shortest return delay time is unexpectedly long
shortest return delay time is unexpectedly long
by Zoid
Mon Sep 13, 2010 4:19 am
Using an oscilloscope I've been clocking the actual return delay time from an AX-12+, and it doesn't seem to match the documentation well at all. First of all, mine appears to be working at about 4uS per value (approaching a full 1mS at a setting of 250), and there seems to be about a 21uS minimum (overhead) even at a setting of 0.
Is this normal?
Is this normal?
by i-Bot
Mon Sep 13, 2010 10:50 am
Yes, this is normal behaviour for an AX12. It may vary for other Dynamixel devices.
The return delay value is a minimum value and not the actual value. It is used to allow the bus master time to change from transmit to receive.
The normal minimum time for the AX12 is about 22us as you found at setting 0. However there is also an interrupt process which can kick in and add another 40 or 50 us to that randomly.
I did not check the accuracy of longer settings, because I only use setting 0.
The return delay value is a minimum value and not the actual value. It is used to allow the bus master time to change from transmit to receive.
The normal minimum time for the AX12 is about 22us as you found at setting 0. However there is also an interrupt process which can kick in and add another 40 or 50 us to that randomly.
I did not check the accuracy of longer settings, because I only use setting 0.
by Fritzoid
Mon Sep 13, 2010 11:25 am
That agrees with my experience. The best response time that you can hope for is between 200 and 300 us. per dynamixel command.
Note, at startup time, the controller sets the delay time to zero on all recognized servo devices. So the firmware is currently running with the minimum turnaround time on the dynamixel bus. All you can do is make it slower.
Note, at startup time, the controller sets the delay time to zero on all recognized servo devices. So the firmware is currently running with the minimum turnaround time on the dynamixel bus. All you can do is make it slower.
by Zoid
Mon Sep 13, 2010 5:04 pm
i-Bot wrote:...The normal minimum time for the AX12 is about 22us as you found at setting 0. However there is also an interrupt process which can kick in and add another 40 or 50 us to that randomly.
Does anyone know if Matt's M1 firmware is any faster in that regard?
by Fritzoid
Mon Sep 13, 2010 8:24 pm
Once a packet is received, the servo must do a checksum calculation then decode the command and create the response packet. Even in the best of circumstances this takes a couple hundred clock cycles. So, I doubt that it can be done at all in less than about 10 us.
The average read command is an 8 byte packet. That takes 80 us. to transmit. The average read response is another 8 bytes and takes another 80 us. to receive. That's 160 us. already. A processing overhead of 22 us. is pretty small in comparison.
The average read command is an 8 byte packet. That takes 80 us. to transmit. The average read response is another 8 bytes and takes another 80 us. to receive. That's 160 us. already. A processing overhead of 22 us. is pretty small in comparison.