badinfluence

Also read the whole thread and noticed you had some similar issues. Make sure you check out the Canbustriple if you havent. I also think you should break off your functions to either multiple ATTiny8
5s@16 or 84s if you need more outputs/inputs. Also an isolated 5v supply and optocopulers and SSRs might help with the noise and errors in the general setup of the car. ( the alternator and voltage regulator make lots of noise and the starter current wears the modules out) Multitasking with the arduino is supposed to work, but you absolutely have to calibrate registers and it gets annoying. AVRs are more flexible and if you run your ATTINY at 3.3v the outputs are 3.3v. If you optocouple or isolate the I2C you can run a MCP23008/23016 or the spi version MCP2S008. That will keep your avr snacking on a AA battery and the MCP taking a hit. It also would work with multiple Chips if you wanted to swap chips for fun testing profit ocd or a sanity check.

The ATMEGA32U4 and 328 are an option, but attinys are more flexible for voltage changes and cheaper....

logalinipoo

holy crap and I tought I was doing good to get a light blinking with them.

Harlan

iTrader: (3)

If that was the problem I would have gone down that route. However it was hit or miss without the PWM enabled or those pins even connected to anything. It seems to be some sort of issue with SPI implementation, and rewriting libraries is beyond my patience right now. I could have tried a logic level shifter, but between the resistors required to trim 5 volt sensors down to 3.3 volts and the logic level shifter it made more sense and took up less breadboard real estate to just add a 5volt logic chip.

Your right there are multiple ways to do this and plenty of options out there. I've almost got this one working right, but next version might have some changes. I've never had problems requiring optocouples for two chips on the same board with the same power supply and ground, but I guess it could happen. Right now it appears to be an issue with the SPI and WIRE libraries and how they are implimented for the teensy. Everything works fine on the arduinos.

Also networking multiple ATtinys is kinda silly because of the pins you lose for networking, easier to just start with a Atmega328 if it has all the pins you need in the first place. That's why I'm using the Arduino nano. It has the Atmega328 with a nice form factor and a usb connector for programming.

By now I could have gotten away from the Arduino platform and just started programming the chips directly without a bootloader, but again it's above my level of patience or time. At this point I'm so close I just want it to work.

badinfluence

Shifters are easier to make than trying to get a chip for them. The Chip just saves space really. Resistors do not work for I2C very well because it is a pissy protocol. https://www.adafruit.com/products/757
Arduino Playground - I2CBi-directionalLevelShifter

I would make your own and test it for a few bucks. The optocoupler also would allow for the same shifting with a completely isolated power supply and zero interface ability to the RC on the uC. How the reset and RC is doing dictates how the uC operates. No matter how much Teensy or Arduino buttons it up and makes it look pretty, the chip still has hardware level demands that need to be met in AVR automotive type scenarios. On Digikey you can buy AT controllers with Automotive AVR stability, and save a headache. The noise rejection is what you really want, unless you do the math to find the correct ceramic caps to kill the noise you show on the scope while driving. (Not safe....)

I wasn't talking about networking them. the MCP230017 does all the work really, the microcontroller just give the orders. So 4 Tiny85s at 16 or 20mhz wwith an 8 pin package fits 4 microcontrollers than run on 1.7 to 5.5, so you could avoid level shifting all together. The 85s are programmed through SPI pins also so just make a header and a reset switch. If they talk to all the devices through I2C, what microcontroller you use doesn't matter really. You could even use a DPDT relay and switch it with the push of a button.

For 87 cents I would try the optocouplers, spending 2 minutes and 5 dollars is worth saving hours in the field.

Another thing is Teensy is trying to hard to distance themselves from Arduino, and using sales/marketing bullshit to do it. The 2 is just an ATmega32U4 with different software.


I use the ATTINY10s SOT23-6 (bottom), with the 85 8 PDIP(top), programmed with an Arduino Uno ISP and the 6 pin AVR/SOIC port.




What your doing doesn't take much power, but multitasking with the software is annoying because even the Teensy programmer doesn't really like it much. Better 2 uCs running separate code than 1 overloaded uC that ***** itself right as you run out of WMI and boom goes the engine.

If you get worried about that in particular, I use the AEM WMI Failsafe Gauge and a tank float switch. Devil's own sells a good float switch.



