Can't configure Mesa 7i92

OK, made a cable and got some results, but three problems:

1. The axes only move one direction. All 3 will move one direction but do nothing for the other direction. Now if the DIR pin wasn't connected properly, it would move left if you pressed the left or right arrow. But it's moving left for the right arrow and no motion at all for the left arrow, so it's not simply a malfunction on the DIR pin. So confused on this point, because the STEP command should pulse for both left and right arrows.

I think the sample hal file uses quadrature instead of step/dir, to change to step/dir change the step_type of each axis to 0

(0 is step/dir output mode read "man hostmot2" in stepgen section for details )

2. The directions are reversed. Can I fix that in software? Resoldering cables would be very problematic at this point. I just need to selectively change DIR to active-low.

Yes, the axis directions can be changed a couple of ways in software but first I would fix the step type and recheck.

3. The gantry needs second stepper- the 4th axis side- to be driven along with the X. And it really needs to be driven slave but with homing alignment like in Mach3. Google is telling me LinuxCNC has historically had problems doing this.

It may be that you need the JAx branch for this. I'm not sure what the current status of slaved axis gantry homing is (do you have 2 X axis home switches?)

What are the ways to reverse stepper direction?

Yes, I have dual X-homing sensors. I'll look up the JAx branch. Google tells me a lot of people have been frustrated because all they can do is hard-link the step generation on both steppers in LinuxCNC and had no way to align the axis, which is very surprising. It's a super-common setup.

There are ( at least ) 2 ways to reverse axis direction depending on whether the machine coordinates are correct
(That is does the DRO read correctly)

If the DRO reads correctly the easiest way is to invert the DIR pin

Inverting the direction pin requires setting the invert_output parameter for the proper GPIO pin
You can determine the specific GPIO pin(s) to invert by looking at the pinout that
mesaflash --device 7i92 --addr 10.10.10.10 --readhmid
returns

for example:

setp hm2_[HOSTMOT2](BOARD).0.gpio.004.invert_output true

will invert a g540 configs stepgen 0 direction pin

If the DRO reads in the wrong direction you change the SCALE parameter in the INI file

I think LinuxCNCs main focus has been milling machines and lathes and its
less used on routers, plasma cutters and the like that are likely to use dual drive gantry configs

You might open a new thread on gantry homing and see how others deal with it

OK, got it moving, and the directions are correct-

Now, though, I have 2 weird problems:

1. The stepper is VERY noisy at lower speeds on the Z axis and X axis (which moves the whole gantry, of which only one side works). Weirdly the Y does NOT make any unusual sound, and it's the SAME physical rack-and-pinion as the X. I checked the .INI file and the info for the X and Y are identical. The Z is a leadscrew, totally different, but makes the same noise as the X. It doesn't matter if it's a manual jog at 60ipm or an MDI command at 60ipm, it moves but makes a LOT of rattle/grinding sound. It goes away with a faster move.

2. The second X-gantry motor isn't just unstepped, it's unpowered. This I can't understand- this a G540, there's no way to disconnect power, it should be locked in place in the absence of a STEP pulse. I know the current drops to 70% of its rating when it's idle for a few seconds, but this is much weaker than that. The unpowered side is actually being dragged by the powered side so there's no current to speak of. I'm baffled, it didn't do this on the Ethernet Smoothstepper, and the motor wiring hasn't been touched.

EDIT: ok, I think the noise is from the G540 itself being poorly tuned- LinuxCNC was jogging at 60ipm, a speed I basically never used in Mach3.

I found something odd, though. I went to change my timing parameters to match the G540 limits- STEP 1000ns high, 2000ns low, DIR 200ns setup/hold

What I got was the axis stopped working entirely when STEP was below about 2300 ns HIGH. Stopped working entirely. Is that just due to a weak output being too slow? It this an open collector output?

Basic calcs are showing me I don't need less than a 28us total STEP period so I guess that's not a thing.

As far as timing goes, in my experience the G540 timing specs are not valid at 3.3V.
I would use 2.5 or even 5 usec for step timing, as there is no disadvantage of longer pulses
until they limit your maximum step rate.

The 7I92 outputs are 3.3V push pull signals with 24 mA of drive and > 50MHZ bandwidth
(they can be programmed for open-drain mode but that is not the default)

For the 4th axis issue, you might try enabling the 4th step generator in the .ini file, the
4th axis G540 pins will be both weakly pulled high if only 3 axis are enabled
(since they will be inputs if not enabled as stepgens) and this might cause strange behavior

Hi all..

I'm currently trying to decide between the 7i92 and the 5i25. I'm tempted by the 7i92 because I could connect to an existing LAN port, but have slight hesitation since support has only just been added to LinuxCNC (AFAIK).

Are there any pros/cons between the two regarding performance and reliability etc?

I've always struggled to find a PC that could provide a low base thread times. Will these Mesa cards allow me to run PC hardware independent of the CPU base thread jitter, whilst maintaining stable stepping?