Just a quick heads-up: Maxim's new MAX9924 chip (and its two-channel partner, the MAX9926) is utterly awesome.

Suck on this, LM1815:





Holy noise floor, Nocturnal Echolocating Flying Mammal Man!





Money shot:

 

i totally havent a clue what im looking at, lol

 

New? Welcome to last August

I would feel more smug if I had actually gotten it installed in the car by now.

 

Geez.. I wish I were more electrically inclined like you guys. Looking at these threads make me feel stupid

 

One day I will understand this.

 

Interesting. Which operating mode did you try? Was this on the bench or on the actual engine?

Dimitris

 

Since Joe wasn't nice enough to actually link to the site, http://www.maxim-ic.com/quick_view2.cfm/qv_pk/5822/t/al

Datasheet: http://datasheets.maxim-ic.com/en/ds...24-MAX9927.pdf

Maxim does make some pretty cool ICs for all kinds of stuff.

 

Huh... You'd think I'd remember that.

I didn't realize they'd been out that long. When I ordered mine a month ago, none of the usual distributors even had them in the catalog. Had to sample 'em directly from the factory in Durkadurkastan. I believe that DigiKey now has them in stock.

Turns out that Jean (of JBPerf fame) is also using this device in his new decoder circuit. And as a happy coincidence, it also works with a 0-5v input, so it's great for stock Miata sensors too. I'd probably recommend the 9926 as the standard input decoder for any MS build.

There are just so many awesome features here. For starters, this chip has a differential input, so any common-mode noise on the line is moot. None of the other circuits, including the 1815 and the stock MS3.0 circuit, have this feature. You could probably run unshielded wire to the sensor, though I haven't tried this. (Mine was already wired with Belden 9451, which is two conductor plus shield.)

And the adaptive threshold detector actually works. The big problem I had with the 1815 was that it would miss teeth if there was a major change in amplitude from one to the next. As it turns out, the wheel I bought from DIY wasn't quite perfect. A problem, I suppose, with laser-cutting in general. The teeth weren't all the same height, so the amplitude of the various pulses varied a bit. Here's an example of what I'm talking about:




I later put the wheel into a lathe and trimmed down all the teeth so that in addition to being true, they have a 2-3mm flat section at the tip, rather than a point. I'd suggest this modification to anyone considering the use of DIY's 4" wheel. The larger wheels already have flat teeth, though they also might not be 100% true.


In that first scope trace in my original post, you'll notice that the amplitude of the various pulses varies greatly from one to the next. The 1815 choked on this and would miss the smaller pulses. The 9924 has no problem catching them.



The one downside is that the chip is a tad small. Here's a picture of it:





Now, unlike kday, I didn't really want to spin a new board to put this on. Proto-Advantage to the rescue:



Any fool who happens to have access to a Leica M50 laboratory-grade varizoom stereomicroscope and a Metcal SMC rework station can easily attach the chip to this carrier board, which then turns it into a standard 10 pin DIP. From there, you can just mount it on a standard perfboard and build the rest of the circuit around it.

The required parts count is surprisingly low- much simpler than the stock VR circuit on the 3.0 schematic. Just three resistors and one capacitor, plus a few power supply filtering caps. No trimpots or other BS adjustments. You just built it, throw signal at it, and it figures out the rest.




The first and second captures in my original post are of me waving a screwdriver in front of the sensor on the bench. I was deliberately trying to produce the ugliest waveforms imaginable, and the damn thing just grabbed 'em and did its job every time. The third capture is of the finished circuit on the car, with the engine running at idle.

I'm running the chip in mode A2, and my circuit is a copy of the A2 reference schematic.




Heh, don't worry about it. In all of these images, the upper (green) line represents the signal coming out of a VR crankwheel sensor. That signal goes into the chip and gets converted into a squarewave, which is the lower (yellow) line. That squarewave signal is what goes to the CPU. So the key points here are that the chip needs to make a pulse on the output line every time it sees a wave on the input line, and it needs to do this reliably, even if the input waves are small, distorted, noisy, etc. In the second trace I posted originally, you can see that the input waves are barely even visible on the scope, yet the chip detected all three of them perfectly.

 

I would bet $100 I could solder that chip to that board with a radio shack iron covered in oxidation in under 2 minutes.

I used to add memory to xboxes from borked boards to double it up for people using emulators and linux. Used RS iron and a flashlight to detect jumped legs.

