Cherry hall sensor RPM spikes

[ZVOLV]

Yeah airgap could be wrong. Need to put a labscope on it. have seen build your own scope kits for $40. Go build one.

Oh, if your sender gets "beat" by the trigger wheel. Airgap might be wrong. hahaha

 

[142 guy]

My wheel is .25" thick, and I was using that cherry sensor that you listed above. I had no tach signal issues until the sensor itself started getting beat up by the trigger wheel.

So begs the question - if everything was just fine with the Cherry sensor until it got mauled by the toothed wheel, why would you not go back to the Cherry sensor and address the issue with the sensor mounting (or whatever caused it to get beaten up). If everything really did work just fine with the Cherry sensor, its not a firmware / software / configuration issue and probably not a noise issue (at least noise coupling to the signal line between the sensor and the MS).

 

[gross polluter]

So begs the question - if everything was just fine with the Cherry sensor until it got mauled by the toothed wheel, why would you not go back to the Cherry sensor and address the issue with the sensor mounting (or whatever caused it to get beaten up).

Agreed.

This is why I asked if there was a picture of the sensor mounting. See what we're working with.

 

[cwdodson88]

The sensor in question.

The reason for swapping to the Tabbed DIY sensor was that the threaded Cherry sensor was difficult to set the gap and the gap requirements were smaller .5mm to 1.5mm compared to 1mm to 4mm, I figured that since the characteristics were similar it wouldn't have been an issue. Mounting bracket is almost identical to the one I used before, just further away from the wheel, tooth lineup was the same, and the sensor itself is more centered fore and aft of the trigger wheel.

 

[142 guy]

The sensor in question.

The reason for swapping to the Tabbed DIY sensor was that the threaded Cherry sensor was difficult to set the gap and the gap requirements were smaller .5mm to 1.5mm compared to 1mm to 4mm, I figured that since the characteristics were similar it wouldn't have been an issue. Mounting bracket is almost identical to the one I used before, just further away from the wheel, tooth lineup was the same, and the sensor itself is more centered fore and aft of the trigger wheel.

Your images didn't work!

If you are running the sensor with the maximum gap listed by DIY, you may have the answer to your problem. Cherry lists a maximum gap of 1.5 mm and Honeywell lists a max gap of 2 mm (and a much lower test speed!). In the DIY test video of the sensor, the gap between the sensor and the teeth is no where near the maximum of 5 mm that they are listing on their website. At the very end of the video, it looks like the gap is set as tight as possible without having contact, which would be consistent with reliable performance.

 

[cwdodson88]

 

[142 guy]

Going back through your posts, I see you had the gap set to 2 mm. Pertronix aside, I am not a hall user; but, your installation looks reasonable. Based upon DIY's data the clearance should be good. Kind of leaves you with the conclusion that you should go back to the Cherry sensor in an equivalent tabbed mount if you want to stick with that style mount.

 

[cwdodson88]

Not what I wanted to hear 142guy, not what I wanted to hear.... but I'm going to try a few more things tonight. if all this fails, I'll just have to break down and install a DSM CAS.

 

[gross polluter]

That sensor bracket is concerning as well.

 

[cwdodson88]

any suggestions on a bracket?

 

[carbonmike]

Something like the bracket from this kit could be made with hand tools from a chunk of aluminum or nylon. Or you could pony up for the $1200 kit

 

[gross polluter]

any suggestions on a bracket?

If you can weld on it, Welding some gessets at the ends of the angle would help significantly. Even though it seems sturdy enough by feel, 4cyl engines have some considerable vibration that'll exploit the weakness in even the strongest brackets.

 

[cwdodson88]

That electro motive, lol, looks like an overpriced microsquirt.
I'm not able to weld on aluminum. How about a brick of aluminum 1" x 1.5 or 2"?

 

[bobxyz]

Try setting IgnitionOptions - Wheel Decoder Routine to "025 Style" I'm sure hoping it's set to 024s9 Style, which appears to have very little tolerance for extra tach pulses.