EDIT: Also found this for 3.3v MAP sensing via I2C.

http://www.digikey.com/product-searc...030PG2A3&v=480

And 5BAR via Analog Voltage
http://www.digikey.com/product-searc...004BGAA5&v=480

Harlan

iTrader: (3)

Here is what my current system does:

Takes 6 0-5volt sensors and makes them available to the CAN network.

Reads a TI knock sense chip via SPI and makes that available to the CAN Network.

Fails open the wastegate if knock limit is exceeded.

Reads RPM and MAF off the CAN network.

Takes RPM, MAF and two 0-5volt MAP sensors to control wastegate operation with PWM.(Currently proportional control, may add an integral component later.)

Takes MAF to control a water injector with PWM.

Takes a 0-5v fuel pressure sensor and a 0-5volt MAP sensor to generate a PWM for fuel pump control. (Written but not implemented.)





The problem has NEVER been multitasking. A single ATMega328P can handle the processing, the problem is having a smooth CAN interface as well. And splitting function to multiple chips only makes sense if the functions are truly independent. Also I wasn't referring to using resistors for shifting i2c voltage. I was referring to using resistors for shifting 0-5v analog inputs.

Using a MCP23017 makes sense if you have a lot of digital inputs or outputs. I'm dealing with all analog and PWM. I could use an servo driver chip and an analog input extender then connect it all with i2c or I could use single Atmega 328P and put fail-safes in the code.

Right now I'm kludging some serial communication between the chips. Even doing it the sloppy way it still gets a refresh across in <100ms which is more than good enough. At this point the only vital input is MAF, and updating WI rate 10 times per second is good enough.

Also fixing a oil leak on the turbo feed line. I get tired of feeling like 007 with a smoke screen coming from under the hood.

badinfluence

[QUOTE=Harlan;4718475]Here is what my current system does:

Takes 6 0-5volt sensors and makes them available to the CAN network.

Reads a TI knock sense chip via SPI and makes that available to the CAN Network.

Fails open the wastegate if knock limit is exceeded.

Reads RPM and MAF off the CAN network.

Takes RPM, MAF and two 0-5volt MAP sensors to control wastegate operation with PWM.(Currently proportional control, may add an integral component later.)

Takes MAF to control a water injector with PWM.

Takes a 0-5v fuel pressure sensor and a 0-5volt MAP sensor to generate a PWM for fuel pump control. (Written but not implemented.)





The problem has NEVER been multitasking. A single ATMega328P can handle the processing, the problem is having a smooth CAN interface as well. And splitting function to multiple chips only makes sense if the functions are truly independent. Also I wasn't referring to using resistors for shifting i2c voltage. I was referring to using resistors for shifting 0-5v analog inputs.

Using a MCP23017 makes sense if you have a lot of digital inputs or outputs. I'm dealing with all analog and PWM. I could use an servo driver chip and an analog input extender then connect it all with i2c or I could use single Atmega 328P and put fail-safes in the code.

Right now I'm kludging some serial communication between the chips. Even doing it the sloppy way it still gets a refresh across in

badinfluence

What the **** did the thread do with my reply?

(Here it is)


What I2C and SPI rates are you achieving with your setup? I don't have any idea how hard you can push those and I would like to know for my project. I think in some areas our projects will align.

Ah I misunderstood. Somewhere in the thread (I read the whole thing Sunday), you referred to the Arduino struggling with running a few items at once in the sketch. It sounds like that isn't the case anymore.

The MCP23008 or MCP23S08 is the 8 I/O version, but my understanding WAS they are Digital and Analog. Someone mixed up the A for Analog and the A for Address on the forum I was reading and I got bad info. It does do PWM on the GPIO according to some application notes and the interwebz

I have been looking at SOT23-6 I2C ADCs (and DACs) for some time for a I2C or CAN network setup for my projects... I have found that most of their designs are cheap enough to generically design the entire board, and be able to hook up multiple things to the I2C bus, including a Raspberryi Pi2 for changes, info logging, GPS integration, and HUD.

Harlan

iTrader: (3)

Never measured the i2c rate, all I know for sure is that it was negligible overhead transferring a couple dozen bytes back and forth on every iteration. SPI has never been a bottleneck either, but something about the teensy implementation is off.

