Zero-crossing output spikes from DAQ (pci-6229).

[MUSCRick]

We are using a pci-6229 M series to generate auditory stimuli (100-10,000 Hz).  We have noticed spikes at the positive and negative zero crossings of about +/- 100 mV (positive for the rising slope, negative for the falling slope).  Becasue the spikes are constant in voltage, at lower voltage swings of the D/A, they become audible to the listener (which is how we found out about them).  The spikes are short - about 1 or 2 samples in length.  So the question is: how can we smooth out the sinusoid at the zero crossings?  

We are running XP Pro SP2 on a Dell 2.X GHz machine and the 6229 is in a CA-1000 external box.  We are running Labiview 8.5.  The 6229 output is driving an amplifier with about 10K input impedance, but does the same when attached directly to a scope.  The spikes are also present when we use the test panel software to output a sinusoid in the Measurement and Automation Explorer suite (MAX).

I've taken over this project and am just feeling my way around the LabView system.  Any advice would be appreciated.

Rick

[PBear]

This sounds interesting and I'd like to see a screenshot of this phenomenon. Also could you describe the output charecterstics (shape, amplitude, output range, sampling rate, etc.) and more about the tools you are viewing this with. You said you can see this is MAX and I'd like to recreate this.

Again, any screenshots and details about the DAQ settings would be helpful. In the meantime, you could build a lowpass filter and this should help you out. Thanks!

[MUSCRick]

 

Hey PBear,

Thanks for the reply.  I've been out of the office on vacation so I apologize for the delay.  Attached is a quickie iphone shot of the waveform on a Tek osc at 250 us and 50 uV per division.  This is the direct output from one of the channels (both channels show the zerocrossing spikes).  The output is from the MAX which was set up with the following parameters:  sa rate 500000, 10V max pp, freq=500 hz, 200 mV pp out.  We have tried using an off the shelf low pass or bp filter, but they add to the noise floor. 

Any more ideas would be greatly appreciated.

Thanks

MUSCRick

PS:  I cant seem to get the image linked, but it can be found at:

http://rickschmiedt.smugmug.com/gallery/5769456_ekUZf#356389200_HLqu7

[Quantumbyte]

Hi MUSCRick,

The problem you are seeing is in fact the dissipation of charge that accumulates in the DAC during normal operation. Depending on your impedance, this will result in the glitches you are seeing. The Glitch Energy is define in the specs of this board as having a magnitude of 100mV and a duration of 2.6us as can be found in the specs: http://digital.ni.com/manuals.nsf/websearch/8117DF4C5A29C95C862573020061023B

If you are working on audio applications you could add filtering between the 6229 and your hardware. A low-pass filter should be able to get rid of that glitch, without affecting your audio application too much. On the other hand, the 6259 has a specified glitch of 10mV with a duration of 1us. 

I hope this helps.

Quantumbyte

DAQ Software R&D 

[dousley]

So these spikes are audible. And you are using a 100Hz-10kHz bandwidth. And you're going from the 6229 to an analog amplifier right? And you tried an off-the-shelf filter and it's noise performance was unacceptable.

Since the glitches are audible in your case, using a filter to get rid of the glitch noise might also start to cut in to your audio bandwidth. I wonder if you could get around the glitch effect in another way. If you didn't mind dropping down to 15-bits of dynamic range, I suppose you could prevent DAC glitching by cutting your signal in half and giving it a DC offset of 1/2 full scale. That would make your output unipolar and prevent zero-crossings from the DAC. Of course you would have to AC couple your amplifier downstream.

Another hardware solution, in addition to the 6259 already mentioned, for generating audio would be something from the dynamic signal acquisition product line. Something like a PCI-4461, for example, would give 24 bits at up to 204.8kHz and has an architecture and specs more focused on audio and vibration type applications. 

Hope that helps.

Dan O

Conditioned Measurements Hardware 

[MUSCRick]

Quantumbyte and Dousley:

Thanks to both of you for clearing up some of our questions.  First, unfortunately when this NI package was spec'd, the person responsible didnt realize that there were boards better suited for auditory use.  Second, the programmer is not optimizing the board's output very well in that a better setup would be to output only large voltage swings so that these mV artifacts stay reletively far down in the noise floor.  Again, the program attenuates the signal by reducing the voltage output of the board rather than controlling an outside attenuator to do the job, which is a lousy way of keeping any kind of S/N ratio.

Anyway, a passive low pass filter might do the job and certainly biasing the waveform up with a DC offset to get rid of the glitches would also work.  Or maybe it would be better to just order the right board!

Thanks again, I really appreciate your time and efforts.  Now I just gotta scratch my head a bit more.

MUSCRick