Guest
Show Posts
1
Mach4 General Discussion / Re: Short output pulse from Mach4 + ESS
« on: August 15, 2016, 02:01:33 AM »
Thanks for the replies - I think a stand alone pulse generator communicating through modbus or some such would be nicest (and most reliable) way to go, however I didn't have the time to build up the hardware and interface it, so I ended up going quick and dirty...
Here is what I tried*:
- Just using a basic IO enable/disable with a G4 pause to set the pulse length. This worked OK for longer pulses but broke for anything below around 25ms. Of the around 100 pulses I tested I saw a variation of generally less than +/- 2ms. Running the PC (i5) under full load (cinebench CPU bench mark running) did not affect the pulse lengths. I changed my exposure setup so the shortest pulses needed are now 400ms so I think this method will be fine, subject to some more testing of course. Very simple and no extra hardware.
- Using a "motor" set up with 100000 counts per unit, max acceleration allowed, and commanding a move of for example 0.001 units at 6 units/min feed rate (10x pulses at 1ms intervals). This method requires some basic external hardware to shape the pulse train, but on the plus side it appears to be accurate to <1ms (seems to be a very small error on the last pulse). Again putting a high CPU load doesn't seem to change the timings.
* My tests weren't very scientific - definitely wouldn't use on anything timing critical!!
Here is what I tried*:
- Just using a basic IO enable/disable with a G4 pause to set the pulse length. This worked OK for longer pulses but broke for anything below around 25ms. Of the around 100 pulses I tested I saw a variation of generally less than +/- 2ms. Running the PC (i5) under full load (cinebench CPU bench mark running) did not affect the pulse lengths. I changed my exposure setup so the shortest pulses needed are now 400ms so I think this method will be fine, subject to some more testing of course. Very simple and no extra hardware.
- Using a "motor" set up with 100000 counts per unit, max acceleration allowed, and commanding a move of for example 0.001 units at 6 units/min feed rate (10x pulses at 1ms intervals). This method requires some basic external hardware to shape the pulse train, but on the plus side it appears to be accurate to <1ms (seems to be a very small error on the last pulse). Again putting a high CPU load doesn't seem to change the timings.
* My tests weren't very scientific - definitely wouldn't use on anything timing critical!!