What I can say is that the teensy handles the CAN bus like a beast. Once I had filters set up it never missed a read request and the bottleneck became my Bluetooth CAN adapter. I would highly recommend it over the MCP2515 based interfaces because it can handle responses natively.

badinfluence

I have been looking at CAN native Atmels, and there is quite a good selection with native AVRs for 8bit and even better for 32 IF you can get them soldered to something.

I have a 3W laser cutter I am constructing for PCB creation that will hopefully end my endless fabrication and need for adapter boards.

I really think you should check out that Canbus Triple. It may be possible to run the Teensy software on that setup since it mirrors the 2/leonardo. Or just use the Bluetooth integration with a UART or Binary interface.

Harlan

iTrader: (3)

Oil feed line is fixed. And I'm swearing off cheap Chinese turbo oil feed lines.

OMP issue has reoccurred. At this rate you may need to start calling me 4k RPM... It's definitely a connector issue, but I can't reach it to solider the connector pins with everything else in place. It stays in limp mode until I reseat/clean the connector then it's fine for a while. I'm looking at either bypassing the connector or running a new plug from a different connection.

Still need to fix the hardware, but I've been busy with other projects. When the OMP issue is fixed I'll go back into boost.

Going into an outage at work, so won't have much time, but will have a bit of extra cash. Looking into either 3d printing my intake parts or having a shop weld up an aluminum intake for me. Also I eventually want to put more cash into a better turbo.

badinfluence

3D Print? As in the thing that Melts plastic to make objects on a XYZ plane? Umm Heat?

Even the newer resin based printer objects break down in heat. I would go for a CNC if your looking for that route. It was be cool as **** with all the benefits of your 3D printing.

You could even cad it to be modular so it is easy to install and uninstall.

If you really want to be awesome find the expansion of the given metal and leave the clearances so it can be assembled, but as it heats up, it meshes together till it cools down and needs no gaskets.

Harlan

iTrader: (3)

ABS should be ok for non pressure (preturbo) applications. For pressure nylon is a better choice. I didn't say I was committed to it, just that I was looking into it. If I just get a mock up in plastic then have it made in aluminum it's a win.

Also the stock intake manifold will hold at least 17psi without breaking. Just FYI.
Edit: Also from the same log, highest IAT was 135F. With zero knock. It would have been massive power if timing hadn't gone way retarded and AFR hadn't leaned out. The lean out was because it wasn't tuned, the timing was because I knew it wasn't tuned. As it was it just bogged had a momentary high EGT. Spike lasted all of 1.5seconds. So I'm getting the software right before I try again.

Harlan

iTrader: (3)

Ok, here is a log from my drive to work yesterday. You can see boost hitting at under 3000rpm. Also you can see the spikes I'm talking about, and after going through with a fine toothed comb I'm convinced that it isn't a hardware or software problem.

The issue is the wastegate is trying to open way too late, and the system spools too fast for it to catch. I'll try adding a differential component to my boost controller and try limiting the max signal to the WG solenoid and see if that works.

Oh and for the knock reading, based on previous runs anything under about 200 is ok, and anything under 50 is perfect. I'm not really worried about EGT or AFRs at this time. More tuning to come.

Gonna be a few days before I get back to it, it went into limp mode again this morning and had to drive home at 55MPH.
Edit: OR I can up the premix and bypass CEL in ATR.

And still tracking down the cause of boost spikes. Might be a hardware issue after all...

logalinipoo

Watch out man, you're getting awefull lean, but I guess that's the true test of your water injection. It's nice to see you building boost that early. Hows your cap pack working?

Harlan

iTrader: (3)

I took the caps out and balanced the bank a while back. No real difference from a battery except the car starts a little quicker and less fluctuation on voltage. Kinda forget it's there.

Yeah it is lean, I don't want to tune it too much until I get the fuel pump replaced and the controller for it working. If there is not detonation I'm not worried. Our understanding of proper AFR is pretty limited, and I plan on challenging some of the widely held beliefs.

Need to get a good baseline of knock vs RPM so I can reference WI and boost limiting from a known value of knock.

logalinipoo

I agree, but our knowledge is largely based on piston and a lot of rx7 experience also. Best power should still be In the upper 12_ low 13's. Your almost 14 that I saw.

I like where your going with it. I'd love to see some timing play. With less overall timing and less split. I think there's a lot of power to be had with that combination.

I'm going to play a bit on my end but once it looks like your knock stuff is going good I'm in.

How much water do you carry for injection? Is your boost controller arduino based? Common code or custom?

