lquiroz

I found a solution to adapt the pulse width to needs of external drivers.
Warning! I have not evaluated yet the impact on the performance (and I am not a programmer).
I modified stepper.c adding a delay as a function

#include <util/delay.h>

void delay(int us) {
while(us–) { // loop adds some overhead
_delay_us(1);
}
}

and inside of each “if” in ISR(TIMER_DDA_ISR_vect) I put a delay of 2 and followed by step bit off:

if ((st_run.mot[MOTOR_4].substep_accumulator += st_run.mot[MOTOR_4].substep_increment) > 0) {
PORT_MOTOR_4_VPORT.OUT |= STEP_BIT_bm;
st_run.mot[MOTOR_4].substep_accumulator -= st_run.dda_ticks_X_substeps;
INCREMENT_ENCODER(MOTOR_4);
delay(2);
PORT_MOTOR_4_VPORT.OUT &= ~STEP_BIT_bm;
}
Without the delay I had a pulse width of 1.4 us . Now pulse width for all motors is 3.4us. I measure this with an oscilloscope.
Now J19 and J20 are working and I know why.
I hope it will be helpful for all people that tried to use external drivers with tinyg.

cmcgrath: Thank you for your answer.
Before to send my post here I read all posts related to external drivers and pulse width.
You are right. My parameters are for a screw drive machine.
I do some test with all $_tr=40, as you suggested, but I have the same results.

Exploring the code I found on stepper.c
// pulse stretching for using external drivers.- turn step bits off
PORT_MOTOR_1_VPORT.OUT &= ~STEP_BIT_bm; // ~ 5 uSec pulse width
PORT_MOTOR_2_VPORT.OUT &= ~STEP_BIT_bm; // ~ 4 uSec
PORT_MOTOR_3_VPORT.OUT &= ~STEP_BIT_bm; // ~ 3 uSec
PORT_MOTOR_4_VPORT.OUT &= ~STEP_BIT_bm; // ~ 2 uSec

between 530 and 534 lines. I am not a programmer but I see some code done to provide longer pulses for external drivers. Comments show me that the pulse width is not the same for all ports. Am I right?
cmcgrath: May be I must start a new thread on pulse width? PW of 4-5us is fine but only if we have the same in all ports.

Thanks you again.

Thanks cmcgrath for your answer.
The setup on both boards is a X,double Y,Z. One board is set with 1.25 mm/rev and the other one with 5 mm/rev. There is some differences with speeds also.
The firmware on both boards is 440.20 version 0.97
I use Coolterm or ChiliPeppr. The behavior is the same (same pulse width)
Here are the links to both boards parameter configuration.
https://www.dropbox.com/s/jfz9g7c8n3fl5jg/TinygV8-1.txt?dl=0
https://www.dropbox.com/s/bhdagnianrsqmk8/TinygV8-2.txt?dl=0
I have connected four stepper motors to one board outputs. Of course the four motors work fine because the on board drivers can handle pulse width of less than 1μs (at my knowledge)
Thanks you for your time.

Thank you for your answer.
Both boards are 131v8h
I tested each external driver (four) connecting each one to only one conector (J18) and every one works fine. No one of the four external drivers works fine connected to J19 or J20 connector. That means to me that the driver is not the problem.
I don’t know very much about the input side of the external drive, except the connection information http://www.omc-stepperonline.com/images/typical-connection-m542.jpg. It seems to have opto-isolation.
I tested again with 3.3V to 5V level shifting for J19 connector. I have the same behavior. The same driver connected to a level shifted J19 doesn’t work but connected to a level shifted J18 works fine.
When I send movements commands without drivers connected the four green leds are lit and then go off, no matter what axis is driven. I have $3pm=2 and $4pm=2.
I was thinking about timing or pulse width. Could be different for J19 or J20 compared to J17 or J18?
The driver specification says: “For reliable response, pulse width should be longer than 1.5μs” and “For reliable motion response, DIR signal should be ahead of PUL signal by 5μs at least” I will try to measure pulse width.
Thanks again for any help.