Latency test questions

Toward the end of the latency test results, there are a couple of lines that read like this:

XXXX secs servo min:-XX.XXX uS max: XX.XX uS sdev: X.XXX uS

I understand the first XXXX secs as being where the reading is taken from, 3600 is one hour in, etc

What do the rest mean?
servo min
uS max
uS sdev
uS

To me it seems that the lower the value on those 4 parameters, the better... but that is all I have derived. Thanks.

txtrone wrote:

Toward the end of the latency test results, there are a couple of lines that read like this:

XXXX secs servo min:-XX.XXX uS max: XX.XX uS sdev: X.XXX uS

I understand the first XXXX secs as being where the reading is taken from, 3600 is one hour in, etc

What do the rest mean?
servo min
uS max
uS sdev
uS

To me it seems that the lower the value on those 4 parameters, the better... but that is all I have derived. Thanks.

If you mean lcnc-hw, then those two lines at the end are just exactly the two lines of the xyz.lat file the script generates. I am currently struggling to ad a graphical output.

AFAIK for servo based systems (ie.: mesa ethernet) tommylight used to call anything below 100 uS as good. From my own experience, with a rather slow (1800mm/min) bridge mill on parallelport, I had no issues with 50 uS.

These figures are just orientation values. As others have pointed out:
a) your system may behave somewhat different when making chips
b) looking at the myriad of variations (CPU, GPU, Disc, Mobo, Bios version, OS flavour and version and update state plus bugs, interfaces, electronic components, environment...) there cannot be a single good value. But I do agree, the current span is too wide.

As for the individual labels:
"XXXX secs" is the runtime of the latency histogram run. I set the default run time to 7200 secs (2h) because my small bridge mill rarely has jobs that last longer and more important, I did not see one case where big changes occurred after 1 hour. So to play it safe: make it two hours

"servo" means the line addresses data for the serve thread. You could trigger also a base thread. That is only suitable if you plan to use a parallel port based set up

"min" is the minmal time say "shift" away from expected response and
"max" the positive

"stdev" is the standard deviation of the data, I am not sure on how many samples this is based

hth

chris

PS: I'll put a link to this in the thread for lcnc-hw , where I feel your question would also have been nicely at home

seuchato wrote:

there cannot be a single good value. But I do agree, the current span is too wide.

By that do you mean the current span of 50uS - 100uS ?

Thanks for the clarifications and your hard work on this for the community!

Yes and no. Yes because I figure tommylight has much more knowledge than mysel. No because I saw figures accepted as good in other threads that are well over the said 50 for base and 100 for servo.

... and thanks for the flowers.

Tommylight: Thanks for the clarification. Is it OK, if I use our above post as a hint to reference indication from my script?

For parallel based systems, there is the Step Timing Calculator and valuable reading here . I recommend reading the whole chapter .

I will add this as reference hint, if the script is run with a base thread.

greez

chris