Maybe new rusefi target? TMS570

[roccomarco]

May I ask what's the current problem with STM32? I know it is not an automotive grade MCU but I am curious about what are the problem you are experimenting and why you are migrating...

[AndreyB]

May I ask what's the current problem with STM32?

Short answer: there are no major problems with stm32

Longer answer: to some extent, it's a holy way between old-school 5v automotive and our current mp3-player-grade 3v.
A lot of vehicle sensors are 5v - we need a few fast analog inputs, I believe external 5v ADC would be impossible or complicated. At the moment we have op-amp and divider - see http://rusefi.com/forum/viewtopic.php?f=4&t=19

Another 5v use-case would be taking and old OEM ECU, removing old chip and soldering in a tiny daugter board with rusEfi brain. For such a use-case 5v chip would produce a small daugter board.

Aslo safety CPU would help with safety-specific features like watch dogs etc. Also in theory we can use more hardware timers and more input captures - stm32 is kind of limited on isolated hardware timers, the whole "8 timers x 4 channels" is a bit of a mess in my view, I'd rather have 32 really independed input and output timers.

[roccomarco]

I'd rather have 32 really independed input and output timer.

sounds costly... But I understand your point of view... I will take a look to TMS570 in my spare time.

[kb1gtt]

Hello @ and please to hear from you.

Another limitation we are faced with would be more noticeable when we try common rail direct injection. When that happens we need much more precise timing. It is common that an injector will be fired 4 or 5 times during the combustion stroke. In that situation, very small amounts of jitter become a problem. An low latency processor become really helpful in simplifying that kind of setup.

[stefanst]

[...]
A lot of vehicle sensors are 5v - we need a few fast analog inputs, I believe external 5v ADC would be impossible or complicated. [...]

Is there any good reason why we don't use an external SPI ADC like the MCP3008, http://www.microchip.com/wwwproducts/en/en010530 or for more resolution the MCP3208? They're fast, accurate, cheap, 5V tolerant etc....
I'm not aware of any automotive analog signal that requires true real-time sampling. The fastest sampling on standard use cases would be needed for MAP and TPS and there I'd say if we get every ignition event on a 4 cylinder spinning at 6000rpm, we are definitely fast enough. So that would be 200Hz sampling. That'd be approx. 1/1000 the rate the chip is capable of, if I read the datasheet correctly. SPI comm is already implemented, so the effort in programming shouldn't be too intense.

This would open us up to simplified circuitry, avoiding the Op-amps and we'd be more independent in terms of usable processors. All the timing-sensitive stuff is digital as far as I can tell, so here we can still use interrupt-capable pins.

I'm sure I'm forgetting some important facts, but wouldn't mind getting educated on them.

[AndreyB]

So that would be 200Hz sampling.

Right now we are 10Khz since we are averaging MAP within a certain angle window (maybe for no good reason). Also I really do not want to think about SPI and fast SPI.

[kb1gtt]

Fast SPI commonly causes unintentional radiators. We can deal with this by treating the traces like a coax, but it becomes more of a problem. Right now we are reasonably low frequencies, and reasonably low harmonics. Once we increase frequency such that we can pump the data over an external bus, the PCB routing become harder to do properly. So problem 1, is PCB design resources / time. Also keep in mind that we not only increase the frequency, but we also lengthen the antenna.

Second problem is that the ADC has variable impedance. I'm not sure of this chip, but it commonly ranges from 5k to 100k ohms. This chip is one ADC with a MUX, as you change the MUX switching frequency you change the impedance. So if we set it to MUX at say 10kHz, such that your ADC is a fairly constant impedance, your 10hZ signals need allot of signal processing to low pass filter. By using the op-amp, we keep a constant impedance seen by the sensor circuits while allowing for the ADC's to vary impedance. Often OEM's can get around these issues by controlling the MFG process, such that they know what sensors are being connected. However we do not have those quality controls. So second issue is that op-amps offer impedance matching.

Third problem is that if we consider external ADC's we should consider something that simplifies the analog design. The below chip includes low pass analog filtering, as well as digital FIR filtering. Take note there is no MUX.
http://www.quickfiltertech.com/qf4a512.html