20:52] <EvdZ> Action68, I measured the step pulse witdh. It is > 5us without an additional delay.
[20:53] <EvdZ> I dont think this is the problem. Sorry for the false lead. The performance of this processor is less then the ones i'am used to.
[20:54] <EvdZ> I have also missing steps.
[20:55] <EvdZ> Action68, Are you using motord from mendel-parts?
[20:58] <03Kliment> EvdZ: The step pulse width is sufficient. That's the main reason to use the arduino pin set functions rather than the registers
04[20:58] <03caru> Kliment, i read... yeah that makes no sense, but if it works, great, will add a #define for that
[20:58] <03Kliment> But Action68 added the delay to the command processing code
[20:58] <03caru> i know
[20:59] <03caru> weird
03[20:59] <03Kliment> caru: No, I want to ask you something else
[20:59] <03caru> tell me
[20:59] <03Kliment> previous_micros_y += interval;
[20:59] <03Kliment> Did you put that in?
[20:59] <03caru> yes
[20:59] <03Kliment> Can you explain it to me?
[20:59] <03caru> that makes sure speed is real speed
15[21:00] * Robsato (~Robsato@wetering.xs4all.nl) Quit (Read error: Connection reset by peer)
15[21:00] * RoTorIT_AFK is now known as RoTorIT
15[21:00] * Arc_Phenom (~Arc@nv-71-50-64-245.dhcp.embarqhsd.net) has joined #reprap
[21:01] <RoTorIT> AHOY! RepRap`ers :-)
[21:01] <03Kliment> So it says "the time this step was takes is INTERVAL microseconds after the previous step
[21:01] <03Kliment> But that's not true, is it?
15[21:01] * prusajr (~prusajr@prusa.cust.centro-net.cz) has joined #reprap
[21:02] <03caru> previous_micros_xxx is set at the beginning of the move to current time
[21:02] <03Kliment> Yes
[21:02] <03caru> so each time we take a step, we just add the ideal interval it should take to it
[21:02] <03Kliment> Can it be that because of this we take steps too quickly?
[21:03] <03Kliment> As in, actual time significantly less than interval?
[21:03] <03caru> no matter if the loop is behing with his homework, it will catch actual speed in the next while loop
[21:03] <03caru> it can be this causes missed steps...
[21:03] <03Kliment> Yes, I think so
[21:03] <03caru> but we must deal with it, if we want real speed
[21:03] <03Kliment> Ideally, if we need to catch up we reduce the interval
[21:04] <03Kliment> Rather than setting it to zero
[21:04] <03Kliment> You understand?
[21:04] <03caru> yeah... very difficult to do while keeping the average real speed though
[21:05] <03Kliment> Why? Just do faster steps to catch up.
[21:05] <03Kliment> And after you catch up do slower steps again
15[21:05] * maxbots (~maxbots@76.121.174.186) has joined #reprap
[21:05] <03Kliment> That's what we do now, but the interval between steps when catching up is zero
[21:06] <03Kliment> You understand?
[21:06] <03caru> sure
[21:06] <03caru> but then there is always the issue that during acceleration and deceleration, the interval always changes
15[21:06] * DrDavidWeb (~Adium@xdsl-78-35-83-17.netcologne.de) Quit (Quit: Leaving.)
[21:06] <03caru> also, missing steps maybe caused by the fact that the interval is not calculated at each step during acc and dec
[21:07] <03Kliment> Or add a delay to the bottom of the while loop.
[21:07] <03caru> because we have while(timediff > interval) instead of if(...)
15[21:07] * jebba901 (~jebba@143.231.249.135) has joined #reprap
[21:07] <03Kliment> The while is run more than once only if we are catching up
[21:08] <03caru> which happens more often than you think
[21:08] <03Kliment> I know how often it happens
[21:08] <03caru> just try to change "while" with "if" and you will see
[21:08] <03caru> ok sorry
[21:08] <03Kliment> Now, at the end of the while loop, we could put a delay
[21:08] <03Kliment> Conditional on timediff>interval
[21:09] <03caru> nice idea!
[21:09] <03Kliment> So if we are in catchup mode, we wait a bit before the loop is repeated
[21:09] <03caru> i like it
15[21:09] * da_poller (~speed@p579958C2.dip.t-dialin.net) Quit (Quit: WeeChat 0.3.4)
[21:09] <03Kliment> Will you do it?
15[21:09] * jebba900 (~jebba@32.144.121.175) Quit (Ping timeout: 258 seconds)
[21:09] <03caru> i will
[21:09] <03caru> not sure i can do it today, but i will
[21:09] <03Kliment> Thank you! Do you think you can add Z interpolation also?
[21:10] <03caru> i'd rather wait i merge ramp acceleration for that