Noise preventing of PCI 6713

[Tomas_E]

The load is 1MOhms, and i find the analog output of 5 v has much noise, about 15mv rms, how can i get rid of this without the SCXI module, and if i use, what kind of product should i bu? What is the noise limit of the product?

[<Mike Ross> mross@smpcorp.com]

Tomas,

Noise is alway difficult to solve. Can you be more specific regarding the characteristics of the noise?

For example if the noise has a well defined frequency you can use a particular set of tricks. If the noise is random and the 15mv is the maximum, that is another set of tricks.

Also, the information you are trying to gain from the acquistion dictates how the noise should be removed/suppressed.

A first step would be to average the data. This can work very nicely, and can be done as a continuous process, or applied to a data set after acquisition.

Mike Ross
mross@smpcorp.com

[Tomas_E]

Thanks. the noise contains the following frequence: 800KHz and a few of MHz. I should get the bandwidth for 1 MHz. I guess the noise come from the space of the room. But it is strange that the noise is much larger than the specific noise of PCI-6713 which is only 200 uV rms.

[<Mike Ross> mross@smpcorp.com]

Tomas,

You did not explain what information is contained in the signal of interest.

For instance, if you are trying to find the exact time or position of some event like the crossing of a voltage level, then filtering, by analog or digital means might be inappropriate since they would cause an offset in time.

If you only wish to visually monitor the signal, then a simple running average would be a good choice.

I have a motor drive that generates bursts of 2 MHz to 4 MHz sine wave noise. This radiated noise can easily exceed 15mV (actually it can exceed 800mV). There are all sorts of power line transients that can be greater than 5V in my system.

There is a rule you should consider: Stop or shield the noise, do not filter it.

Question: If you keep the noise (all +/-15mV worth of it), what problem is caused by this uncertainty of measurement? If it does not cause a problem, then do not fix it.

If it is truely unacceptable, can it be stopped, removed, shielded or obscured?

Stop the noise by finding the source and turning it off.

Remove the noise by using differential input techniques (subtracts the noise common to the signal and its ground).

Shield it by moving the noise far away, or blocking it with the appropriate cabling, enclosures and grounding techniques.

Obscure the noise. One great way to do this is to boost the signal to a higher level. Then the 15mV noise is a smaller percentage of the signal. For example you could change from a 0V to 1V sensor to a 0V to 10V sensor. 15mV is 1.5% of 1V, but 0.15% of 10V. You should always try to use the entire resolution of your DAQ card. If a channel has 12 bits of resolution over 10V, and you acquire a 5V signal then 6 bits of resolution is wasted.

The last thing to try is filtering the data itself. This will introduce distortions to the signal of interest.

I advocate an averaging routine if possible. Consider a noise signal of 1kHz. If you sample this at 10kHz you will see 10 samples of every cycle of noise. If your signal of interest is a 100Hz signal, the 10 samples would only include 1/10th of the signal of interest.

If you took the average of 5 samples before a datum plus 5 samples after the datum, the 1kHz signal would be averaged into nothingness without significantly affecting the signal of interest.

By averaging before and after the point of interest you do not affect the timing of the signal.

To do this you need to acquire the signal into a buffer and post-process it to generate the average.

You can filter the noise with analog circuitry. Be aware that this can distort and shift the signal of interest. If the signal shifts in a predictable manner you can correct for it, but the distortions would be come part of the data collected.

If you use the digital filtering available as part of LabVIEW, you should familiarize yourself with how these filters change the signal of interest. You could simulate the signal of interest and apply the various filters and compare their inputs and outputs. This is very educational.

These filters can change your signal in very important ways. A square wave may be very much less than square after filtering, for example.

One last comment about the noise with respect to the 200uV rms of the card: The noise is a real phenomena. It is really that loud and it isn't really strange at all. It is something to become accustomed to.

It will be a big help if you use the best oscilloscope you can lay your hands on in trouble shooting your system.

Mike Ross

[Filipe A.]

Tomas;

Noise is always a difficult issue to debate. I'm attaching a good Application Note that talks about noise and some methods to avoid it.

Hope this helps.
Filipe A.
Applications Engineer
National Instruments

[Tomas_E]

Thanks for your advice. I use PCI-6713 to generate a sequential control signal to some equipment which is very voltage-sensitive. I agree with you that filtering is not advisable to use, because it will get the Slew rate down, which i don't want. It is really hard to find the eletrical noise source, for there are too many equipment here in the room. It is not the noise coming from the space, i have use shield and the result is the same. Moreover, can it come from the 2-meter cable of NI?

[<Mike Ross> mross@smpcorp.com]

Tomas,

I think you are saying that you are using analog control:

I use PCI-6713 to generate a sequential control signal to some equipment which is very voltage-sensitive.

