yes Stirling has the nub of it - you can use a serial connection if it can send the information fast enough - and in this case we are talking about 16 times or more faster that the parallel port.

Sorry Jim  - but if you read my post again I'm sure you'll see that this was not what I was saying at all - quite the opposite in fact.
I'd also make the point that when it comes to discussing the relative speeds of parallel v serial transmission we need to establish some units - otherwise we're not comparing like with like. In fact if we use the same units for parallel and serial we'll find that - surprise surprise - to get the same throughput, by definition they both have to work at the same speed. (actually this isn't 100% true - one issue is that the serial has to work slightly faster to take into account the time needed by the operation of the shift registers in the external hardware to re-parallelize the data).
Art did NOT choose the parallel port because it was faster than any serial port or any other port for that matter - he chose it because Mach did not then need any external clocking hardware to re-parallelize the data and could therefore be sold as a 100% software motion controller - something others were struggling to do or had already given up on.
Cheers
Ian

Yes - Mach 3 is simple, because it uses the parrallel port. I think that was what attracted me to it.

I have read your post again - and in the last two lines you are saying "The only SPEED issue is that if you use a USB and the associated and required external hardware then the USB must be able to send data fast enough that the SERIAL data can be parallelized in the same time as a parallel port could have sent it."

Perhaps we are coming from different angles - but that is what I was saying.
The PC collects date for the LPT1 port in three addresses. It amalgamates that into 17 lines of data and puts it out on the parrallel port. It is hard to know the internal workings of the CPU, but somewhere between 4 and no more than 8 moves would be required. For serial work, the same data has to be amalgamated, then moved to a buffer for one bit at a time for transmission. The transmitters I am familiar with only go at CPU clock speed. It follows, therefore that in simple terms the CPU will take at least 25 moves more to put out the data.

I must admit - and I know nothing of USB systems - if the CPU can put the parrallel data into a buffer, and then a seperate system moves the data serially at an enhanced rate - the the time difference may be small.

With the dissapearance of LPT1 ports, it would be a great benefit if someone could produce a USB2 to Parrallel output that could guarantee the speed and reliability of the LPT1. Even better if it had reversible pins. The plug and play type of connection could be a great enhancement.

I think you both ARE saying basically the same thing.
And THANKS for putting in such clear and understandable terms.
I learn something here every day. BRAVO
Thanks much,
RC

Might be worth you guys without parallel ports waiting a few more weeks then getting a Smooth Stepper, its USB and does the pulse external to the computer. It will do up to 4MHz per axis and is ultra clean compared to the software/parallel pulse. I have been Beta testing the Smooth Stepper and its working great, should mean you can basically use any computer (laptop etc) as all Mach has to do is the calcs and not worry about the pulses. I have had the Smooth Stepper running from a pico motherboard (total size of pico is 100mm x 70mm) its 1GHz embedded CPU and I have 1GHz memory in it. I will be using this setup in my Beaver mill that I am retrofiiting at the moment, another great thing is I will be able to use the 2000line encoders that my motor/drive combos have without having to resort to the drives electronic gearing

Hood