Happy to find this forum!
I bought a 1994 Miata last summer. It has a 1.8L VVTI, turbocharged, and is managed with a plug and play MS3.

I've been updating and repairing it slowly, when it developed this odd problem, video below:
End result is a really jumpy tach. Does not do it all the time, but when it does it makes it pretty difficult to find out what RPM the car is at.
I've plugged into Tuner Studio and it does not affect the RPM the computer is seeing.
At higher RPM it can do full sweeps like if around 4-5k it can sweep from 3k-8k which can be alarming.

If it was losing pulses I would expect to see it drop understandably, but at cruise it jumps up like it is seeing extra pulses. Part of me thinks that somewhere there is too much noise getting into the circuit for the tach, but not sure where it is coming from.

Reading online on possible culprits so far I have:
Replaced the crank angle sensor with a delphi replacement
Replaced the coils with NRG coils
Replaced the sparkplug wires
Replaced the cam angle sensor (genuine Mazda)

Thoughts would be appreciated

 

It’s a simple pwm output from the ECU with your setup, no crank/cam sensor or coil/wires will fix it if the RPM the ECU sees is smooth as you say. Try testing the tach output in TS, with the engine off, the tach should move to ~2500rpm at 72(?) hz. It might just be a bad wire connection or faulty tach/driver.

 

Oh, interesting. I thought it may have been getting a signal off the ignition circuit somewhere. I'll have to try test mode and see if I can get the tach to move as you suggest.

In the meanwhile, I found a dodgy ground on the intake, removed bad crimp and re-crimped, added a ground from the head to the intake and then from the intake to the chassis in case it was a grounding issue. I even took apart the cluster in case somehow the gauge face install was making the needle stick, but it it seems to rotate freely. Still jittery which would make sense if the ECU is feeding the signal.

Will update with results of test mode tomorrow

 

Tried test mode in TS, confirmed ECU can send signal by turning the fuel pump on and off. Tach output does not cause the tachometer to work. Now it sits at 0 most of the time and briefly flickers to a random RPM.

Getting to the ECU is unfortunately a bit of a nightmare with the fixed seats. Would be lovely to order the specific driver if anyone knows the correct model.

 

Mines not as bad as yours, but jumps from 3-4k often, then wakes up after 4k and is accurate after that. Super weird, my cars a 90 and MS is driving the tach. I thought it might be ground related but after going over the grounds a few times, changing stuff around, etc there has been no difference. I'm leaning towards something in the cluster itself causing it. But IDK. Curious to see what you find, mines not too bad so I mostly just ignore it. And of course just like yours the ECU sees no change in RPM and is always accurate.

 

The fuel pump was never in question. You've confirmed the buttons in Test Mode work, but nothing else. Check continuity and resistance from ECU to tach. I would try another cluster before blaming the ECU. Make that seat easier to R&R, this won't be the last time you have to access the ECU.

 

Agreed, fuel pump test was to just make sure test mode was working. Testing tach output did not net a result.
Will order or find a spare cluster, would indeed be silly if it was just a tired cluster that is causing me to chase my tail.

Short of cutting up the seat or the door card, not many options without losing a lot of headroom. Could change mounting location of the ECU, I suppose no hard reason it has to stay where it is.

Will update with findings.

 

Hz was set to 72? It’s only the highest and 2nd highest that make it move, below that it barely budges.

 

Mine has 39hz and 78hz as options for the top two settings, no twitch of the tach with toggling either either in test mode. Is the pulse duty cycle relevant? I set tested at 25%, increased it to 50% to see if it made a difference and did not notice any.

Purchased a tested gauge cluster from a parts breaker, should arrive next week. If that also fails, then will pull the seat and ECU.

 

Updates:
Replacement tach arrived. Replaced and exhibits same behavior. Unlikely the gauge cluster unless this one also is dying or died in a similar way.

I tried reading the tach signal with a simple oscilloscope which shows a really noisy signal. Would not auto capture a signal and kept showing a fluctuating signal from 600KHz to 1.3 MHz with 30-80% duty cycle (still learning this piece of equipment)
To me that means the ECU is sending a signal, but an erratic one which coincides with what the tachometer is doing.

Seat and ultimately ECU removal was not as bad as I recalled. I think re-install of the seat is the rough part, but ECU is out.
Circa 2015 board, AMP EFI MSPNP (MS3).

It is a cold night, so I brought the ECU in for some inspection and continuity testing.

If nothing sticks out as the problem, plan is to go back and plug it in, play with test mode and see if I can get a stable signal out of the ECU, then maybe bypass the harness and run a jumper wire to the tach input and see if the harness is the problem somewhere.

 

My 95 had a jumpy tach before and after installing a MS2, i cleaned the ground wire connection near the throttle body (typical issue) with no improvement but it was an intermittent issue. 2nd year of running on the MS2 it started to happen more frequently. I found the plug wires at the coil pack had backed out a bit and there was evidence of the connector being hot (black and sooty), replaced the wires, making sure the connections were fully seated in the coil pack and its been good since then. Based on what others have said in another form, some times the boot at the coil pack is glued on too short, making the connection at the coil pack difficult to fully seat, or stay fully seated.
TLDR - Might be worth wile checking your plug wire connections and grounds if you have not.

 

Grounds are good, extra good if I can toot my horn with the rewire and addition of extra grounds.

You may be onto something with the wiring, I peeled back some of the plugs in the engine bay and found them soldered. Also noting only two plugs going into the ECU which seems odd for a plug and play unit, but it does have a separate plug on top. Based on the wiring diagram I found the missing plug is mostly sensor information which megasquirt is reading so it must be coming in possibly from that separate plug on top?

Also found I have a 96-97 ECU (based on the plug and pinout) in my 94 car, running a 2003 engine. Some splicing must have taken place. Going to go with my original plan and test the output and jumper before I start peeling back all the harness. I'm almost afraid of what I would find.

 

Well I did not get to try out my tach wire bypass.

As mentioned I noticed some of the plugs in the engine bay had been spliced to the harness. I decided to peel back the electrical tape on the coil plugs and investigated more based on Houleo’s post. Did not think it was a problem as whatever they did had been holding up. But dear god….



I know that whoever comes after me is also going to criticize my wiring that replaced this, that’s just how amateur electricians are, but AhHhhhh. Joints were cold and some were visibly loose like outer couple strands soldered and you could move the wires.

Cut those all out and replaced with quality naked crimps and marine heat shrink and tessa tape to keep it all together. Could have soldered, but I am more in line with the vibration theory these days and refrain especially in the engine bay unless pigtails are too short.

Wish I tested at that point but I was convinced it would not be the only problem since the ecu was seeing the signal fine and was not triggering the tach with test mode. So I spliced in a wire from the ecu plug off the tach signal wire, sat down to test the car to get a baseline and it started easier than it ever has and tach snapped right to cold idle holding rock steady.

Kudos to Houleo, and thank you to those who offered ideas, will have to see if my original cluster works, but celebratory break time today.

Theory is this must have been exacerbated when I redid the valve cover, probably the manipulation broke some of the cold joints and it was barely holding together with the tape? I fear I'll have to go revisit the other plugs now. Lots of homework in the days to come.