7i76e and opto ancoder TCST2103

Thanks for tips !!

I think is better for halscope parts and not so bad for speed displayed

Same parameters but return to quadrature and scale 400

Regarding to my old hardwar i see one difference
old 24v kill switch + 24v PWM + Ena + 24V + 5v + Gnd + free wire + encoder in the same shielded harness (isolated by DAC and Opto inside spindle box)
now 5v Kill switch + Analog signal + Ena + Com (+unused 15v) + 5v + Gnd + encoder in the same shielded harness

both have shield to Gnd

It is possible EMI from unisolated Analog ?


For speak about positive things the 7i83 seem to be very good with correct power supply : without any PID the actual speed are really close to the commanded speed and are really helpfull for easy 0v at 0rpm without any offset component ! (including watchdog !!)
The 7i96 are really cool regarding to the input/output i have now only one pull-up resistor for probe !
I do not speak about step generation because they are already know for quality !

That last plot doesn't look like quadrature (no 90 degree between phases)
It really doesn't look so much like EMI but rather bad signals from the encoder

The earlier plot (in count mode I think) might be about as good
as you can get if the signal is as bad as it looks it (its a bit hard to
tell without going slower so the 1 ms quantization is less of an issue)

That is if you have a 20% variation in edge position in the cycle,
at low speeds this can result in a 40% velocity estimate error
(because if one cycle is late by 20% and the next is early by 20%
the time between edges will be off by 40%. This should get better at
higher speeds where you have multiple edges per servo thread so
the velocity estimate has more than 1 edge-edge cycle to time
for the velocity calculation

Well

Ok i will continue to search the correct positioning, but before changing a few hours ago this (original) encoder work fine with older hardwar.

Changing optocoupler can be a good idea ? (if possible to found for better signal)
Changing the whole encoder for some chinese manufactured holow shaft encoder ?

Not sure why the difference,maybe the optical thresholds changed do to the
differing load resistance or similar

if you want to verify the encoder velocity estimation you can
loop a free stepgen back to the encoder (either just A in counter mode
or A,B with encoder and stepgen in quadrature mode)

Oh yes, soon i can rewire my tailstock unused encoder to the mesa encoder and remove the belt

if i understand the goal is turn the stepper motor with fixed velocity and check encoder accuracy

Thanks a lot

Well

It seems to be not really better with nema23 1000ppr encoder , scale set ENCODER_SCALE = 4000, quadrature mode

Wire are short, i have unconected all the input/output and other stepper

attached my modified hal file


velocity set to 0.5 rps :


velocity set to 15 rps :



At the end displayed RPM seem to be in a tiny range but encoder signal is for my eyes not really better

The encoder signals will not look good because they are not representative
of the actual signals (A and B are 15 KHz signals at 15 RPS so you
will just get aliased signals when you sample them at 1 KHz)

Even at 1/2 RPS you get 500 Hz square waves (1 ms high 1 ms low if perfect)
the1 ms sampling of linuxcnc/halscope will cause massive distortion of these waveforms

You would need a real oscilloscope to see the encoder signal at this speed

at 15 RPS you should get a quite accurate RPM reading

With spindle encoder 100ppr

Velocity set to 30 rpm



Velocity set to 450 rpm