1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 <2017> 2018 2019 2020 2021 2022 2023 2024 2025 | Index | 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 <2017> 2018 2019 2020 2021 2022 2023 2024 2025 |
<== Date ==> | <== Thread ==> |
---|
Subject: | Re: Recommendations for EPICS Based Data Acquisition System |
From: | Matt Rippa <[email protected]> |
To: | Mark Rivers <[email protected]> |
Cc: | Rolando Rogers <[email protected]>, Talk EPICS Tech <[email protected]> |
Date: | Fri, 8 Dec 2017 09:06:32 -1000 |
Hi Matt,
It's possible that there are low-cost USB solutions that would meet the specs.
Analog input
https://www.mccdaq.com/usb-data-acquisition/USB-1608G- Series.aspx
Analog output
https://www.mccdaq.com/usb-data-acquisition/USB-3100- Series.aspx
Digital I/O
https://www.mccdaq.com/usb-data-acquisition/USB-DIO32HS- Series
I have good experience with products from this company. I have an EPICS drivers for the analog input models above.
<http://cars9.uchicago.edu/software/epics/ >measCompMultiFunctionDoc.html
http://cars.uchicago.edu/software/epics/ measCompMultiFunctionDoc.html
All of these models also have digital I/O and some have a few analog outputs as well.
I seems like the processing required might be done in a multi-core modern CPU?
Much cheaper than VME or the development costs of FPGA.
Mark
________________________________
From: Matt Rippa <[email protected]>
Sent: Thursday, December 7, 2017 4:29 PM
To: Mark Rivers
Cc: Talk EPICS Tech; Rolando Rogers
Subject: Re: Recommendations for EPICS Based Data Acquisition System
Hi all,
Thanks for the replies.
Here's more info:
Can you describe a bit more the specifications and how it operates?
- How many inputs at 5 kHz?
Attached is a cut sheet. There's 38 analog and 14 TTL inputs.
- How much calculation to determine the outputs?
We have two core processing tasks for mirror position and vibration control.
This results in a total of 5, 3x3 matrix vector multiplies per cycle. This is to
transform for linear sensor space to AZ/EL tip tilt coordinates.
The legacy DSP's are TI-TMS320.
- How many outputs to be written, and at what rate?
All the stepper motor control is implemented with TTL lines. So a 20 Hz stepper clock
is realized by skipping sufficient 200 us cycles to generate the clock.
Again, the IO cut sheet shows this as 35 outputs. But only the motors are using the sample rate.
- Is this just for feedback control, or do you need to log the inputs and if so at what rate?
I'm keen on logging inputs up to the sample rate in an engineering mode if it's easy to do.
This would be very useful for tuning the system. At a minimum I'd want to log the inputs at
1 kHz. The closed loop bandwidth of the system is about 100 Hz.
I have a wealth of information if you're interested to understand this better.
-Matt
Mark
________________________________ From: [email protected]<
mailto:tech-talk-bounces@aps. anl.gov > <[email protected]<mailto:tech-talk-bounces@aps. anl.gov >> on behalf of Matt Rippa <[email protected]<mailto:mrip[email protected] >>
Sent: Wednesday, December 6, 2017 2:54 PM
To: Talk EPICS Tech
Cc: Rolando Rogers
Subject: Recommendations for EPICS Based Data Acquisition System
Aloha,
I'm looking for your recommendations on the latest data acquisition
hardware, compatible with EPICS.
Our Secondary Mirror system at Gemini has a 1 meter diameter,
55 kg mass. There's 3 degrees of freedom and a 5 kHz control loop.
The system was designed in the 90's by Lockheed Martin, based on
a DOS PC with an ISA backplane (PICMG 1.0) and a DSP data
acquisition board with 16 bit ADC's and DAC's. The DOS PC interfaces
with an EPICS IOC via Reflected Memory.
In 2010 I did a conceptual redesign based on a Hytec Blade 9010 using
IP cards and an Intel CPU running RTEMS. Today it seems FPGA's are
a popular choice for data acquisition. Perhaps a Xylinx zynq solution?
Would anyone still consider VME?
I have no requirement for 5 kHz EPICS control. Although that's intriguing.
Thanks for your time.
-Matt