Max Number of Grinch boards

[RonnieK]

Is there a max number of Girnch boards that can be used on one computer? I would like to use four, should this be a problem?

[Wayne J]

4 boards is not an issue for ON/OFF control on one parallel port. I don't remember the exact number of max channels, but I think it was something like 1,024

[RonnieK]

Thanks Wayne, looks like I need to get to work on another one. I used two for July 4th and have a third built and tested but I was not sure if a fourth would work.

Ronnie

[kostyun]

You should able to run four without a problem. Two years ago I ran 256 LPT channels on a 595 (same principal as the Grinch)

[nathank]

I think those boards are "infinitely scalable". The 1024 channels was used in a sample calculation of processing speed if I remember rightly?

[RavingLunatic]

I think those boards are "infinitely scalable". The 1024 channels was used in a sample calculation of processing speed if I remember rightly?

The original number of 1024 channels comes from the original info posted by Peter Olsen over at ComputerChristmas.com for his 595 based design. That might have been his limit with his original software he was running. He also used the term "infinitely scalable" which isn't accurate.


There is a definite limit on the number of channels possible which is driven by the speed of the clock signal.


If I remember correctly from discussions held in the past:

Vixen doesn't have access to a high fidelity timing capability, so it has to rely on a certain number of instruction cycles executed to create the pulses on the parallel port. So in theory a faster CPU should/will create a faster signal.

However, since you are relying on the instruction cycle of the CPU, anytime that the CPU gets tasked with other Windows taskings it affects the timing of the pulses. Which you can see on an oscilloscope quite clearly.



Real world testing today:

With a 1.2GHz CPU, the clock signal is running at about 100Khz (average) which would give you a theoretical 2500 channels at a 25msec interval (if the clock was stable/reliable). However, while watching it on the oscilloscope there are huge variations in the output from the parallel port, so the real number capable on a reliable/repeatable basis is a good bit less.


So what does this rambling mean?

I believe that we should continue to use the 1024 number (at 25msec interval) just as an easy to remember number even though it is probably possible to run much more than that.

For someone to actually try to push that limit they would need 16 Grinches/Coop Olsen 595s and they will most probably see problems with the signal quality going to the boards before they exceeded Vixen's ability to send out the data signals.


So RonnieK you are more than safe to use a 4th Grinch.

[RonnieK]

Thanks RavingLunatic, I was kinda thinking like that but I just wanted to make sure. It's good to be associated with others willing to help. Thanks again.

RonnieK

[d2frette]

Just for the record, he said:

"This How To describes how to build an infinitely scalable controller.
...
A 486 PC driving a 1024 port system could update all outputs 88 times per second, so there are no practical speed limitations."

[darkforcesjedi]

The original number of 1024 channels comes from the original info posted by Peter Olsen over at ComputerChristmas.com for his 595 based design. That might have been his limit with his original software he was running. He also used the term "infinitely scalable" which isn't accurate.


There is a definite limit on the number of channels possible which is driven by the speed of the clock signal.


If I remember correctly from discussions held in the past:

Vixen doesn't have access to a high fidelity timing capability, so it has to rely on a certain number of instruction cycles executed to create the pulses on the parallel port. So in theory a faster CPU should/will create a faster signal.

However, since you are relying on the instruction cycle of the CPU, anytime that the CPU gets tasked with other Windows taskings it affects the timing of the pulses. Which you can see on an oscilloscope quite clearly.



Real world testing today:

With a 1.2GHz CPU, the clock signal is running at about 100Khz (average) which would give you a theoretical 2500 channels at a 25msec interval (if the clock was stable/reliable). However, while watching it on the oscilloscope there are huge variations in the output from the parallel port, so the real number capable on a reliable/repeatable basis is a good bit less.


So what does this rambling mean?

I believe that we should continue to use the 1024 number (at 25msec interval) just as an easy to remember number even though it is probably possible to run much more than that.

For someone to actually try to push that limit they would need 16 Grinches/Coop Olsen 595s and they will most probably see problems with the signal quality going to the boards before they exceeded Vixen's ability to send out the data signals.


So RonnieK you are more than safe to use a 4th Grinch.

It looks to me like it'd be possible to connect several in parallel (different parallel port DATA pin to each chain of Grinches). So shouldn't you be able to program 8 in parallel at the same frequency for a grand total of 8192 channels (provided the computer can switch all the data pins in under 1 clock cycle) if the plugin were updated to allow it?

[nathank]

It looks to me like it'd be possible to connect several in parallel (different parallel port DATA pin to each chain of Grinches). So shouldn't you be able to program 8 in parallel at the same frequency for a grand total of 8192 channels (provided the computer can switch all the data pins in under 1 clock cycle) if the plugin were updated to allow it?

I believe you'll find there is already a plug-in to achieve that, the '8 line 595'.

[darkforcesjedi]

Oops... my bad. So given that, there really is no practical limitation on how many channels you can drive, no? 3 parallel ports * 8 595 chains per parallel port * 1024 channels per chain = 24,576 channels. The setup would cost as much as a house to assemble, consume more power than a residential distribution line can output, and take around 6 years at 8 hours a day to sequence a single song assuming 10 seconds/channel/second of music.

[nathank]

Haha, so, in short - "infinitely scaleable" is accurate enough :)