General 2nd order EQ, no output, possible instability
homeCinema_v4_EQ_test4_kteq.dspproj
Fullscreen
FL_sigmadsp.txt
Download
0.999608536682770
-1.997355600291492
0.998027213657203
-1.997355600291492
0.997635750339973
0.999652539507289
-1.998533998473304
0.998954310532900
-1.998533998473304
0.998606850040189
0.999608517507141
-1.997355573382208
0.998027206822863
-1.997355573382208
0.997635724330004
Fullscreen
FR_sigmadsp.txt
Download
0.999113297428514
-1.995381481638307
0.996652887137440
-1.995381481638307
0.995766184565954
0.998999096263852
-1.995978767947492
0.996987759712467
-1.995978767947492
0.995986855976319
0.999394158831979
-1.997349258659638
0.998176708600197
-1.997349258659638
0.997570867432176
Fullscreen
EQ_test4-sub-input_sigmadsp.txt
Download
0.996611345269132
-1.995750317058764
1.000000000000000
-1.995750317058764
0.996611345269132
0.999813814903769
-1.999090002824030
0.999309310843528
-1.999090002824030
0.999123125747297
0.999540990111309
-1.998032423219722
0.998618600341722
-1.998032423219722
0.998159590453031
0.999778469273957
-1.998918462836794
0.999185940817843
-1.998918462836794
0.998964410091799
0.999865989509558
-1.999230285928081
0.999390742523215
-1.999230285928081
0.999256732032773
1.000110815322156
-1.999074975174754
0.999033221399557
-1.999074975174754
0.999144036721712
0.988415212221241
-1.000000000000000
0.000000000000000
-0.988415212221241
0.000000000000000
0.999899661219836
-1.998967658608988
0.999124621184411
-1.998967658608988
0.999024282404246
0.999921855066129
-1.999637527671358
0.999721040466995
-1.999637527671358
0.999642895533125
0.999940400531700
-1.999583032696322
0.999651459711273
-1.999583032696322
0.999591860242972
Fullscreen
SUBL_sigmadsp.txt
Download
0.991647431553728
-1.991629816096504
1.000000000000000
-1.991629816096504
0.991647431553728
0.999666000253696
-1.997839381278943
0.998191644383539
-1.997839381278943
0.997857644637235
0.999688256887761
-1.998713596031630
0.999061802720753
-1.998713596031630
0.998750059608515
0.998439061609961
-1.993521091772239
0.995302323954836
-1.993521091772239
0.993741385564797
0.999912701324058
-1.999272767935728
0.999372542179162
-1.999272767935728
0.999285243503220
0.998691859330308
-1.998691431242260
1.000000000000000
-1.998691431242260
0.998691859330308
Fullscreen
SUBR_sigmadsp.txt
Download
0.972190490094641
-1.971994411852773
1.000000000000000
-1.971994411852773
0.972190490094641
0.996952354360038
-1.996256049259608
1.000000000000000
-1.996256049259608
0.996952354360038
0.999185946996812
-1.996640767744438
0.997550090819109
-1.996640767744438
0.996736037815921
0.999761387648591
-1.999340206797406
0.999587307227953
-1.999340206797406
0.999348694876544
0.999125925062263
-1.996429669892274
0.997369453578130
-1.996429669892274
0.996495378640393
Fullscreen
SUBBR_sigmadsp.txt
Download
0.179597950399366
-0.353056807079249
0.173520097727176
-1.985525898938991
0.985587139986284
0.178571108268053
-0.354554269991437
0.176044662519803
-1.993947352187398
0.994008852983817
0.997210966537496
-1.996958214478877
1.000000000000000
-1.996958214478877
0.997210966537496
0.999952234496432
-1.999705311626191
0.999757314773260
-1.999705311626191
0.999709549269691
0.999821735142894
-1.998826599413335
0.999067587065327
-1.998826599413335
0.998889322208221
0.998993400000205
-1.997175402955129
0.998318072400452
-1.997175402955129
0.997311472400658
0.999345715621202
-1.000000000000000
0.000000000000000
-0.999345715621202
0.000000000000000
Hi
I experience a problem that is partly similar to this:
ez.analog.com/.../general-2nd-order-eq-does-not-pass-signal-after-project-is-closed-and-then-reopened
Background:
I have a rather big project and just wanted to change some parametric equalizers blocks to using custom IIR coefficients instead.
