RMT - 2 Channels syncron?

[zmaier24]

Hello,

i have a "problem" with RMT. I want to use 2 channels just for sending a sequence. For test reasons the sequence is just 1,0,1,0,....
And the 2 channels should start synchron. I am using Micropyhton 1.18 on a Wemos Mini ESP32.

Now, the problem

1. Both channels are not starting synchron and
2. even the sequence is the same, the output is not the same, it is inverted?

Any ideas why?

Here the code:

from machine import Pin
try:
    r = esp32.RMT(0, pin=Pin(21), clock_div=80)
    r1 = esp32.RMT(1, pin=Pin(22), clock_div=80)
    r.loop(1)
    r1.loop(1)

    r.write_pulses((100, 100), 1)
    r1.write_pulses((100, 100), 1)
except:
    r.deinit()
    r1.deinit() 

[peterhinch]

It would be interesting to try this with single-shot sequences.

[zmaier24]

Unfortunately i killed the device, i have to wait till the new ones are arriving :(

[jonathanhogg]

It's unlikely to be inverted. It's more likely to be ~180° out of phase. A non-symmetrical sequence (e.g., (120, 80)) would confirm this.

I can't think of any way to start two RMT channels in sync. The best you would get is starting them very close to one another, but that would require large changes in the MicroPython RMT implementation: in the current design there is a not-insignificant amount of setup being done when you call .write_pulses() before the hardware is triggered.

[zmaier24]

You are right, it is not inverted, there is just out of phase, since there is a time delay. In my config, the time delay is exactly as big as that the signal looks like it is inverted :)
I played with a time.sleep_us between the two `.write_pulses(). to sync the two streams, but this was also not really working, because the waiting time needed was not constant (After a reboot). Which is strange, i thought the programm is deterministic?
I need between 135 and 145 µs time delay.

r.write_pulses((100,50),1)# 100 = 100µs
time.sleep_us(140)
r1.write_pulses((100,50),1)

Any idea how i could generate a sequence and the exact inverted sequence (this is actually the usecase for me).

[robert-hh]

At the moment, the only reliable method is using an external inverter. If a high impedance/moderate rise time can be accepted, then it could be as simple as a single 2n7000 transistor and a resistor.

[zmaier24]

Thx for the hint, i made a simple Switch with a BC337, it works perfect :)

[mattytrentini]

A Sync Manager allows multiple channels to be started simultaneously, but such configuration isn't yet exposed through MicroPython. See Multiple Channels Simultaneous Transmission.

I think if we put our heads together we can come up with something to provide this functionality!

[russhughes]

That looks interesting, it would require using a newer version of esp-id.