Is there maximum value for product of avgCount and FFTSize in SlowSweep?

Hello,

I am using the linux API and have been reading the manual and testing a few things with the SlowSweep function. The manual states that the product of avgCount and FFTSize must be evenly divisible by 512.

Here is the code I am running:

#include <cstdio>
#include "CUSBSA.h"
using namespace std;

int main()
{
    CUSBSA mySignalHound;
    mySignalHound.Initialize();
	mySignalHound.Configure(0.0,1,2,1,0,0);
	mySignalHound.SetPreamp(1);  //turn preamp on

    int i,returnCount = mySignalHound.SlowSweep(162.390e6, 162.56e6,8192,16,0);
	printf("#RBW: %f Hz\n", mySignalHound.m_dCalcRBW);
	
    for(i=0; i<returnCount; i++)
		printf("%3.8e   %3.3f\n", mySignalHound.dTraceFreq[i], mySignalHound.dTraceAmpl[i]);

    return 0;
}

I can run SlowSweep with FFTSize=4096 and avgCount=30 and successfully get the result. But if I try to increase avgCount to 32, it looks (judging by the front panel LED) like the SA44 does the measurement, but hangs at the point where it is sending data back to the host.

If I lower FFTSize to 2048 and avgCount to 32, it works. If I run 8192 and 14, it works, but 8192 and 16 fails. Based on these and other tests, it seems that
(FFTSize x avgCount) / 512 must be less than 256
for the SA44 to successfully perform the measurement and return the data.

Is this correct? and expected?

Thanks,
DAN

I think I answered this on Facebook last week. In essence, for the legacy API, yes. I recommend keeping FFTSize * AvgCount <= 65536, although up to 512*255 will work. The new Spike API has no such limitations.

[Author]

Posts