Open Source "Lab instrument system" Idea

[mrflibble]

*grin* I was almost expecting this kind of reply. I can see the future, because I keep making the same mistakes. And sometimes even other people repeat my mistakes.

Transport is trivial. It's the visualization & ease of applying math & updating that is the thing IMO.

Sooo, do an fft in an easy manner with rrdtool? Yes yes, I know it can be done, been there done that (well, not fft, but the not-so-trivial math). Averaging + error bars? Histograms with binning that doesn't suck?

The thing is, (I think) I know what you roughly can do with the snmp + rrdtool and I see zero benefit over boring old industry standard methods. And there are drawbacks, such as not being able to integrate with tools that actually deal with the problem domain. On the upside you can integrate with all manner of infra-management tools. Not that this will get you anything tangible in a good time spent vs features obtained way, but what the hell!

That said, if all you want is read out a voltage every minute, store it, and do a simple plot then sure it'll work.

Anyways, if you see a benefit or just think it's fun, go for it. 

[linux-works]

I'm not looking to do anything complicated other than setup gear and data log.  I think that will do most of what I would want at the capture phase.

math - that would all happen post-capture, no?  analysis and whatever - I see that as a post-process.  capture is just data logging of raw values and time.  no?

I realize this won't directly plugin to the existing tools in this domain.  that's not my goal.  I'm avoiding some of the existing scpi management tools since they are non-free and I want to give a fully free way to talk to test gear in a simple and lightweight way.

[Bored@Work]

I want to give a fully free way to talk to test gear in a simple and lightweight way.

But SNMP is not simple. Yes, I know what the S in SNMP stands for. But it it is really only in relation to other network management protocols.

[mrflibble]

math - that would all happen post-capture, no?  analysis and whatever - I see that as a post-process.  capture is just data logging of raw values and time.  no?

I realize this won't directly plugin to the existing tools in this domain.  that's not my goal.  I'm avoiding some of the existing scpi management tools since they are non-free and I want to give a fully free way to talk to test gear in a simple and lightweight way.

Well, there are more mature free infra-management tools than there are free mature T&M tools, that's for sure. 

And yeah, math would be post-capture. And inter-capture. And pre-capture. As in control your instruments based on real time measurements and such. Rrdtool and friends are not really built for that. So open loop data logging, sure. Anything even remotely closed loop, not so much.

[linux-works]

I want to give a fully free way to talk to test gear in a simple and lightweight way.

But SNMP is not simple. Yes, I know what the S in SNMP stands for. But it it is really only in relation to other network management protocols.

hey, its a new idea and I'm going to give it a try.  I'm not saying that its going to be a new fad or anything but snmp is a perfectly fine remote control protocol; and to me, scpi is still just another tree-based command language and so mapping scpi to snmp won't be that hard and I don't think it will be that hard to use, either.  snmp is easily scriptable and as long as you have ip connectivity, snmpwalk and snmpset should work just fine.

the other comment about not being closed-loop; I'm not sure how much that will matter to me.  if I happen upon that situation, I'll see how best to deal with it, but I'm more thinking about doing a batch setup of gear, starting a 'run', collecting data and then closing down the run.  for that, I'm pretty sure this paradigm will work out well enough.

the net-snmp cli tools would be a first way to control things.  something like:

% snmpset -v2c -c private 192.168.1.1 KEITHLEY2500::trigger.count.1 = 1
% snmpset -v2c -c private 192.168.1.1 KEITHLEY2500::sample.count.1 = 5
% snmpset -v2c -c private 192.168.1.1 KEITHLEY2500::genCmd.1 = "init; *opc?"
% data = $(snmpget -v2c -c public 192.168.1.1 KEITHLEY2500::response.1)
% echo $data

etc.

in this case, I'm talking to instance 1 of the keithley.  I'd first do a walk of the equipment 'slot table' to find out what gear I have and what their instance id's are.  here, I found it was '1' (I only have one of these boxes on the network).  then I use the instance ID to point to which box the cmds are directed at.  do set's to change things and get's to fetch things.

if I want to get fancy, I could even have an snmp trap come out when an experiment is done so I don't have to poll for a status variable.

just some initial thoughts.  I know its quite a different take on remote control, but I'm going to work thru it and see if its something that can actually be useful or if its just a tech demo.

[Bored@Work]

Good luck in writing your own MIB in pseudo-ASN.1. Good luck in implementing the basic SNMP protocol, several of the standard MIBs and then your own MIB.

[mrflibble]

the net-snmp cli tools would be a first way to control things.  something like:

% snmpset -v2c -c private 192.168.1.1 KEITHLEY2500::trigger.count.1 = 1
% snmpset -v2c -c private 192.168.1.1 KEITHLEY2500::sample.count.1 = 5
% snmpset -v2c -c private 192.168.1.1 KEITHLEY2500::genCmd.1 = "init; *opc?"
% data = $(snmpget -v2c -c public 192.168.1.1 KEITHLEY2500::response.1)
% echo $data

etc.

For your chosen SNMP path that's as good a way to start as any. If you find your shell scripts get a bit convoluted after a while you could move to your script language of choice. I know the Net::SNMP perl module is quite usable (apt get install libnet-snmp-perl). For python you could do apt get install python-pysnmp4-apps and check that one out. Quick check shows it to be a viable candidate but haven't used it myself.

