Changes in Probe plots when sampling frequency modified
Hi,
I have a very strange behaviour with the Probe and Stimuli blocks.
I am using SigmaStudio 4.5.
With an ADAU1701:
- When using an Nth Order filter between input and output, the plotted frequency response in Probe window changes when I modify the project sampling frequency.
- If I modify a parameter of the filter (type, freq,...), the plot is corrected in the window
- the changes in frequency seem to be multiples, so I guess that they follow the ratios between sampling frequencies
- when using a crossover filter, the plot is also changing with the sampling frequency
- modifying parameters has no impact on the plots
- the changes do not seem related to the sampling frequencies ratio
With an ADAU1466:
- with an Nth Order filter, the plots do not change, but I have an error pop up message ('CELL: NFILTER1')
- the the crossover filter, same behaviour than the 1701
I suppose that the problem with the Nth Order filter is a bug.
Crashing for the 1466, and needing a 'refresh' for the 1701.
But with the crossover, this is very strange.
And I would not be surprised to have identical problems with other cells.
So the questions are:
- are there known bugs or limitations with the Stimuli/Probe feature?
- are the results depending on the sampling frequency of the project?
- if so, which frequency should I use to have a 'real' overall frequency response chart?
- can I be sure that the XML exports generated for several sampling frequencies are correct?
Many thanks for your help.
Raphaël
Audio Signal Processors
SigmaDSP Audio Processors
adau1701
adau1466
Hello Raphael,
The Stimulus and Probe is calculated in SigmaStudio and not on the DSP. But it does not seem like you are running into this issue.
When you change the sampling rate in SigmaStudio, do you propagate the new rate to all the cells? This needs to be done. You should not have to recompile but it may help in this case. So the probe cell will use the sampling rate shown in the top of SigmaStudio but the coefficients need to be recalculated for the new rate. So there may be a bug where it is not recomputing after the sampling rate is changed. This is why I want to know if you propagated the sampling rates to the cells.
So if the coefficients have not been recomputed then you will see a difference in the curves that is correlated with the difference of the two rates.
I would like to see your projects and test it here if you can zip them and attach the projects.
Dave T
Hi Dave,
I do understand that it is only simulation.
In that case, no DSP was connected.
I tried to propagate and/or recompile, but without success.
I join one example of the problems I have, with the 2 cells I talked about.
I am convinced that other cell types are surely affected, but I have not tried so far.
You will notice that the behaviour is opposite, with different proportions:
- doubling Fs will double Fc of the Nth Order filter
- doubling Fs will reduce Fc of the crossover Lo output, but not half (seems like 1/√2 )
And changing a parameter of the Nth order filter corrects the plot (refresh of the computations?).
Whereas this has no effect with the crossover cell.
Raphaël
Probe bug example.zip
Hi Dave,
I am sorry to insist, but I have no news since last week.
I know the situation is unusual with Covid-19 and lockdown restriction.
But I still hope to have some support...
I use Analog Devices DSP to realize filtering in active speakers with streaming capabilities.
These speakers are able to play from 44,1ksps to 192ksps audio files.
Therefore, the DSP program is updated each time the sampling frequency changes.
That is why I need to export the SigmaStudio project for several Fs.
And be sure that the audio characteristics of the speaker are constant.
Obviously, there are some problems with Stimulus/Probe feature.
But I can confirm that the Nth Order Filter block in the Export XML is unmodified when I change Fs.
Apparently, I need to change one of its settings for the XML to be modified.
For the Crossover Filter, I am not sure if the parameters are correctly computed.
The XML seems to change with the sampling frequency.
But the simulation results are wrong.
Parameter computation problem or simulation problem?
Many thanks for your help.
Raphaël
Hello Raphael,
You have some support. The Covid-19 situation is certainly a challenge and it has taken some adjustments for me to work from home. The family is here which is good and bad. When the leadership of ADI told us we needed to work from home for two weeks if possible, I knew it would be longer so I actually packed up my entire lab station and took it home so I can do a lot here. I just cannot run things at hot or cold temperatures but I rarely have to do that.
Now, your issue is a bit tricky to track down and so this is not a quick answer. I really need to see what you are doing in the project when you change sampling rates? I think I have mentioned this before but just changing the rate at the top of the screen is not enough. So let's get more detail. Do you mind sending me the three projects for the three rates? We can do this off of the forum.
I do have one other question regarding the master clock. Are you changing the master clock frequency when you change between some of these rates? The reason I ask is because to reproduce 44.1kHz and 192kHz require two different master clock frequencies. You will need one rate for the 44.1k, 88.2k and 176.4kHz rates and a different master clock rate for the 48k, 96k, and 192kHz rates.
Dave T
Please log in to reply.