[rusEfi] '99 Miata in Eastern PA #17

[tomiata]

I found this stock timing table for miata 1.8 awhile back, but have not tried applying it yet. Does this look correct?

stock18timing.JPG (104.64 KiB) Viewed 19432 times

[AndreyB]

Is that psi on the Y axis? We would have to convert to kPa.

Looks pretty realistic but I have pretty tiny experience.

[tomiata]

Oh, I see where I found the picture from this thread:
http://www.miataturbo.net/megasquirt-18/lets-discuss-spark-advance-69662/page2/

and a comment there says "The load scale is next to useless however."

Maybe someone can translate the Japanese?

[Rhinoman]

It looks like the load scale is simply 0 to 16 which is the number of cells. If you are tuning a stock ECU with an emulator then the actual scaling isn't that important, you just tweak the value in the cell that you're hitting.

[stefanst]

Stock ECU is using MAF, not MAP, so the y-axis scaling is totally pointless for us anyway. But I would assume that the top row is idle and the bottom row is full throttle (equivalent to 100kpa).

My timing is a little more aggressive up top (see 99.5kpa row). But I am running 93 octane and checked for knock with knock sensor and headphones. Stock is tuned for 87 octane and presumably very safely so.
So for whatever it's worth, here's my map:

timing.png (132.09 KiB) Viewed 19392 times

[AndreyB]

Welcome back

[Rhinoman]

Stock ECU uses separate tables for idle.

[stefanst]

Back from vacation and caught up with most of the other BS. So last night I made some small progress on my Frankenso install:

- Fitted board to my 'new' BPE2 housing
- Modded housing for external USB Type A plug
- Modded housing to add a microphone plug (see below)
- Decided to use the signal wire for the stock MAF sensor for my new external MAP sensor

Microphone plug:
Why add a microphone plug? Knock detection! I have not yet heard of a reliable implementation of automatic knock detection on a Miata engine. The problem appears to be that the head, especially on the NB engines with the solid lifters, is so noisy, that even the more advanced knock detectors can't reliable tell the difference between knock and regular noise. Experienced human ears on the other hand have much less problems with this task. The traditional solution for tuning is the to attach some 'det cans' to the block near the head and listen in. That's very old school and works surprisingly well. Picture of det cans:

My car already has a knock sensor installed which as far as this application is concerned is nothing but a piezo microphone. So if you hook that up to a piezo amplifier and some headphones you can bond with your engine by ear in a more modern fashion. A cheap way to obtain a piezo amp is from harbor freight. Buy their mechanic's stethoscope, modify it by adding a resistor and some microphone wire and Bob's your uncle. Here's a quick video on the mod: https://www.youtube.com/watch?v=t4mZ2FMLC4A. So I added a microphone output to my frankenso that allows me to plug in my HF det cans.

[kb1gtt]

Long term I personally feel we should simply use a PGA to get the signal into the STM's ADC, then software process it. However that requires a bunch of work from russian, who has limited resources. This is why we looked to use the TPIC.

The det-cans are an interesting approach. If I knew the schematic for that device I might attempt to make a board that does it, instead of relying on some MFG who might change the design at any moment. However I also say I might do that design, as I have many items that I'm not getting done.

[stefanst]

I'm suffering from various problems:
- "While I'm At It" syndrome
- severe cold that lasted for 4 weeks

About 4 weeks ago I decided to finally toss out the old five-0 (*) 850cc injectors, so I don't have to tune the Rusefi with old injectors. I already have 650cc Flow Force Injectors in hand. No problem- right? Since I have to remove the upper half of the intake manifold to get to the injectors, I figured while I'm at it, I might as well replace the intake with one that makes a little more power (squaretop). I was completely unaware how bad removing the lower half of the intake is for your knuckles. Took me 3 hours to get the stupid thing off! I was sporting five bandaids on my hands for days!

Once the lower half of the intake was off, I finally found the root-cause of the small puddles of oil that kept collecting under my engine: the 'tee' for getting oil for the turbo (sourced from the oil-pressure sensor). While trying to tighten the connections, the brass stud that connects the tee to the block sheared off- flush with the block. This was trying to tighten it by hand- it just broke right off. I'm very happy that didn't happen on the track.

After extracting the broken off thread from the block, which was quite a nail-biter, I installed a tee that is not made from soft brass- hopefully that will last longer.

Now I'm slowly putting it back together. Fortunately the squaretop manifold leaves a little more space to fit the nuts, so my knuckles look a little better. Hopefully I'll be back up and running after the weekend.