[There also looks to be a code bug in 025 where the ~2x missing tooth period is included in the average tooth time calculation without dividing it by 2 first, but this is still better than the 024s9 style that doesn't do any averaging and uses simply 1.5x the previous tooth's time to determine that the current tooth is the "missing" tooth on the wheel. If you get a noise spike in the 024s9 style, the next standard tooth can pass the 1.5x criteria and be mistakenly declared as the "missing" tooth.]

For logging, can you run with a Tooth Logger interval of 500ms or 750ms, instead of 250ms? I'm guessing that the logging code isn't interlocked and that too frequent of an interval can sometimes result in a single screenful showing both new and old samples from multiple different non-sequential revs.

If the logging and tach signal are working correctly, you should always see 2 or 3 long teeth per screen. [It looks like a screen is 93 events and, with a 35 tooth wheel (1 missing), it will show ~2.6 revs. Depending on starting time/tooth, it can cover 2 or 3 missing teeth.]

 

[gross polluter]

That electro motive, lol, looks like an overpriced microsquirt.
I'm not able to weld on aluminum. How about a brick of aluminum 1" x 1.5 or 2"?

That would work fine. Basically, if something can flex it will when attached to an engine. A sensor bracket that's overbuilt will hold the sensor true regardless of operating condition. That'll allow you to run the sensor with a closer gap without trainwrecking the sensor.

 

[Redwood Chair]

 

[cwdodson88]

Try setting IgnitionOptions - Wheel Decoder Routine to "025 Style" I'm sure hoping it's set to 024s9 Style, which appears to have very little tolerance for extra tach pulses.

[There also looks to be a code bug in 025 where the ~2x missing tooth period is included in the average tooth time calculation without dividing it by 2 first, but this is still better than the 024s9 style that doesn't do any averaging and uses simply 1.5x the previous tooth's time to determine that the current tooth is the "missing" tooth on the wheel. If you get a noise spike in the 024s9 style, the next standard tooth can pass the 1.5x criteria and be mistakenly declared as the "missing" tooth.]

For logging, can you run with a Tooth Logger interval of 500ms or 750ms, instead of 250ms? I'm guessing that the logging code isn't interlocked and that too frequent of an interval can sometimes result in a single screenful showing both new and old samples from multiple different non-sequential revs.

If the logging and tach signal are working correctly, you should always see 2 or 3 long teeth per screen. [It looks like a screen is 93 events and, with a 35 tooth wheel (1 missing), it will show ~2.6 revs. Depending on starting time/tooth, it can cover 2 or 3 missing teeth.]

I'm running 025 style, so I guess thats the best option I can get with ms1 as far as wheel decoding.

As for logging, the extra missing teeth and/or too many teeth between triggers is only showing up under 250ms, it doesnt happen at 500, 750 or 1000ms.

Redwood - I had gone back through Dan Haili's build and saw that bracket he did, I'm going to try to get over to the scrap metal shop and see if I can find a hunk of aluminum to try. I'm just going to drive it as is until I can tackle more this weekend.

 

[bobxyz]

The tooth log captured by TunerStudio with a setting of 250ms is garbage and does not match what's actually happening. This explains why your logs look so unbelievably bad. Please try some 500ms or 750ms logs and see if you see anything other than the correct 36-1 pattern. It may be hard to capture a glitch, but it would help confirm if there really is a spark/injector induced noise problem, or if the hall sensor is dropping out.

Can you also try the 024s9 setting? It sure would be nice to find something that definitively changes the behavior.

 

[cwdodson88]

The tooth log captured by TunerStudio with a setting of 250ms is garbage and does not match what's actually happening. This explains why your logs look so unbelievably bad. Please try some 500ms or 750ms logs and see if you see anything other than the correct 36-1 pattern. It may be hard to capture a glitch, but it would help confirm if there really is a spark/injector induced noise problem, or if the hall sensor is dropping out.

Can you also try the 024s9 setting? It sure would be nice to find something that definitively changes the behavior.

I've grabbed several logs at 500 and they all seem fine, maybe 1 or two with hiccups, but it seems that may just be from trying different sensor gaps. I forgot to log what was what in my notes, so I'll do some more investigating this weekend after I make up a new bracket... we'll call this one Hall Bracket V3.0

 

[ZVOLV]

Why not just use a hall sensor built into a distributor? Like from a 75 240. Just an easy guaranteed squarewave and you won't have to mess around with airgaps.