TNT 4882 chip read errors

[roni]

Hi,

I have a problem with the TNT4882 chip I am using.

First of all ,let me give a description of the system:

On one side of the GPIB cable, i have a computer controller using PCI GPIB
by NI, under WinXP. the NI.4882 Ver in the PC is 1.70.

On the other side of the GPIB cable I have a board we developed, using the
TNT4882 chip in Turbo+7210 One-Chip mode.

The sequence The PC performs is as follows:

ibwrite(string1) - no ACK on this string.

ibwrite(string2) - ACK on this string

ibwait for serial poll

serial poll

ibread.

...What I noticed is that depending on the GPIB communication load, the TNT does not receive the full string2, but misses the first character  -  (this happens 1/a lot of times this seq is  performed

.

As soon as the load is LOWER - it happens more often !!

Using the spy utility on the host computer shows that the string is sent OK:

1. Is it possible that although the spy says it was sent, it actually was not ? - Could an analyzer be of help ?

2. How is it possible that only a partial of the string is received ?

3. Can it be to do with the T1 timing parameter ?

I am a listener the whole time

Thanks,

Roni

[Collin M]

I need to ask a few clarifying questions first:

1. You said you were using "Turbo+7210 One-Chip mode". These modes are actually mutually exclusive and cannot be enabled simultaneously. You are probably using One-chip mode but can you verify this? There is a "One Chip Mode" bit in one of the registers (don't remember which one) that you set to enter One-chip mode.

2. Can you clarify what you mean by an ACK following an ibwrite? Does this occur at your device or at the host. Do you get any errors from the NI-488.2 driver at the host in either case? Do you get any kind of status or error information at your device? Also, attaching a few NI-Spy captures of the different ibwrite cases might be helpful.

I'll answer your specific questions:

1. Is it possible that although the spy says it was sent, it actually was not ? - Could an analyzer be of help ?
NI-Spy just looks at the status of the transfer from the host system. As long as there was something at the device-side toggling the appropriate handshake signal then NI-Spy will report a successful transfer (as it should). An analyzer may be helpful but don't think that would reveal the problem.

2. How is it possible that only a partial of the string is received ?

If NI-Spy and the NI-488.2 driver report a successful transfer I am pretty sure the entire string did get transferred and was received correctly. There may be a device-side firmware issue that only causes part of the string to read from the TNT.

3. Can it be to do with the T1 timing parameter ?

Again, if NI-Spy and NI-488.2 don't report any errors I doubt this is a physical layer or bus timing problem. Also, the T1 parameter is used by the  GPIB interface that is transmitting the data (ibwrite) rather than the device receiving the data. If you are using an NI GPIB controller and  TNT4882 in your device then just about any T1 setting should work for any reasonable cable length.

It sounds as though your system is almost working and just needs a few tweaks to get everything working correctly. Would you be willing to post your device firmware to this message or email it to me?

[roni]

What is your email for sending you the source code ?

[Collin M]

You can email me on yahoo. My user name is ckmurphy2000.

[Collin M]

This issue is a known problem in the ESP-488 source code and is documented at the following link:

http://digital.ni.com/public.nsf/allkb/e9dbf30a78c1e27b8625646c0080b39b

[roni]

Hi

Issue still not solved. The ibread has nothing to do with it - happens on unfixed code as well, sorry for the misslead.

The same issue occurs - any more ideas ?

Benjamin

[Collin M]

I have a few more suggestions/questions:

1. Have you made any other modifications to the ESP-488 code?

2. Can you run NI-Spy on the host system and capture the failing case?

3. Also, are you still using NI-488 v1.7? This is a rather old version 
of the driver. Can you upgrade to v2.4? The latest version is here
http://www.ni.com/support/gpib/versions.htm

4. You made a comment that adding a cable changes the frequency of the 
problem. What is the original cable length and the new cable length?

[roni]

Hi

The initial modifications definately improved matters ,I think it is in that direction.

1. No. You have recent code.

2. Yes, will send to your yahoo acnt.

3. I'm currently downloading. Do you know of any fix ?

4. Just by adding a standard GBIB connector with cable to back of connector makes problem appear more frequently. The other side of the connector is in the air. Length of cable is 1 meter.

Any newer fix to ESP 488 ?

Benji

[roni]

Hi

The 2.4 ver does not help.

Any ideas ?

[Collin M]

Let me clarify my last question. Are there any modifications to the original ESP-488 code you received from NI?

Also, I attached your Spy capture to this thread, renamed from capture.spy to capture.txt

Collin