Can't capture all rising edges using timer CH3 while using CH1/CH2 as encoder counter. Chip bug?

[MSipo]

I'm using an STM32F429ZI (NUCLEO-F429ZI)

I have setup Timer 1 CH1/CH2 in encoder mode TI1 and TI2.

I have also setup CH3 as input capture a rising edge. I want to capture the encoder counter every time I have a rising edge on CH3. But I have trouble catching all the rising edges. I can see them on my logic analyzer but often the STM32 misses to capture all of them.

I then tried to configure a GPIO pin to capture the rising edge using an external interrupt. And then manually capture the encoder counter inside the interrupt. Looks like this works.

But why can't I do this using CH3 on the timer? Am I missing something, chip bug?

I haven't measured the rising edge signal yet, so I don't know if it is a problem there.

[Tesla DeLorean]

A signal on CH3 should allow you to latch CNT into CCR3. The count's going to depend on the quadrature input. Signal placement, might want to use a 32-bit free-running counter.

Solid grounds, inputs pulled-up? Signals CMOS, or Open Collector?

What do the signals look like on a scope? Clear edges, no bounce?

What kind of frequency?

[MSipo]

I'm going to measure the signal with an oscilloscope tomorrow. The signal gives a rising edge for every 4096 count on the encoder.

[Tesla DeLorean]

So an Index / Top-Dead-Center type third signal?

Are you using the full 16-bit range of the timer?

Any setting of ClockDiv or Input Prescalers?

[waclawek.jan]

What's the system/timer frequency and what is the frequency/period of input signal on CH3?

> STM32 misses to capture all of them.

How do you observe that, exactly?

Read out and check/post TIM registers content.

JW

[MSipo]

I have a timer capture callback that reads the captured values and places them into an array. I can for example see more than one pulse on my logic analyzer but software has only captured one value to the array. And if I examine the TIM1 registers I can see that the latest captured value in my array is the same as in the TIM1 capture register. I will investigate further tomorrow.

[MSipo]

The encoder is connected to a motor that software can control. Yes, it's an index puls (8 per turn). And I also have a center signal (from an optical sensor) that I use for initial centering (but it isn't as precise as the index pulses). At optical center I reset the timer counter to 0. From there I can count right or left (32767 to -32768). Then I search for the index pulses to again calibrate the position more accurately.