But then I think you are using digital control:

because it will get the Slew rate down,

I need to know the definition of these variables to help efficiently. For example, can you define voltage-sensitive? What is the threshold of poor performance?

What is an unacceptable slew rate?

Can you describe the ideal control signal and its sequential information?

Two ways to gain improvement are proper impedance matching for digital and analog control, and the use of schmitt triggers for digital control. Also, the noise can b
e on the ground line of the DAQ system.

I found a good article at the NI website. It is about using a unity gain follower to isolate the signal from the transducer and match impedances.

http://zone.ni.com/devzone/conceptd.nsf/webmain/CD57A73721E0612586256BAE0055CDD9?opendocument

I have used schmitt triggers to clean up digtal signals with good effect. If this interests you let me know I will tell you more.

Pay extremely close attention to grounding of all parts of the system. A single point ground could minimize ground loops. I had a problem with noise from a motor drive. The best reduction in noise came from driving a dedicated ground rod into the floor of the lab. I used this as a saftey ground for the DAQ system, oscilloscope, power supplies, etc. The mortor drive uses the building ground.

I once acquired some terrible noise from a small exposed piece of cable shileding brushing against a ground conductor. You cannot be to thorough with basic wiring practice
s.

Mike

[Tomas_E]

Dear Mike
Thanks so much that you help me a lot. I am using an analog control. The term "Slew Rate" indicates the time property between two updates. E.g. at time 0.000000s the output be 1V, and 0.000005s the output be 10V. If i use some kind of filtering, the voltage changing process from the two time point may be different.
In fact, i am trying to use the analog output to control some kind of laser, which is very current-sensitive(of course there is an exchange of voltage to current). My boss suggested me to get a relative stability of the voltage to 0.00001, and now i only get 0.001. I think i could describe the work i do is a high stable computer-control power supply.


Tomas

[<Mike Ross> mross@smpcorp.com]

Tomas,

I will try to restate you description, please tell me if I am interpreting it correctly.

The characteristices of your analog control signal is:

from a steady state of 1.00V +/-???V the signal changes state at maximum slew rate of 1.8 V/microsecond to 10.000V +/- 0.00001V
after some variable period of time the signal changes state at a maximum slew rate of 1.8V/microsecond to 1.00V +/-???V and so on.

Your problem sounds difficult and I am going to pass you on to higher authorities, so to speak. You need to be very specific in describing exactly the effect you want in order to get good advise on this subject. At the level of control you want, almost no detail is trivial.

For instance if you have an extremely current sensitive device to control that is very important information. In my statement of you problem I made some assumtions that may not be correct and could be crucial to your solution. I assumed the 1V low signal is always the same, if it needs to vary that is important. The speed of switching is importatnt, and I assumed that it happens slowly enough as to be insignificant.

First I think you should subscribe to the info-labview mailing list. The quality of advise may be is much higher. The email address to post a question is: Mailto:info-labview@pica.army.mil

To subscribe to this list go to (you should absolutely, definitely do this):

http://www.info-labview.org/resources.html#FAQ

Also, there is some good literature you should have. Keithley Instruments, Inc. publishes two good handbooks you should acquire: Low Level Measurements, Precision DC Current, Voltage and Resistance Measurments, and Switching Handbook, A Guide to Signal Switching in Automated Test Systems (your application sounds more like a high speed switching app than an analog control app - you may not be able to do what you want with the DAQ board you have). These books are free if you contact Keithley. You will have to call or write to get these books.

Their website is: www.keithley.com

In the US Phone:
Direct: 440-248-0400
U.S. Toll-Free: 800-552-1115
Fax: 440-248-6168

From Amazon.com you can get Grounding and Shielding Techniques, by Ralph Morrison.

http://www.amazon.com/exec/obidos/ASIN/0471245186/104-3492923-2811911

You should also, contact National Instruments directly. They have very good phone support services at 1.800.433.3488. Also, learn your way around the Knowledge Base at NI:

http://search.ni.com/?col=alldocs&layout=TechResources&ql=a

What you are trying to do may me difficult, voltage control of +/-10microVolts is not going to be easy (it is not easy to measure this accurately, and certainly not easy to control it accurately). I am not sure you can do this with thte DAQ card you have, and certainly not without some very carefully selected and assembled accessory equipment.

A last piece of advise, try to describe your problem fully to get good advise. The impedance and other characterists of the load (laser amplifier?) are going to be very important. This equipment is going to react with the current source significantly at the very low voltages you describe. If I were you, I would just present all the specifications of the laser immediately.

You are going to know a whole lot of new stuff by the time you get this one figured out.

Good luck,
Mike