Remora - Rpi Software Stepping Using External Microcontroller via SPI

Bari wrote:

I'm waiting for hardware to be available. I will try to find out why the step rates are not 2-3X higher than they currently are. The Novosun cards reach 200KHz step rates for each axis with stm32f2's.

Step rate is determined by the base thread frequency, which is now configurable. It really comes down to how long the code spends in the base thread (doing steps) vs being out doing the servo thread an main loop. In the dev branch there is some thread debug code to allow some spare outputs to be used to show the thread activity.

Also, we could double the step rate by completing the step in one thread cycle vs the two that are currently used. I just have not seen the need yet for super high step rates on a real machine... But maybe I should for advertising purposes.

AlessandroEmm wrote:

Apparently MX_master has looked already into these and found them not suitable. ( forum.linuxcnc.org/38-general-linuxcnc-q...-card-with-linux-cnc ) That said I found a thread with/of him on the russian cnc forum, he seems to be onto something new as far as my browser was able to translate. I will look up the link again and post it here.

Yes, that's were I noticed the card. I think his issue was wanting to use hardware step generation. For Remora, no issues with pin assignment with software step generation.

 Also, we could double the step rate by completing the step in one thread cycle vs the two that are currently used.

I thought that was already the case.

I forgot to mention that many servos offer Step Dir inputs now and have torque curves that reach in to the many thousands of RPM vs ~1K RPM for steppers.

In our quest for higher step rates, it's important to realize that most step/dir servos max out at 500 KHz step rate. DMM servos for example.

The JMC integrated servos which are fantastic for hobby or even prosumer machines max out at 250 KHz, which is the same as most external stepper drivers. MHz step rates are really not required for the vast majory of machines. So there is definitely a point of diminishing returns after 200 KHz or so.

I think even a 100Khz would suit 95% of uses cases. I personally needed 45KHz using 4mm pith and 16x micro-stepping to get the speeds I needed. So the LPC17xx SKR 1.4 turbo didn't work out for me it was maxed around 30KHz which is close to the step rate a Pi can do with its own I/O. So I'm watching closely at the ARM 32-bit progress here. I know there are different ways to generate the step pulses but I still struggle with why the same hardware fluctuates so much. the LPC17XX for example in Grbl is 150K X5 axis, in marlin it's 200k, in Mach 3 its 100k x4 axis but in Remora it 30k.. or the stock Rpi I/O on Linux CNC is around 35KHz but on pandapi is making step rates in the 10+ MHz using the rpi4 gpio. The pi does not have a PRU like the beaglebone or orangepi so what voodo is going on?

Yes 100KHz stepping for 3-5 axis would cover the majority of applications.

Pandapi is a new one to me. Looks like it uses a stm32f103
www.pandapi3d.com/forum/pandapi-fdm-3d-p...ns-about-the-project

is there a schematic somewhere? Found it seconds later   gitee.com/markyue/pandapi_wiki/raw/maste...atic_PandaPiV2.8.pdf