(*) These five-o injectors SUCK. They are EV14s, but I don't know what they did to them. Their dead-times are all over the place. They change with temperature, humidity, time of day, moon-phase and a few other seemingly random variables. My dead-times vary from 1.3ms to 1.8ms. Since I'm idling at less than 1ms, a change in dead-time of 0.5ms severely messes with the Air-Fuel-ratios. DON"T BUY FIVE-0 INJECTORS. DON'T!!!!

[stefanst]

I'm back in the game!

Live happened. Lots of distractions. Trying to sell our house and build our own - I'm shooting for a five-car garage
But after my Miata was sitting in the garage, partially disassembled, for the last 4 months or so, I finally got it back together. Slightly improved even. Now sporting the fancy squaretop intake and the supposedly totally awesome flowforce injectors: http://www.flowforceinjectors.com/. They are Mustang GT500 injectors, flow-matched at a whopping 610cc/min and come with all the necessary HW to install in several different cars.

8:00PM: Car fires up (still on Megasquirt, because: never change three things at a time- stick with a max of two)
8:00PM and 20 seconds: pfshhhhhhhhhhhhhhh sounds from the engine bay. Turns out fuel fittings must be pushed on ALL the way. Fuel everywhere.
8:30PM: garage floor and engine bay are cleaned of fuel
9:00PM: After removing upper part of intake and re-attaching fuel line, the car fires up and runs. No more obvious problems

To do:
- re-tune with new injectors and manifold
- Install external MAP sensor in engine bay
- Discard Megasquirt and run 100% on rusefi

Wish me luck!

P.S.: Somebody, who shall not be named, doubted that the car would be running tonight. To whom I say: Neener neener!

[stefanst]

I *LOVE* the new squaretop intake. With the old VICS the engine really bogged down after about 5krpm. And the VICS is pretty much the best intake available in any NA/NB Miata in the US. With the squaretop it feels like the engine just keeps going and going and going. Look how the power curve keeps climbing, even above 7krpm!

[AndreyB]

Stefan, can you post a two minute log of engine not running on your Miata? Maybe both MS and rusEfi with ExpAverage=0 and ExpAverage=0.3333?

ExpAverage is defined on the bottom of Analog Inputs dialog, ECU reset after changing parameter just to be sure.

[stefanst]

Got some logs off the '99 with MS3 here- didn't have time to plug in the rusefi.
Even with the engine running and no firmware smoothing applied there is no noise worth mentioning. With just a little software smoothing the noise goes away and still the values react to real changes quite snappy. The noise observed is really just oscillations of the smallest step-size. That's to be expected.
Keep in mind that MS3 hardware essentially just uses a classic voltage divider, followed by a low-pass. None of the snazzy Op-amp and diode-clamping we have with the frankenso.

MS3_Running_With_Smoothing.png (20.6 KiB) Viewed 19070 times

MS3_Running_No_Smoothing.png (21.99 KiB) Viewed 19070 times

MS3_Not_Running_With_Smoothing.png (13.66 KiB) Viewed 19070 times

MS3_Not_Running_No_Smoothing.png (24.61 KiB) Viewed 19070 times

[kb1gtt]

I don't know what the cutoff of the MS is. I would guess something like 1kHz with a bunch of variable signal delay. AKA a 1Hz signal has a different phase delay than a 1kHz signal. We also have a low pass filter. It would be R213 and C210 on the Frankenso R0.4 schematic. Our cut off knee is about 150kHz, which is mostly just an EMI filter and high end anti-aliasing filter for the max that the ADC can receive. After that the filtering is a software feature. The STM32 has the brain power, so the hardware is allowing the software to change the filters programatically. You can see more details of our physical low pass and it's lack of signal delay at the below page. Put in 100 for ohms and 0.01u for F.

http://sim.okawa-denshi.jp/en/CRtool.php

[stefanst]

Your guesses at the MS circuits seem to be largely correct:
MS3 filters MAT and CLT with 2k2 and 1.0uF. Since we don't have to have millisecond accuracy here, that's sufficient (cut-off frequency is 73Hz). TPS is filtered with 1k0 and 0.22uF (cut-off frequency is 720Hz). Vbat seems to be an unfiltered voltage divider right off of the sanctioned 12V supply.

[AndreyB]

OMG such a blast driving your car today! Sorry for tire smoke. I believe this is the most power-to-weight I've driven!

[stefanst]

Seems like every time I ride with you, you generate some tire smoke. It must be me then. I promise I'll do better next time!

[stefanst]

Remember how @ wrote in his Hunchback thread that it's important to tighten lugnuts? http://rusefi.com/forum/viewtopic.php?f=3&t=906&start=60#p20813
Today I decided to attempt empirical verification of his theory and <drumroll> he's right! If you leave the lugnuts on even just one wheel untightened, this wheel will come off and, if it's a front wheel, steering problems ensue. Quite embarrassing. Even more embarrassing is that EVERYBODY passed me while I was parked trackside. @ was right behind me and saw sparks fly. Will need a gopro to record these things.
Damage seems minimal, we even found all four lugnuts. But I did decide to not track the car for the rest of the day. We just decided to share my backup car that @ was driving which worked real well.

