Tutorial SPS - Sabertooth Packet Serial PID motor driver sketch w/SoftStart

you have to edit the sketch before uploading

Yes lol but where in the sketch

edit: I found it however there are a lot, do i change them all ?

#ifdef SECOND_SERIAL
mySerial.begin(115200);
#endif

to

#ifdef SECOND_SERIAL
mySerial.begin(38400);
#endif

I just sent you a PM with a new sketch to try with instructions how to change the sabertooths baud.

@BlazinH, I only had a quick look so this isn't a comprehensive response but I notice the packet serial requires about 4bytes minimum per motor. At 115200 baud that's at least 1ms to send all the data to 3 motors without allowing for any processing in between.

I would get rid of the ST[1] commands for motor 3 and just make it a two motor controller. That will give you a bit extra headroom.

It also means that you don't want the PID loop running at its full speed. I note in SilentChill's screenshot Fpid is set to divide by 5 which is good - maybe even go divide 6 or more if I allowed it? If you run faster than that I expect it will be overrunning the tx buffer to the sabretooth.

As for the baud rate to the sabretooth, this also means you want the fastest baud rate possible to get the commands flushed before the next time round the loop.

You may also find that adding #pragma GCC optimize ("-fast") at the top of the code might help speed the code a bit (sometimes it can have negative effects though). Although the main limiting factor is the rate at which commands can be sent to the sabretooth this might at least ensure the rest of the code is as fast as possible.

@RufusDufus I have just tried from 5-10 Fpid and it didnt make any difference still each motor twitching in turn and then goes crazy when I try to move the sliders in simtools

Thanks for your sharing your insight @RufusDufus. I must not be figuring the baud rate correctly; 4bytes x 8 = 32 x 3 motors = 96 x 820 (fpid/5) = 78720?

Also, I just added motor 3 for the hell of it but it has never been tested so removing it is a good idea.

Btw, if the sabertooths tx buffer is overrun it will just ignore additional data. Sabertooths can handle up to 2000 updates per second though.

Probably don't want to use larger than Fpid /6. That will give 410 motor updates per second and I don't think less than that would be good. You could try using 4 too just for kicks.

CORRECTION: I was incorrectly thinking in binary terms. Fpid /10 = 410 updates per second and Fpid /6 = 683.

There are 10bits for a serial byte at the hardware level (includes the start and stop bit)
so that is 1/115200 * 10 * 4 * 3 for three motors with 4byte packets = 1.04ms assuming the bytes are sent without any interbyte delay which might be optimistic for softwareserial?

The PID loop is about 4000Hz at full speed from memory so Fpid/5 = 1.25ms round loop assuming nothing else slows it down.

@RufusDufus today I tested again the UNO with the latest sketch from @BlazinH and I can confirm that 1ms packet rate works flawlessly on the UNO. Maybe there was a bug earlier when testing different betas so at that time I came to this conclusion.

Updated sketch posted.

Quick update, I have been using the latest sketch now for a few weeks and it is working had to set Fpid 4 to get rid of an annoying twitching that was going on but other than that it works great.

Big thanks to @BlazinH for all your hard work making this and helping me set everything up and @RufusDufus for his amazing SMC3

Think this should possibly get a sticky @Historiker @value1 ??

Agreed and done.

UPDATE: This sketch is further optimized for speed and can now use an Fpid setting of 1. If you experience motor twitching or issues at Fpid = 1 then try increasing it.

Btw, added a SOFT START motor routine too so I need some testers to let me know if it needs any tweaking. At the top of the sketch look for
//****************************************************************************************************************
//SET SOFTSTART SPEED
const int SSloop = 8;
//****************************************************************************************************************
I estimate SSloop = 8 to be around 2 seconds to achieve full motor power. Increasing SSloop will slow it down more and decreasing it will speed it up.

Is this so that the when it starts up it moves slowly into position ?

Yes, on initial start.

Set this up on my test bench with a new Arduino and 2x32 today. Everything went smoothly. Looking forward to trying it with some bigger DC motors on the weekend. Great work guys.

Great work !, working great with my 2x32

Just ordered the 2x60 Sabertooth, so this is really helpful! Thanks

FYI - if using the new Simtools 2.0, your interface output should be: [A<Axis1a>][B<Axis2a>][C<Axis3a>]

Note the Axis1a not just Axis1

Might be worth updating the original post.