3 phase alternator output load power sensing

[beduino]

Hello,
Another challenge is to sense 3 phase alternator load power output with desent accuracy - it could be nice to have +/- 1 W accuracy at power ranging between 0W - 500 W, but +/-10W will be not o bad too.
Its original internal voltage regulator was bypassed and coil is powered at around 2A 6VDC while resistance of this coil is ~ 3 Ohms.
Alternator spins and a few car H4 light bulbs are connected directly to 3 phase output bypassing DC rectifier to have electromagnetic resistance from lowest (0 RPM) to higher alternator RPM.
Current can be sensed by MPU using hall effect sensors, but unsure howto deal with voltage measurement while load is in triangle configuration.



First thought is is add additional resistors to make star configuration and connect MPU ground to the center and try to measure voltages in each star ends using opamp biased somehow?
Output voltage range will be something below 16V at higher alternator RPM while at lower speeds as low as 2.5V, so that is why DC rectifier is not used to avoid its diodes Vf influence.

AC voltage measuremnt is always tricky especially in 3 phase :/

[beduino]

Sorry, I didn't notice this link in search results earlier - probably I've found solution here for AC voltage measurement on MPU pins by usinng differential amplifier like LM258:
AC Voltage measurement using PIC16F877A microcontroller: http://microcontrollerslab.com/ac-voltage-measurement-using-microcontroller/



It is time design PCB for this - probably sensing each phase by own MPU (the same 3 PCBs) will be more accurate since each MPU will sense own phase in parallel than 0.1s - 1s average will be send by 1 wire or I2C to master device to calculate power at giver RPM speed since custom Hall effect sensor is inserted inside alternator with glued small neodynium magnet to alternator rotor 

What do you think about idea of using 3 MPUs instead of one with higher number of ADC I/O pins? We need to sense voltage and current (voltage from Hall effect sensor) 2 ADCs x 3 phases = 6 ADC readings at changing AC voltag efrequencu from 0 Hz to not sure what AC voltage frequency can be at lets say 1000 RPM but by running MPUs in parallel should help sense voltages at similar times at each of 3 phases, I hope , as well as measure AC voltage frequency itself.

[jbb]

You might only need to measure 2 phases (2x V, 2x I), depending on your setup.  Can you sketch the circuit diagram of the alternator and load?

On sampling: it's best to use simultaneous sampling where possible.  This means that all Vs and Is are sampled at the same time, and avoids phase shifts due to delays.  You can get external ADCs which do this.

I would also recommend sampling at a moderately high frequency - at least 10x the highest AC frequency.  Then doing the final filtering in the digital domain where it's easier to get identical filter response across channels.

[tpowell1830]

I would be very interested in seeing the wave forms directly coming from alternator (the AC part), to see how clean it is and phasing. Do you have a 3 channel scope so that you could possibly take a screenshot? That would be cool.

[beduino]

You might only need to measure 2 phases (2x V, 2x I), depending on your setup.  Can you sketch the circuit diagram of the alternator and load?

Nothing fancy but car H4 light bulbs are used where usually they fail this way that daily light is broken but long distance lights are still fine, so by connecting this way in parallel we have 3x50W~150W per each of 3 alternator phases at 12V in triangle configuration, since at lower alternator RPM speeds less AC voltage is available, while alternator is powered by.... human during indoor cycling sessions by using 80T chainring and 13T gears custom made from stainless steel by laser cut
So, we have below 1000 RPM speeds, however efort to obtain lets say 90 RPM chainring speed is noticeable and similar to road bike cycling conditions at around 30 KPH - 35 KPH so about 200W-250W and it is interesting to know how much power is generated using alternator.
Losses in drive train can be about 10% so we can assume 90% efficiency of this 1x1 chain drivetrain to estimate human effor on chainring, i believe

I could draw some circuit but it is as simpke as it is since alternator 3 phase output is connected directly to those 3 terminals od light bulb load ~150W per phase in delta configuration if we were able output 12V, but desent power is also needed at lower voltages, since cold light bulb has lower resistance, so it is not linear relationship between power and voltage but it is summy load only - I'm not interested in good light from those light bulbs at its nominal voltages, but to dissipate output power as heat and some light

[beduino]

Do you have a 3 channel scope so that you could possibly take a screenshot?

I do not expect nothing special in output currents since light bulb resistive load is in delta configuration
Anyway, I should be able sense 3 phase currents at the same time, since 3 MPUs will be used in parallel and AC waveforms frequencies estimated at real time in MPUs itself.

1k resistors probably not needed when 3 differential amplifiers were connected on 3 separate PCBs, but with common GND and Vcc with MPUs for sensing AC voltages between VAC12, VAC23, VAC31 and Hall effect current sensors (galvanic insulation from MPUs) to sense currents I12, I23, I31 - than P12=VAC12*I12, P23=VAC23*I23, P31=VAC31*I31, so this what we are looking for total power: P= P12+P23+P31, if I didn't missed something