Can we simply replace 5v with 3.3v?

[AndreyB]

Would it be possible to completely eliminate 5v in our ECU? Let's reiterate all sensors and how they would be transitioned to 3.3v

Thermistors - 3.3v compatible as is?
TPS - just a potentiometer, 3.3v compatible as is?

MAF, Mine (1996 Ford Aspire) outputs current, so it's just the question of changing the pull-down value to get it into the 3.3v range
MAP. What about MAP? These output 5v, but what is their input - is it 12v or 5v?

Hall sensors - should work with 3.3v as is

Do I miss anything?

[blundar]

Most hall sensors are open collector - they'll work with whatever voltage (within reason) you pull them up to.

There are a lot of engine sensors (MAP most notably) that expect a 5V reference signal. You could use a minimal LDO reg and current limiting resistor setup there. That's probably the hardest thing I can think of.

All the resistive sensors (air temp, coolant temp) could be refactored to 3.3V easily enough.

All 0-5V inputs can be relatively easily dealt with using a voltage divider. You're going to sacrifice some accuracy here due to multiple resistor tolerances being involved, but with 1% and 0.5% resistors it shouldn't be too murderous. I'm attaching a schem snippet I used for this that also has a protection circuit. I will be the first to acknowledge this is far from perfect but it should be enough to prevent fryage from +12v or -12v being connected to an ADC.

The biggest challenge IMHO with a pure 3.3V design is finding FETs / H Bridges that are worth a damn with a 3.3V (or lower?) Vgs drive. FETs traditionally do much better with a higher (close to 12v) drive. One of the biggest advances in past 5 years or so has been more FETs on market that have usable RDSon values at low-logic level drive. I put my favorite one up in another thread.

[kb1gtt]

I agree, that resistor sensors at 3.3 will be fine, or fine enough, however the issue we have is that many sensors look for that 5V reference like hall pickups. I like the idea of a high power 5V with a lower power 3.3V for sensors and MCU operation. The high power 5V can have ripple and such, while the 3.3V would be a good clean signal.

[blundar]

Most Hall sensors have open collector outputs (i.e. they sink current to ground so you can use whatever pull up voltage you like) and they also generally can be powered by 12V. The OC output means the supply voltage isn't connected to the output voltage.

[abecedarian]

Still have issues with things like MPX4250 needing and outputting 5v signals, O2 sensors being similar.

Might be worth noting there are MAP (Honeywell comes to mind) and O2 that output over I2C/IIC and/or SPI.
So even if they are 5v devices, logic level translation could be used to de-couple the device from 3v3.

[kb1gtt]

Also most hall's can't actually pull to 0V. You'll typically end up around .5 to .7V. For 3.3V systems this might be an issue, you really have to verify what the low threshold is for the input pin. I believe that 3.3V MCU's have a lower voltage requirement. So a 5V system might be more likely to register as a 0 instead of a 1, while a 3.3V might register as a 1 instead of a 0. Also don't forget that the bottom transition will be rounded as it discharges some capacitance or inductance. So if your DC pull down is .6V and your MCU threshold is .7V, the rounding of the bottom of the transition will likely be exhibit a fair bit of jitter. This jitter is an elusive issue. Ultimately we need to measure the DC off voltage, and we need to verify it's OK with the MCU minimum to register as a 0 instead of a 1.