Harlan

iTrader: (3)

Yeah 14 MIGH be leaner than peak power. 12-13 is a good number for a piston engine, but it's probably not the best for a rotary. Because reasons...

I'm running 25% water to fuel as the default, but I can select it from the Torque App with an OBD2 command. I can also select my 4k and 8k rpm boost targets. So as soon as I get the boost spike issue cleared up I'll run some back to back injection showing IAT, knock and EGT differences. The amount of water commanded is in the log in ml/min. The injector I have right now flows 800ml/min at 80% duty cycle. My tank is >3 gallons, but don't quote me on the exact capacity.

All the control is done on an Arduino nano, but the can interface is on a teensy. It's all custom code by me, but I've put most of the pieces on the forums already. I'll post all the code when I get home.

Harlan

iTrader: (3)

Here is my current code. It has a lot of old unused variables and is otherwise very kludgy.

Disabling the OMP CEL Codes did not prevent limp mode... So I got the harness side of the OMP connector pulled up high enough to split the pins out and soldered them. Couldn't do the OMP side of the connector, but if/when I get the exhaust manifold off again I'll try. this fix worked permanently on my power steering to I'm hopeful. I think it comes down to corrosion where the wire is crimped.

Still tracking down the cause of boost spikes. I'm starting to look for problems in the electronics.

Harlan

iTrader: (3)

Think I found the boost control issue. Grounding issue on the Mosfet board. Also found some erroneous behavior with the boost target code. Will be testing tonight.

Edit: Test is Satisfactory. Fixing boost lockout issue and getting a knock baseline at 0% water 0 Boost to 35% water and 0 Boost.

Edit#2: Boost lockout issue is fixed. Found the ID I used for requesting data from the ECU was wrong, must have had a typo at some time and error carried forward. Only showed up when I wasn't logging or the OBD2 scanner skipped a beat. Also got knock logged VS RPM and verified I was getting near 0psi with the WG failed open. More to come!

logalinipoo

Have you ever pulled the intake apart shortly after a boost run with a lot of water and checked to see if you have pooled water in there?

Harlan

iTrader: (3)

Only time I've ever found water pooling in the intake is after a malfunction. There is enough air flow to sweep the water through the intake when it is spraying at correct times and flow rates. My biggest concern is water pooling pre-turbo and then slugging into the turbo causing blade damage but even that is drawn in quick enough to prevent problems. So far so good.

Edit: And the OMP put it into limp mode again. I'll be pulling the downpipe here in a few days to re insulate and add some v-band gaskets, and I'll fix the OMP at that time.

badinfluence

Sounds like interface bullshit that can happen sometimes. Have you considered a diode to ground and that optocoupler I suggested? The diode is supercheap, and might save the fet if you ever have some kind of issue with voltage cross that happens with Nchannels. I always put a diode from ground to source (- to +) so it can flow back and save the diode, but you might want a diode from fet to ground to keep voltage off if you lose your ground connection. Some people use a resistor from Gate to ground on the 510s. Also you could make a LED circuit for your fet and source to ground that would warn you of your ground cutting in and out.

Harlan

iTrader: (3)

Mosfet ground was just inadequate, fixed now. OMP issue was a no strain relief on the wires to the sensor extension I added it's now been fixed. It had been hanging on by a thread for a while.

Fuel pressure sensor is in and should be reading for logging. Code is in for fuel pressure control, and should be testing it soon. It's set to maintain 68psi at the injectors until 0psi manifold then ramp with manifold pressure maintaining 68psi across the injectors. We'll see what it actually does...

More updates when I get more time to work on this. Almost have the system solid enough to tune.
Edit: Whoops... Should have done the actual fuel rail pressure of 58... Now I have to redo math. But just got my first readings from the fuel pump pressure sensor.

Harlan

iTrader: (3)

I had another slight air filter malfunction, so I now have oiled air filter in front of the radiator, and life is good.
Fuel pressure control has been tested and will be going in with the new pump. Logged some pump pressure dips which were probably causing the lean outs. Might get that all done tonight.

And... still gathering knock vs water injection percentage data. I don't think doing two back to back runs will be enough considering how much can change. I'll post when I have enough to show a solid trend.

Harlan

iTrader: (3)

And it's a fuel leak at the pump hose... Trying to figure out if it is the hose or the nipple. This could explain a lot. Fix one thing and break another.