertISaar]

well, i did resolder the points for the IC that deals with the check engine and low fuel light(i believe it is a comparitor?), and they now work correctly roughly half the time....

 

when i pull the cluster again, i guess i could trace where all the 4K PPM circuit runs to, and retouch anything it comes into contact with.... or i could trace the ground pin on the connector and see where all it leads....

[RobertISaar]

ok, pulled the cluster last night in about all of 5 minutes thanks to only needing to remove 2 screws to pull the dashpad and 3 screws holding in the cluster...

 

took some messing around to get the cluster into it's 3 main pieces, but it felt like i was missing some fasteners, nope it's the little shafts that connect to the motors that drive the guages... just have to wiggle it out side by side, and then they come apart...

 

resoldered EVERYTHING on the PCB in about an hour, the only difficult portion is the coated section, that stuff takes a while to either burn off with the tip or makes the joints look like ass.

 

then i decided to start marking stuff to try and reverse-engineer the board to get an understanding of what does what, and to see if any of the discretes were out of spec... turns out i have quite a few resistors(all of them being the metal-film kind with the 1% tolerance) that were out of spec, ALL of the out-of-spec having less resistance than they should, so it wasn't a coating/contact issue. i haven't figured out what all of them are for yet, but i get the feeling that they are contributing to some of the erratic operation of this cluster...

 

on that note, there are a bunch of little yellow components that i can't quite identify... they don't test well as resistors or diodes, i'm thinking they might be caps, but i have no way of testing that theory. can't really test the pair of transistors or IC's either, but oh well.

 

i found the large white blocks with the laser etching on them, but i'm not sure which terminals to connect the potentiometers to yet, or of they will even be necessary if i replace all of the failed compnents.

 

here's some pics for anyone interested. i typed in specs for the stuff i could identify, but i can't get any useful numbers off of the diodes and the little yellow pieces.

[GnatGoSplat]

Testing resistors is tricky. You can't reliably test a resistor in-circuit because they're often wired in parallel, not just with each other, but with other components that also have a resistance and will affect the reading. To reliably test resistors, you have to at least pull one leg. However, those fixed film type resistors almost never go bad, and when they do, they will have more (higher) resistance not less. If they're not blackened or burned, then most likely they're probably fine.

 

Little yellow components? The tannish-yellow looking things usually near ICs with axial leads? If so, those are definitely caps. Polyester film type. Those also almost never go bad. The only caps that are likely to be suspect are the electrolytics. They can dry out over time, with heat, because the electrolyte inside is water-based.

 

You can use the diode check function to get a good idea of the condition of a transistor. Once you've identified the pins E-B-C, E-B should read 0.6V one direction and infinity the other direction depending on whether it's an NPN or PNP type. B-C is the same. If you have 0V in both directions, then it's likely a shorted transistor. Usually you don't even need to bother identifying the pins, if between any two pins you have ~0.6V or so and infinity the other direction, that transistor is likely okay. If any two pins are 0V both directions (diode check mode) or 0 ohms both directions (resistance mode), that's a definite fried transistor.

 

ICs are hardest to test, because you need to know the function and pinouts of the IC.

 

The black paint on top of those ceramic white blocks is the calibration resistor. That black paint is conductive. The value of that resistor is fine-tuned at the factory by a very tiny cut in the paint that you'll see if you look carefully. The two pins the larger of the black paint resistors connects to is where I soldered my additional adjustment resistor.

[RobertISaar]

i found it odd that only the metal-film resistors tested bad in-circuit, and even then, not all of them... i may have to pull some and test that way then.

 

the transistor on the right acts odd when tested in-circuit, i read .91 and .97V for the E-B measurement, otherwise, they seem to check out.

 

there were a bunch of smudges on the resistor block from my greasy paws, but i believe i have most of it removed, but i can't see any cuts in either, but my eyes suck at that level, so i'm sure i just can't see it.

 

i'm about to go throw it back in the car, see if the resoldering did anything, if it fixed it, great, but i'm not expecting a miracle...

 

assuming it didn't, i have to find somewhere to place the potentiometers so that they can be adjusted while in motion, but not in a place where they can be easily thrown out of adjustment or broken...

 

maybe drive with the dashpad removed, have some ~12-18" wire leads running to the pots, have wife turn the dials while i'm driving and watching the guages while she watches what the laptop shows via the datastream? then stuff the pots in behind the cluster when i'm putting the dashpad back on?

[GnatGoSplat]

Resoldering by itself is often a miracle fix on these cars!

I'm going to guess the resistors are most likely fine. I think the transistor is probably fine too. When it comes to testing components in-circuit, there's almost always something else that will affect the reading.

Running some leads sounds like it'll work.

[RobertISaar]

well, resoldering did nothing AFAIK, but i wasn't watching for the check guages or low fuel light, those may or may not be fixed...

 

so, since the guages were still showing roughly actual values X3, i got the potentiometers ready...

 

time-saving tip: just remove the 2 pins on each resistor block, desolder the stumps from the PCB, and stuff some leads from the potentiometer into the holes. now instead of having to potentially fight with the factory resistor, now ALL of the range is done via the pot. if necessary, i'll have to add in some 1M ohm fixed resistors into the leads to get more resistance to lower the guages further.

 

i guess it will get tested again tonight if i'm feeling up to it, maybe tomorrow?

[RobertISaar]

haz news!

 

the speedo is now 100% correct when compared to the datastream. that 1M pot is REAL touchy in the area i needed it to be.... if i did it again, i would likely use a 100K pot and a combo of fixed resistors iin series, that way the tiniest movement doesn't make a 10% difference.

 

the tach however..... i think i killed the pot.... the only way i could get any kind of response from is was to push down on the knob and tweak slightly, but then the needle would only hand in one place, it did NOT respond to changes in actual engine speed at all.

 

when i pull the dashpad again to replace the dead tach pot, i'll have to record the values, in case anyone else tries this, give them a starting place....

[GnatGoSplat]

Awesome, congrats! If you need something with a lot of fine-tuning precision, you could also get a multi-turn pot. Only problem with those is they need to be adjusted with a tiny screwdriver.

[RobertISaar]

driving is dangerous enough as-is

 

screwdrivers would be a bad idea with me behind the wheel.