[linux-works]

Good luck in writing your own MIB in pseudo-ASN.1. Good luck in implementing the basic SNMP protocol, several of the standard MIBs and then your own MIB.

not sure if you are being sarcastic or not.

fwiw, I have over 25 years of snmp developer experience, so I do think I can pull this off...

I don't have to implement the snmp protocol.  the net-snmp stack works just fine and you can turn a mib file into c code without too much trouble.  once I have a basic engine implemented, the rest is just 'software wiring' to connect leaf nodes to callbacks.

the standard mibs are not useful here; I don't care (in this context) about ip counters or ethernet counters, etc.  the system group, the snmp group (what a master agent always contains) and then vendor mib plugins for each kind of equipment that I have.

I should probably not take up any more space in THIS thread.  once I get a basic POC, I'll post the code and let people play with it, if they so choose.  please wait until you see this before throwing more stones, ok?

[linux-works]

the net-snmp cli tools would be a first way to control things.  something like:

% snmpset -v2c -c private 192.168.1.1 KEITHLEY2500::trigger.count.1 = 1
% snmpset -v2c -c private 192.168.1.1 KEITHLEY2500::sample.count.1 = 5
% snmpset -v2c -c private 192.168.1.1 KEITHLEY2500::genCmd.1 = "init; *opc?"
% data = $(snmpget -v2c -c public 192.168.1.1 KEITHLEY2500::response.1)
% echo $data

etc.

For your chosen SNMP path that's as good a way to start as any. If you find your shell scripts get a bit convoluted after a while you could move to your script language of choice. I know the Net::SNMP perl module is quite usable (apt get install libnet-snmp-perl). For python you could do apt get install python-pysnmp4-apps and check that one out. Quick check shows it to be a viable candidate but haven't used it myself.

how you call snmp is not as important to me.  sure, there are perl binding, python bindings, c bindings.  this is all manager stuff and its not my main interest.

what I care about is more on the agent side.  the server side, if you will.  that will be written in c since net-snmp is a C implementation and C is my best/preferred language.  the platform will be anything linux: from raspi to beaglebone to itx box.  it should not matter; the code will run on any linux system.

[mrflibble]

how you call snmp is not as important to me.  sure, there are perl binding, python bindings, c bindings.  this is all manager stuff and its not my main interest.

Hence the "If you find your shell scripts get a bit convoluted after a while you could move to your script language of choice". Anyways, have fun.

[jwm]

As for a case, how about one of these? CD Duplicator case

Cheap and comes with a power supply and integrated fans for cooling. For a backpane you could just use ribbon cable with multiple IDC connectors, like pre SATA HDD's were connected.

You can even bump up to 7 or even 9 bays easily as you add modules.

[Rerouter]

As for a case, how about one of these? CD Duplicator case

Cheap and comes with a power supply and integrated fans for cooling. For a backpane you could just use ribbon cable with multiple IDC connectors, like pre SATA HDD's were connected.

You can even bump up to 7 or even 9 bays easily as you add modules.

I Really like this idea,

[Felicitus]

There's already such a thing called c't Lab (from a popular german computer magazine). Unfortunately, most of the documentation is in german, but using google translate shouldn't be too hard. And if you have troubles, ask me, I can translate.

Their wiki can be found on http://www.thoralt.de/wiki/index.php/Hauptseite and the project site is on http://www.heise.de/ct/projekte/machmit/ctlab/wiki. It's pretty much complete and I know quite some germans who have built it.

[Bloch]

There's already such a thing called c't Lab (from a popular german computer magazine). Unfortunately, most of the documentation is in german, but using google translate shouldn't be too hard. And if you have troubles, ask me, I can translate.

But is it truly open source ?

I do understand that the software is

But that about the hardware ?

Will you translate this a bit better than google is :-) Thanks !

"Auch kommerzielle Eigenentwicklungen aufgrund der veröffentlichten Schaltungen und Grundlagen sind zulässig, allerdings keinesfalls eine 1:1-Nachfertigung der Platinen, die dem Urheberrecht des Autors und des heise Zeitschriften Verlags unterliegen; hier ist eine eigene Erfindungshöhe der Neuentwicklung Voraussetzung."

[Marco]

They are saying you have to layout the PCB from scratch if you want to use it commercially with your own PCBs.

Basically the firmware sources are attribution only licensed and the PCB layout sources are attribution non commercial licensed (they are just being honest in recognizing that copyright can't cover circuit topology, so a completely new layout renders their copyright irrelevant).

[Bloch]

(they are just being honest in recognizing that copyright can't cover circuit topology, so a completely new layout renders their copyright irrelevant).

So a big differences between 

pls dont copy it |  pls do copy it

[uwezi]

I mean even autoranging might be a challenge, though off the top of my head I think I can figure out how to do it...

Who wants autoranging anyway? 

[mrflibble]

what I care about is more on the agent side.  the server side, if you will.  that will be written in c since net-snmp is a C implementation and C is my best/preferred language.  the platform will be anything linux: from raspi to beaglebone to itx box.  it should not matter; the code will run on any linux system.

Any progress on this?

[linux-works]

I got side-tracked to a few other projects (lots of re-capping and refurbing of used gear I recently bought) but I do still plan to work on this and post the prototype C code (and MIBs, lol) that I end up with.