Adding support for NI GPIB-ENET/100 (would contribute by PR)

[bytewarrior]

I have a National Instruments (NI) GPIB-ENET/100 Ethernet-to-GPIB box in my homelab. I have used it for several years under Windows 2000, but it has reached a certain age and now, and there are practically no drivers available any more. To make things worse, it uses a proprietary Ethernet/TCP-based protocol.

As I do not have the money to keep buying GPIB adapters and I don't want to buy a new one as the GPIB-ENET/100 is working perfectly fine (as long as there is a working software stack available), I have started to reverse engineer the protocol.

I have a working python implementation and I can talk to the devices behind the GPIB-ENET/100.

Question: Would it be possible to contribute a sort of driver/protocol to this project so NI GPIB-ENET/100 would be supported out of the box?

If I made a pull request, would it have any chances of getting accepted, provided that all requirements (code style, functionality, tests, ...) check out?

Guidance beforehand, how and where to implement a rather complicated proprietary protocol would be very much appreciated. I have an idea-- but I'm uncertain if it's correct:

• create a new protocols/ni_enet.py (or similar)
• wire it into tcpip.py with a new Session class (or should it rather live in gpib.py?)
• complete all necessary interfaces, write tests/doc

[MatthieuDartiailh]

This sounds interesting. We added support for prologue adapter in the past so I am not against it. However there a couple things I need to understand better before answerong your questions:

• what is the typical workflow to use the adapter ? I assume it is less invasive than prologix but more so than a USB- GPIB adapter since you must provide an IP.
• Could you use any VISA library with it or did it have to be NI only ?

To give you some visibility anything that deviates from a standard resource will require some work here. However depending on the rest of the integration, a portion of the code may live unde pyvisa-py.

[bytewarrior]

Hi, and thanks for replying to my question that fast!

The Prologix Adapter is like a blueprint for me, because what it does is very similar. I'll use this one as a guideline.

There is a discovery-mechanism built-in, which I can use to find these adapters on the network, so setting an IP address is not necessary (but it is required if the adapter is in another subnet of course, because the discovery will not find it then).

Once found/configured, one can connect to them. Then, you could scan the GPIB bus behind the adapter and automatically find all instruments attached to it. Probably very similar to other adapters.

My current research into the subject revealed that despite there are common wire ids with libnienet, the protocol has major differences in multiple places. The original NI 488.2 library is the only one to my knowledge that can talk to it. But the support for this particular adapter is now long gone, linux versions of the library will not work on recent versions of linux. Same for Windows, and probably also on the Mac version (I don't have a Mac).

I have used a Win98 VM with the complete NI stack installed as a reference.

My current plan would be like this:

1. Put everything specific to the protocol in an own file (like protocols/nienet100.py, because the predecessor GPIB/ENET is very different).
2. Look at the Prologix implementation and the existing code, try to come up with minimal-invasive changes (probably only some hooks required?) in pyvisa-py to make the protocol known and usable to the surrounding code.
3. Probably some minor changes will also be required in pyvisa. But I'll try to keep these to a minimum (same with the existing pyvisa-py Code)
4. Try to keep it as close as possible to existing standard resources.
5. Extend the existing tests, make sure nothing breaks.

I'm by no means an expert with VISA, it's an ongoing journey of learning for me. But it's one I'm willing to make.

[MatthieuDartiailh]

What is the format of the resource name you use ?

[bytewarrior]

NIENET100-TCPIP[board]::host address::INTFC

is what I would propose. Correct?

Specifying a port does not make sense in this case, as the port numbers are fixed, and the adapter uses multiple ones (four to be exact).

I'm aiming for a usage like this (correct?)

import pyvisa

rm = pyvisa.ResourceManager("@py")
rm.open_resource("NIENET100-TCPIP0::192.0.2.5::INTFC")  # binds board 0
inst = rm.open_resource("GPIB0::16::INSTR")

# Standard I/O
print(inst.query("*IDN?"))

And the discovery should look this this, right?

import pyvisa

rm = pyvisa.ResourceManager("@py")

# Auto-find all bridges on the LAN
print(rm.list_resources())
# results in ('NIENET100-TCPIP0::192.0.2.5::INTFC', 'NIENET100-TCPIP1::192.0.2.7::INTFC', ...)

[bytewarrior]

Created the PR: #587