That allowed me to do the EQ design in another tool and just load the IIR coefficients from a file in SigmaStudio.
I wanted to do it this way because:
1) Some filter types in the other tool were not identical to the ones in SigmaStudio. I.e. it could not get the same behavior from the supported biquads in SigmaStudio.
2) It takes less time to transfer the settings and there is less risk of entering something wrong.
The problem:
When I load the IIR coefficients and upload the compilation to the DSP, I have experienced the following behaviors:
1) From one output channels, there is no sound. Even if I substitute the IIR filter block from this channel with the one from a working channel, there is still no sound. If I load another project/the old project, it works fine.
2) I sometimes got random pops from the channel. This could suggest filter instability but the filter coefficients are okay.
3) The output works fine. Even one time where there was no output, I increased the number of biquads from 10 to 11, but disabled number 11. For some reason that made the output reappear. However, I suspect that not all the filtering is correct/that some of it is not performed.
What have I checked:
1) That the filter coefficients provide a stable filter (poles inside the unit circle)
2) I have stimules and probes for all the filter blocks. The magnitude and phase plots in the probes look correct.
3) I have checked the compiler output. According to it, the current project is using 71% of MIPS, 47% of Inst, 42% of Data, 15% of Coef.
4) It got worse when I cascaded more IIR sections, like one block with 7 biquads followed by another block with 10 biquads. I am not sure if it is due to higher load, something about filter precision when cascading more biquads, or something else.
I do have an EVAL-ADAU1772Z as well but I have not tried to reproduce on that yet. It would be helpful to know an easy way to trigger the problem.
I would like to know the root cause and how to fix the problem for good.
Is the problem related to the IIR coefficients? Or is it something with the compiler? Or resources?
How do I debug this problem?
My concerns:
I am afraid to connect the output to power amplifiers, if the output suddenly could become unstable.
Best,
Kim
Edit Notes
Uploaded dspproj and iir coef files
[edited by: HiFiDSP2 at 6:13 PM (GMT -4) on 17 Oct 2023]
Audio Signal Processors
Audio Products
software
SigmaStudio 4.6 and 4.7
adau1442
SigmaDSP Audio Processors
Show More
Hello,
Thanks for supplying the project. It will take a bit to investigate this. I have the file and working on verifying the issue and get a handle on why...
Dave T
Thanks Dave!
For now I will avoid using IIR sections with custom IIR coefficients.
Maybe the problem it related to rounding somehow when cascading many filters.
Kim
I have now experienced a similar problem when using a FIR filter on the FL and FR channels. My project is like the one I uploaded, except that I use do not use custom IIR coefficients in any of the IIR sections and the project worked fine before adding the FIR filter.
I wanted to use 512 filter taps but discovered that caused the output to become silence - even if I just entered 1,0,..0 in the filter. I then tried changing the filter to only 10 taps and the the output was back. Then I tried 128 taps which caused some loud noise to be output to the left channel, even though the filter coefficients were 1,0...0.
The behavior makes me think that the MIPS load is too high and therefore the project is not able to run. However, I find it strange that the 128 taps causes a different behavior than the 512 taps. Also, with the 512 taps filter, compiler output estimations are only 87/56/71/28% of available MIPS/Inst/Data/Coeff.
Let me know if you need more information and if you have reproduced the problem with IIR and/ FIR coefficients.
Should I always be able to trust the compiler estimates or can it be that I have run out of resources without being able to see it in the estimates?
Please log in to reply.