In other news: The Englishtown, NJ road course is decidedly meh. It's short, not much in terms of safety zones and it's super bumpy in some areas. My teeth are still chattering from how rough this track is.

[AndreyB]

Will need a gopro to record these things.

my racing videos are taken with a cheaper https://www.amazon.com/gp/product/B012I6GPOS

Hold on, interesting... Ordered on April 20, 2016 $75.90 - Amazon price today is higher but if you google Xiaoyi Yi Action Camera the right price shows up.

[kb1gtt]

When I was a kid my dad was driving us along, we saw a tire pass us and he commented that some sap lost his tire, we all watched it for a couple seconds, then the back of the car fell at a tilted angle as we hit a sight turn in the road which shifted the weight. It was funny, but scary at the same time.

Yes you need to get the camera thing figured out, such that those of of on the interwebs can get joy from your pain

Perhaps we should develop lug nuts with a pressure sensor in them such that if you don't have pressure on the lug nut threads it sends wireless signals to the ECU just like your tires

[AndreyB]

I've cleaned up NB1 trigger shape definition a little bit - if you ever migrate to current version you need to set global trigger offset to zero (previous value of 276 moved from engine declaration into trigger declaration)

[stefanst]

I just downloaded the latest build and found all kinds of intriguing changes/additions having to do with cam, crank triggers, Miatas and VVTs.

[stefanst]

Please allow me to revive this thread!

So far rusefi integration for Miata seems to be progressing real well with NA6 and NB2.
And if we can reliably run NB2, we can run NB1.

It's time to sell my Megasquirt for big bucks and replace with rusefi.

Since I will be moving to NJ, where the rules regarding emissions are just plain evil and you have to have a working OBDII plug, I was thinking that maybe running the rusefi as a piggyback may be the way to go. Basically give the stock ECU all the inputs and let it think it's running the show, but behind the curtains rusefi does all the actual work.

Things to consider here:
- Bias resistor for CLT: The stock ECU has a built-in bias resistor for CLT already, so I need to remove the one from rusefi
- Bias resistor for IAT: No problem, since I have two IAT sensors- stock and one after the IC for actual intake air temp. Stock feeds stock and second one feeds rusefi
- TPS: Is a voltage divider, so no problem
- O2 sensors: my Innovate LC1 will simulate an analog signal for a narrowband O2 sensor which I can use for primary. Just needs a resistor to simulate the heater. Secondary can stay stock, since rusefi doesn't use it at all.
- Crank/ Cam sensors: The stock ECU has a pullup to 12V if I'm not mistaken. So we'd need to remove the pullup on rusefi and that should do the trick. Both input circuits- stock and rusefi- are fairly high impedance, so this shouldn't be a problem at all.
- Outputs: No issue: rusefi handles all the outputs that are in any way important. Let the stock ECU deal with evap canisters and all tat emissions stuff
- VICS and EGR: I think I'll need to reinstall the EGR pipe. Damn. Also: I don't know if the stock ECU monitors the functioning of the VICS butterflies. Will have to find out. Since I don't have them anymore this may be a problem....

Anything anybody else can think of?

[AndreyB]

Anything anybody else can think of?

With all the op-amps isolation I was running rusEfi as data logger on the hunchback - and stock ECU was happy. Same exact thing as piggy back at least from the sensor side.

[kb1gtt]

I think swapping the ECU when on the road, vs track is going to be allot less labor.

If you try to do them in parallel You'll want to keep the wires between the 2 ECU's short. When you make Frankenso high impedance on things like temp sensors, you'll be more susceptible to noise. Short wires will help remove the antenna.

[stefanst]

In order to have stock ECU run the engine, I will need to at least swap the injectors back to stock, re-install the MAF and wire open the wastegate.

I was thinking about stacking two ECU housings on top of each other and install in stock location. That should keep the runs below 4". And I would wire from stock connector into rusefi and from there on to stock ECU, so the only distance with low current would be on the frankenso board.

Edit: re-reading this makes me realize that I will have to reinstall the MAF even for piggyback operation. Hope I still have it somewhere...

[stefanst]

It looks like the VVT Miata config is more up-to-date than the standard NB config. So when I set up the NB1 (non VVT) should I just use the VVT engine config?

[AndreyB]

Pre-defined configs are not sacret and should not be taken too seriously. Yes set engine_type 47 is pretty good, but the right way would be to take it, fix it for NB1 and hard-code that end result as the new NB1 default config.