How does buffer size affect double buffered waveform generation?

[Vadi]

I had originally posted the following question:

"Why does the double buffered waveform generation pause after the first buffer before continuing?"

"I am using an AT-AO-10 board to generate a multiple channel waveform in double buffered mode. The board's DAC's are updated by an external clock signal. While the waveform generation performs well, I notice that after the first buffer has been generated there is a time delay before the next buffer is output. However the second buffer and thereafter perform well without any time delays. If anyone can provide me an explanation on why this happens I would appreciate it. I am using NI-DAQ API functions to generate the waveforms and my settings for the WFM_DB_Config function are 1 for oldDataStop to disallow regeneration of data and 0 for partialTransferStop to not stop when a half buffer is partially transferred."
-posted by Vadi on 6/7/2001

I received a response from Geneva as follows:

Geneva L. on 6/11/2001 says:
"Vadi,

The first thing is to make sure that you have the latest version of NI-DAQ installed, NI-DAQ 6.9.1. If you need to install it, make sure you completely uninstall any prior versions. Then, you will have examples installed in either the NI-DAQ or the CVI directory. In the AO directory, you should find the WFMdoubleBuf example.


Start with that to make sure the output appears as you expect. Then, you can modify it to apply your external update clock, following the idea presented in the WFMsingleBufExtUpdate example. You might even want to double-check that your external clock acts as you expect using an oscilloscope.


Finally, modify the example such that you can update on multiple channels, remembering that you interleave each channels buffer into one buffer for WFM_DB_Transfer. Whatever data is in the buffer will be updated on the output channels.


Regards,
Geneva L.
Applications Engineer
National Instruments"

I have checked my version of NI-DAQ and it is 6.9.1. I am generating the double buffered waveform according to the format shown in WFMdoubleBuf and with some modifications from WFMsingleBufExtUpdate to allow me to use my external update clock. However I continue to notice the same phenomena again and again. For a buffer size of 7500 or 10000 points there is a time lag meaning after the first buffer has been output there is a noticeable time delay before the second buffer and buffers there after is output. This time lag doesn't exist for the buffers that are output after the first buffer but it does exist for the first buffer. When I decrease the buffer size down to 5000 points the time lag disappears (Note: this phenomena also occurs when I use an internal time base as opposed to my external clock). Is there a reason for this? I am using a AT-AO-10 board and I know the on board FIFO is 1024 points deep. However from the documentation provided it doesn't indicate that double buffered mode uses the on board FIFO at all. In fact, the functions require that the FIFO mode be turned off (in WFM_Load) for double buffered waveform generation. Is there a reason why when the buffer size is increased that there is a time lag after the first buffer? Is this because of the functions themselves or is this because of the AT-AO-10 board?

[Geneva L.]

Vadi,

Make sure that your buffer size is set to the same number of points that you're actually writing to the buffer initially. For instance, if you run the example as-is, the NIDAQMakeBuffer puts exactly the ulCount amount of data into the buffer. Then, it continuously writes out half buffers. Thus, if you are not writing enough data to fill up the buffer the first time, there will be that lag until the section where half buffers are output.

Regards,

Geneva L.
Applications Engineer
National Instruments
http://www.ni.com/ask

[Vadi]

Thanks Geneva

I am such a clutz, I wasn't filling up the buffer.