Arbitrary Waveform Generator!!!

would open up a lot of options for me.

would open up a lot of options for me.

The current sig gen (CS700A) uses an AD9834 DDS sine wave generator chip to make sine, square and triangle. However, we do have control of the DC with a seperate 10 bit DAC, and could use this to make an AWG with a sample rate of 1 MSPS. Would this be good enough for many applications do you think?

The current sig gen (CS700A) uses an AD9834 DDS sine wave generator chip to make sine, square and triangle. However, we do have control of the DC with a seperate 10 bit DAC, and could use this to make an AWG with a sample rate of 1 MSPS. Would this be good enough for many applications do you think?

1Msps is a bit too low for our application.

we need to be able to output a 10usec long chirp from 1-to 10MHz.

1Msps is a bit too low for our application. we need to be able to output a 10usec long chirp from 1-to 10MHz.

If you need a sinusoidal chirp, we can do that with the AD9834. We can manage an update every 1us. We do already have the facility to load a table of frequencies, and intervals. Is that useful at all?

If you need a sinusoidal chirp, we can do that with the AD9834. We can manage an update every 1us. We do already have the facility to load a table of frequencies, and intervals. Is that useful at all?

Now that I've experimented a bit with a new Cleverscope, I find it very useful, but the way the Bode plot / Frequency response is implemented is an ugly kludge (although useable). A much faster and more efficient way would be to generate a stimulus whose spectal content is white, and to obtain the Bode plot from a single capture+FFT cycle. No sweep.

If the CS701A device indeed has the capability to generate a chirp, it sounds to me that all the pieces are available. That's how the HP/Agilent 89410A does the freq response (an extremely useful analyzer, by the way, we have three of those in our lab and regret that they are no longer manufactured).

I'm curious about the facility you refer to - is it distributed (e.g. as a part of the DLL library) or have you created it for your internal use only?

Regards,
Mikko

[quote=BartSchroder]If you need a sinusoidal chirp, we can do that with the AD9834. We can manage an update every 1us. We do already have the facility to load a table of frequencies, and intervals. Is that useful at all?[/quote] Now that I've experimented a bit with a new Cleverscope, I find it very useful, but the way the Bode plot / Frequency response is implemented is an ugly kludge (although useable). A much faster and more efficient way would be to generate a stimulus whose spectal content is white, and to obtain the Bode plot from a single capture+FFT cycle. No sweep. If the CS701A device indeed has the capability to generate a chirp, it sounds to me that all the pieces are available. That's how the HP/Agilent 89410A does the freq response (an extremely useful analyzer, by the way, we have three of those in our lab and regret that they are no longer manufactured). I'm curious about the facility you refer to - is it distributed (e.g. as a part of the DLL library) or have you created it for your internal use only? Regards, Mikko

I checked the DLL driver documentation, and it looks you can update the signal source frequency every 20ms approximately via the DLL call. The utility you mentioned is much faster at its 1us update rate capability.

If we take the extreme high frequency case and wish to measure Bode plot up to 50MHz, the 16384-point FFT would imply ~160us capture time (neglecting complications such as padding and windowing). During the capture time you could change the stimulus frequency 160 times, and get 160 frequency bins to the plot. The single-frequency approach (the Guided Tour p.16) would only give one bin per capture.

Repeating the procedure 50 times with shifted stimulus frequencies (to fill the missing freq bins) would give the full 8000-bin Bode plot in 2.5 seconds, assuming 20 captures-per-second. I guess you could use the same peak-registering trick (Guided Tour) as with the single frequency version, which allows you to fill just the bins where there was spectral content available in the stimulus? In the single frequency version the procedure must be repeated 8000 times to get the full plot and it takes 7 minutes.

I assume there is effectively no settling time when the DDS chip switches frequency, unlike a PLL? Even if there was, and the stimulus spectrum got smeared, that only would help to fill the bins between the 'official' hop frequencies. Because the device input is fed as a reference to the ChanA, the uniformity of the stimulus spectral content is not that important in small-signal measurements, as long as there is finite amount of power at all the frequency bins. SNR in each freq bin would vary but often this is not a problem. Large-signal (non-linearity) measurements are a different story.

It would be great to be able to replace the HP89410A with Cleverscope when tuning filters or struggling with amplifier stability issues. For this to work, the Bode plot update rate should be sufficiently rapid that you see the effect of adjusting a trimpot or moving your hand close to the unstable transistor (causes capacitive loading) in real time. It would be even greater to be able to extend the plot above the 10MHz limit of the 89410A, without having to set up a full VNA.

