Wide temperature drift in 6289

[Alok Damle]

Hi,

We have been using 6289 for one of our application where we have to scan between certain voltage every second. For a long time we were worried over the algorithm that we are using to calculate the voltages. however recently when we did a loopback to check what exactely happening we saw a major temperature drift in the voltages. 

over five minutes period we are experience the drift up to 0.5 mV. For our application, the absolute resolution that we are working with is very low. hence 0.5 mV of deviation leads to a major variation.

The reason that we concluded it to be a temperature drift is that earlierr the drive was very high. so when we placed an external fan near the MXI chasis, the drift started changing by over 2-3 mV in 5 mins. however now we have ensured that the temperature at the MXI is very well constant. the ambiant temperature is near 18 DegC. But still we are facing this drift. 

We have already checked the FAN of MXI chasis and it is working fine. we are operating it at High setting.

Following is the configuration that we are working with.

6289/ao2 -> -10 to +10 V

6289/ao3 -> -1 to +1 V

and the value of 6289/ao2 is configured as offset for 6289/ao3. 

Please see the attached picture of the loopback scan results.

Please tell me how to stabalize the output so that i can use if for my application. with the present drift in voltage its not very much suitable for our application.

Thanks and Regards,

Alok Damle.

[ChrisR.]

Hi, Alok.  The amount of drift you're seeing doesn't seem totally unreasonable for the device, so I'll assume that there's not a hardware fault.  Have you considered using the analog input to monitor the output and either adjusting the output or adjusting your algorithm to compensate for the analog output drift?

Chris 

[Alok Damle]

Chris,

I am very dissapointed to hear that a card which costs near 1500 usd with specifications relating to high precision would have this kind of temperature drift issue. 

Relative to its cost, is the card not supposed to be of commercial specification? I can understand about the in-accuracy of the card but a temperature drift when the thermometer showing constant temperature is something not acceptable from a commercial grade!

This project is about replacing an old electronic hardware (which is about 16 yr old design) with the latest in technology.  Unfortunately, the old hardware electronic is performing way better than this supposed to be high precision card.

Our technical manager indicates there is a design problem with the card.he would like to discuss the issue with your design team. Please send us the relevent contact information.

Regards,

Alok Damle.

Message Edited by Alok Damle on 03-18-2009 05:47 PM

[kikiduu]

---

Alok Damle wrote:

Hi,

however recently when we did a loopback to check what exactely happening we saw a major temperature drift in the voltages. 

over five minutes period we are experience the drift up to 0.5 mV. For our application, the absolute resolution that we are working with is very low. hence 0.5 mV of deviation leads to a major variation.

Following is the configuration that we are working with.

6289/ao2 -> -10 to +10 V

6289/ao3 -> -1 to +1 V

and the value of 6289/ao2 is configured as offset for 6289/ao3. 

Please see the attached picture of the loopback scan results.

---

What exactly do you mean by loop back test?How did you connect the signals?

0.5mV is better than absolute accuracy that the 6289 specification guarantees at +/- 10V.

Also before selecting a card it is always a good idea to check accuracy using the accuracy calculator.

[kikiduu]

---

kikiduu wrote:

0.5mV is better than absolute accuracy that the 6289 specification guarantees at +/- 10V.

---

The 0.5 mV error between readings may be due to relative accuracy of the card. I was not able get the relative accuracy specs from the data sheet.

[Alok Damle]

I understand the absolute accuracy and i can compensate the same in my calibration routine. However if you see the graph this drift is between the scans of same samples values written to the card. that 

Also regarding the loopback, we are using 4461 AI for the application. by loopback, i mean that the output of 6289/ao3 is connected to the ai0 of 4461. 

Now this 'relative accuracy' is new to me. can you please eloborate more? also if you can not find the same from the specification sheet, then where to find this then? cause this is the primary cause for us not being able to stablize our product

All we need is a card which provides a repeatable output accross the scan. I do not know whether this kind of card with drifts accross the scans (that too of unknown quantity over time) would be suitable for any application which demands resolution higher than 12 bit accross +/-10 volts range.

Alok Damle

[ChrisR.]

The 6289 analog output has better DC stability than the 4461 input.  The 4461 was designed for AC applications, and its specs list a DC accuracy of 0.03dB max at room temperature and 0.2dB max over temperature.  I figure that gives a temperature coefficient of something like 400ppm / °C.  Is there a reason you can't use the 6289 for your measurement as well?  What happens if you use it for your loopback test?

Chris

[Alok Damle]

We tried using 6289 AI for measurement and loopback. however the results are similar and noise higher. with 4461 we are getting noise much lower than the 6289. it could be because of the onboard filter.

The drift in AO is  same when loopbaked with one AI of 6289 itself. We are using 6289 in a coupled mode where i have coupled one AO to another and the later one is operating in +/- 1 V range this is to get better absolute resolution. 

we have also tried using 4461 for this output. however with 4461, we are facing another problem as that of accuracy within the scan along with the drift across scan. 

If you see the graph that i posted, i used multi colors to plot each scan. the idea was to check whether the slope of the scan is also getting affected or not. 6289 is providing us with a better result by keeping the slope intact within the scan and across scans. only issue is of a linear drift.

however with 4461 the drift is present and along with the drift another problem is that of the slop of the scan. there the whole scan is getting affected.

I am attaching the graph with 4461 ao as the output used for same scans.

I am really in a fix for getting our team to purchase so much expensive card and now the card is really not functioning up to the specification.Please tell me what to do next.

Alok Damle.

[kikiduu]

Alok,

Before drawing any conclusion we need to consider the following facts.

1. AO of 6289 has an absolute accuracy of only 0.259mV at +/- 1V range

2. AI of 6289 has an absolute accuracy of  0.980mV at +/-10V range

3. AI of 4461 also has an absolute accuracy which I guess will be better than 6289 (if you discount the residual offset of 5mV)

Now you are using AO of 6289 to generate a voltage and using 4462 to acquire  the generated voltage. So the final measured value will have inaccuracies due to (1) and (3), which CAN be in the range of 0.5mV. 

Previously data sheets used to mention relative accuracy, which is the difference between two measurement and is a function of ADC resolution and system noise.

So to accurately measure the effects of 6289 AO, you will need a much more accurate device. 

BTW as mentioned in your first post you are looking for an absolute accuracy much better than 0.5mV. If this is the case, drift or no drift, 6289 is definitely not the hardware that you should be using.

[Alok Damle]

I think some issue with the basics.

As far as my understanding accuracy is related difference in the expected and actual output. however it does not tell anything about precision. We are ok with ANY absolute accuracy till the cards gives precision output (what NI markets them as). 

Here we are getting very low precision of the instrument which is the only issue. reaolution wise for our application both 4461 and 6289 are ok. however stability or precision wise none of them are performing up to expectation. 

Least to expect from cards costing this much is a good precision if not accuracy.

Alok Damle.