In any case, its def a cool chipz0rz.

 

This fool uses a cheap weller temp-controlled iron and a $5 watchmaker's loupe. It seems to work OK down to about 0.5mm lead spacing, which is pretty much the limit of my PCBs anyway.

I'm pretty sure mine is running in mode A2 as well, since A1 needs an external bias voltage and those are the only two modes with the adaptive peak threshold.

 

So Joe I need This and This correct? I am thinking of picking up a few pieces of each to build.

 

Joe, you reckon these would play nicely with MK2 sensors as opposed to ABE's doppelganger circuits?

 

Heh. At home, I've just got a plain-jane WES51 soldering station, and I squint a lot. Been meaning to get a decent magnifier, just haven't come across one I liked which was also cheap. But at the office, well, if ya got it, flaunt it, right?

Yeah, A2 is definitely the way to go. Based on Jean's experimentation, it seems to auto-set to 2.5v when fed a single-ended 0-5v signal, and it's certainly working properly with a balanced VR sensor. I can't really fathom why anybody would need to operate in mode A1, much less B or C.



The 9926 is a 16 pin QSOP device, so you'd need a different board than the one I used for my 9924. Here is the correct unit for that: http://www.proto-advantage.com/store...cts_id=2200032

Pay the extra $2 and get it with the pins assembled. I also just noticed that for $8 you can get it not only with the pins assembled, but also with the IC mounted. (I assume you send your IC to them and they tack it on.) Might be worth the extra money if you don't have an uber-fine soldering tip and/or a good magnifier. If I didn't have the Leica and Metcal here at work, I'd probably take 'em up on it.



Oh, I'm sure I could also. Of course, all five pins on each side would wind up being commoned together.

Aaah, memories. I was in college when the original Playstation came out, and my roommate and I installed modchips on them in exchange for beer.

 

My gut feeling, based solely on Jean B's observations, is that yes, a 9926 is probably an excellent solution for an NA CAS or an NB's crank and cam sensors.

Here's a link to the discussion thread about this device where Jean shares his plans and observations: http://msextra.com/forums/viewtopic.php?f=67&t=35149

Using his board, you could bypass the MSs opto-based input circuit entirely and go directly to the relevant CPU pins.

 

Interpreted, and TX recommendations.

V-tach. Same as above.

 

Cheers Joe,

As much as I enjoy building stuff, my MS board is becoming a PITA, that Dual VR board looks just the ticket

from the basis of that thread you linked to, as the stock 5's sensors are Hall they don't 'go' negative, so you just don't connect anything to the '-' pins and the chip sorts it's self out?

 

I got a decent Pace machine at my work. I figure I buy a few of everything and get some more experience soldering things. Then I will pester you on how to build the rest of the circuit, cause I am a dumbass with understanding circuitry.

 

Hahahaha! I hope I never produce an output that looks anything like that.

You a medic? A doc?




Yeah, for those not interested in rolling their own, I think Jean's new board is going to be the hot ****.

I'd guess that with a single-ended sensor, you'll want to ground the (-) input and connect the sensor's output to the (+) input. Grounding unused inputs is typically done to keep them from floating around. You will also need an input pullup on the (+) side, outboard of the current limiting resistor.

However I am certain that Jean will give directions that are specific to the design of his circuit, once the new boards are released.

More info on Jean's new board: http://forum.jbperf.com/viewtopic.php?f=6&t=78


edit: looks like he now has the boards available for sale. Not on the website yet, but if you contact him, you can buy one direct. $45. Cheap.

 

I've been considering getting a couple 9926's (or a 9926 and a 9924) to condition the Ne+ G1, and G2 sensors on a toyota (20v 4age). What I have there now works fine, but takes longer to sync on on crank than I'd like.

The circuit on the MS3x board works fine too, and the engine starts pretty quick, but I'm curious to see how this new chip does.

For those who don't know, B&G are also working on the "intellitach" which takes differential inputs and does some other conditioning like this max chip and then run the resulting signal through a DSP sampling at once per uSec where they subject the signal to a median filter (and potentially other stuff) and output a square wave at zero crossing.

It's pretty cool stuff; you can hook up a USB cable to it and view the signal from the VR like you're looking at it on a scope. The only problem is that I have no idea when they're going to sell them!

Ken

 

And if they are going to sell them; the synchromap is at least 2 years old and hasn't made it there yet, your new MAP code is probably the reason.

Dimitris