Driver latency?

[chmedly]

I'm curious about issues with latency (delay) between different kinds of drivers used in one machine. For instance, if I use a different brand of driver on the x-axis compared to the y-axis, and there is a difference in timing, cutting moves would not be accurate. Many drivers are, I'm sure, simple enough that there is no "latency" per se. But if there is a dsp chip in the driver then it seems that there will be some kind of latency for any signal being sent through it. Any comments?

Similar Threads:

• Problem- LOL@Linuxcnc 1000000 ms Latency
• need new pc to run emc2 (latency issues)
• EMC2 LATENCY RESULTS
• What to do to get latency figures down?
• Need Help!- Determining Latency...Which #?

[ger21]

Decent drives can handle 200,000 steps per second. I don't think that would be possible if there were any signal delay.

In most cases, you'll at minimum be dealing with thousands of steps/second. If any stepper driver had a noticeable delay, it would be unusable.

[chmedly]

Well, I'm not talking about throughput (or bandwidth or whatever you want to call it). A system can have high throughput but also very high latency. Think about how your machine could be getting 200,000 steps per second but it takes 30 seconds for the first step to make it through the system. Or go to speedtest.net and run a test. The ping time is a measure of latency (it's a measure of round trip time for a packet to the server and back). Then you have the baud rate for download and then upload.

[A_Camera]

I believe latency is in the order of nanoseconds so I hardly think it would affect the end results unless you have an extremely fast machine. Besides, the issue is not latency but the difference in latency between the drivers.

[ger21]

Think about how your machine could be getting 200,000 steps per second but it takes 30 seconds for the first step to make it through the system.

As I said, if there was any noticeable latency, the drives would be unusable.

[chmedly]

As I said, if there was any noticeable latency, the drives would be unusable.

Can you explain why? I'm sorry if this seems obvious to you but I'm not sure I understand your statement.

[A_Camera]

As I said, if there was any noticeable latency, the drives would be unusable.

Actually, no. Speed is not affected by latency. Latency is something else. You could have a driver with huge latency and still be able to use it at 200,000 steps per second. Latency means something else. Of course, you could not mix a driver with huge latency with drivers with very little latency, but as long as all drivers are similar it would not matter.

As an example: Imagine 4 people planning on having a one hour meeting. We could have the meeting starting exactly on time, lasting one hour but we could also have a meeting starting 15 minutes later but still lasting one hour. As long as we know the meeting starts later we can do exactly as much as if we would have no latency (not started later). One hour is still one hour. What we could not have is if each of the four people arrive at different times to attend the meeting...

[ger21]

I was assuming that if there was any latency, it wouldn't be identical among various drives.

But I really think this is a non issue. I've never heard anyone ask this question before.

[A_Camera]

I was assuming that if there was any latency, it wouldn't be identical among various drives.

No, it depends on the parts and the technology used. Different drivers have different latency but I don't think they publish or measure this. I think they all assume that drivers are not mixed, so it is not relevant in real world.

But I really think this is a non issue. I've never heard anyone ask this question before.

I think latency is a non-issue because we are actually dealing with pretty slow events and latency is typically in the nanosecond magnitude probably in all drivers. It would be a problem if it was in the microseconds, or even worse, milliseconds range, in which case it would not be possible to mix drivers.

Perhaps one day, if I get bored or retire from work and look for something to do, I'll measure my drivers to see if they are different in this respect.

[PCW_MESA]

On typical drives there will be random latency in the 10 to 50 microsecond region. It typically cannot be much less than this because the
actual change in stepper drive voltage has to wait for the next (asynchronous) chopper or PWM cycle. (and these typically are in the 50 KHz region or slower).
The slow change in step motor current in response to this voltage change will further slow the actual torque response to the step input.

There may also be significant constant latency on drives that have (enabled) features like step "smoothing"

This makes little practical difference since even a 50 usec timing difference will not make a significant addition to a step motor
driven CNC machines error budget.

[A_Camera]

On typical drives there will be random latency in the 10 to 50 microsecond region. It typically cannot be much less than this because the
actual change in stepper drive voltage has to wait for the next (asynchronous) chopper or PWM cycle. (and these typically are in the 50 KHz region or slower).
The slow change in step motor current in response to this voltage change will further slow the actual torque response to the step input.

There may also be significant constant latency on drives that have (enabled) features like step "smoothing"

This makes little practical difference since even a 50 usec timing difference will not make a significant addition to a step motor
driven CNC machines error budget.

OK, 50 microsecond is quite a lot (definitely much more than I expected) but in any case, as long as all the drivers are in similar magnitude it doesn't matter.

[chmedly]

Thanks everyone for your comments. As Gerry pointed out, no-one seems to have asked this question and since it is possible for it to be an issue, I wanted to find out if it actually was with real systems. To follow up to PCW_MESA (Peter, right?) comments; I did some quick math and 50 usec, even if that was the offset between drives, does seem like it would be at least an order of magnitude inconsequential for a wood router.