Regards,
Mikko

I checked the DLL driver documentation, and it looks you can update the signal source frequency every 20ms approximately via the DLL call. The utility you mentioned is much faster at its 1us update rate capability. If we take the extreme high frequency case and wish to measure Bode plot up to 50MHz, the 16384-point FFT would imply ~160us capture time (neglecting complications such as padding and windowing). During the capture time you could change the stimulus frequency 160 times, and get 160 frequency bins to the plot. The single-frequency approach (the Guided Tour p.16) would only give one bin per capture. Repeating the procedure 50 times with shifted stimulus frequencies (to fill the missing freq bins) would give the full 8000-bin Bode plot in 2.5 seconds, assuming 20 captures-per-second. I guess you could use the same peak-registering trick (Guided Tour) as with the single frequency version, which allows you to fill just the bins where there was spectral content available in the stimulus? In the single frequency version the procedure must be repeated 8000 times to get the full plot and it takes 7 minutes. I assume there is effectively no settling time when the DDS chip switches frequency, unlike a PLL? Even if there was, and the stimulus spectrum got smeared, that only would help to fill the bins between the 'official' hop frequencies. Because the device input is fed as a reference to the ChanA, the uniformity of the stimulus spectral content is not that important in small-signal measurements, as long as there is finite amount of power at all the frequency bins. SNR in each freq bin would vary but often this is not a problem. Large-signal (non-linearity) measurements are a different story. It would be great to be able to replace the HP89410A with Cleverscope when tuning filters or struggling with amplifier stability issues. For this to work, the Bode plot update rate should be sufficiently rapid that you see the effect of adjusting a trimpot or moving your hand close to the unstable transistor (causes capacitive loading) in real time. It would be even greater to be able to extend the plot above the 10MHz limit of the 89410A, without having to set up a full VNA. Regards, Mikko

I must correct my yesterday post: Bode plot above 10MHz is obviously not possible using CS's internal generator, as its max freq is 10MHz. One must use an externally generated stimulus.

Considering the real-time Bode plot up to 10 MHz, an excerpt from this forum...

2010-03-02 21:56:07 BartSchroder
""And now it is my turn to be confused. You say the Cleverscope output file gives 15us. Are you using the Cleverscope application to do bandpass response, or the example 'Bandpass response.vi' which uses fast sig gen stepping inside the acquisition unit to do the response?""

... suggests that the Bode plot with a fast-stepping chirp *has* been implemented, and is available in some corner or other of the CS web page.

By the way, I'm rather impressed by those Cleverscope capabilities which I've been able to figure out so far...

Regards,
Mikko

I must correct my yesterday post: Bode plot above 10MHz is obviously not possible using CS's internal generator, as its max freq is 10MHz. One must use an externally generated stimulus. Considering the real-time Bode plot up to 10 MHz, an excerpt from this forum... 2010-03-02 21:56:07 BartSchroder ""And now it is my turn to be confused. You say the Cleverscope output file gives 15us. Are you using the Cleverscope application to do bandpass response, or the example 'Bandpass response.vi' which uses fast sig gen stepping inside the acquisition unit to do the response?"" ... suggests that the Bode plot with a fast-stepping chirp *has* been implemented, and is available in some corner or other of the CS web page. By the way, I'm rather impressed by those Cleverscope capabilities which I've been able to figure out so far... Regards, Mikko

Hello Mikko,
No, the chirp has not yet been implemented, though if you read the Cscope Control Driver vi description.pdf carefully you will see some new commands for the Link Port, in which we are implementing the capability to do chirp. But right now, when you capture multiple sequential frames (up to 3000), you can automatically step the signal generator between captures. This approach means you get quite a few frequency sample points, and also you can have windowing on each capture. Because the acquisition unit manages the acquisitions it is much faster. We have an example 'Bandpass response' in the labview control driver collection. We will eventually build a fast bandpass response into our Cleverscope Application. Surprisingly demand for it has been low, and so it has taken a back seat to other capabilities (eg Charting and Linux).

Hello Mikko, No, the chirp has not yet been implemented, though if you read the Cscope Control Driver vi description.pdf carefully you will see some new commands for the Link Port, in which we are implementing the capability to do chirp. But right now, when you capture multiple sequential frames (up to 3000), you can automatically step the signal generator between captures. This approach means you get quite a few frequency sample points, and also you can have windowing on each capture. Because the acquisition unit manages the acquisitions it is much faster. We have an example 'Bandpass response' in the labview control driver collection. We will eventually build a fast bandpass response into our Cleverscope Application. Surprisingly demand for it has been low, and so it has taken a back seat to other capabilities (eg Charting and Linux).

I have wanted an ARB capability, or even a swept frequency chirp waveform, for a long time. How is this development coming along? Determining the bandpass characteristic of a network seems like such a common need, I'm still surprised there hasn't been more demand. I hope that you are still willing to pursue this development.

I have wanted an ARB capability, or even a swept frequency chirp waveform, for a long time. How is this development coming along? Determining the bandpass characteristic of a network seems like such a common need, I'm still surprised there hasn't been more demand. I hope that you are still willing to pursue this development.