Gang,
Using the new ESS dac chip and not real keen on the filter selections. Has anyone done a slow roll off or a minimum phase slow roll off in an ESS DAC via the registers?
Matlab is $880 a year and I know the QED2000 is probably cheaper, but I haven't done this stuff in decades.
Thoughts?
Thanks,
Gordon
Using the new ESS dac chip and not real keen on the filter selections. Has anyone done a slow roll off or a minimum phase slow roll off in an ESS DAC via the registers?
Matlab is $880 a year and I know the QED2000 is probably cheaper, but I haven't done this stuff in decades.
Thoughts?
Thanks,
Gordon
https://www.gnu.org/software/octave/index free, not sure how the filter design capabilities compare but I would tend to go 'from scratch' for this type of thing anyway.
One interesting feature of the ESS PRO series 8-channel dacs is that when operated in stereo mode, the filter can be reconfigured to use twice the default number of taps. None of the built-in filters make use of capability.
I've never done anything with ESS DACs, but if you PM me your e-mail address, I will send you two FORTRAN filter design programs that run under gfortran. One is a ported version of the classical Parks-McClellan linear phase equiripple program, the other is for Wilkinson's maximally flat passband filters. I don't have anything for minimum phase.
So this is the new ES9219 which is 2 channel. I have a program I wrote back in the 80's for FIR/IIR types. But man that PC has not been booted for at least 22 years.
I did a couple of dacs with the ES9038Q2M and the slow min phase really sounded best to me.
If anyone has downloaded some filter files into one of these chips I would like to hear your feedback.
So I know a couple of engineers are using short number of taps and getting really good results. So one question... say you load 24 coefficients, does the damn thing use all 127 regs or only 24?
Thanks,
Gordon
I did a couple of dacs with the ES9038Q2M and the slow min phase really sounded best to me.
If anyone has downloaded some filter files into one of these chips I would like to hear your feedback.
So I know a couple of engineers are using short number of taps and getting really good results. So one question... say you load 24 coefficients, does the damn thing use all 127 regs or only 24?
Thanks,
Gordon
I can't help you with that, but you can always add 103 zero coefficients when you have a FIR with length 127 while you need only 24.
i tried the digital filters on ES9018/28/38 (2 ch , q2m) and made also a test with "my" filter , custom loaded coeff. on ES9018.
the chip always runs the 127 coeff. , but you set all unused to 0x00 , so thats "normal" behavior for a FIR filter.
was a bit try and error, to get a working setting, because the documentation is a bit thin in the datasheets.
Sound for my ears is far less different, than i expected ; "slow" filter setting sound is better in "natural sounding" high range , so some kind of short or very short FIR filters are the way to go, because there is the shortest (pre-) ringing.
the chip always runs the 127 coeff. , but you set all unused to 0x00 , so thats "normal" behavior for a FIR filter.
was a bit try and error, to get a working setting, because the documentation is a bit thin in the datasheets.
Sound for my ears is far less different, than i expected ; "slow" filter setting sound is better in "natural sounding" high range , so some kind of short or very short FIR filters are the way to go, because there is the shortest (pre-) ringing.
I agree on the datasheets and really an errata would really be nice. I found a number of mistakes on this one as I think they use a data sheet from another part and then just make changes. Why not a PDF on filter designs with some examples, why not a PDF on THD? Why not a PDF of testing. I emailed Dustin but no answer as his group is pretty much locked down.
There is even a statement in datasheet to see the examples filters for more information???
So stage 2 did you guys just blast 0s into those?
So I was cleaning out my office the other day and came across the first ESS product the ES1204. This was an upsampler with SPDIF input and I2S & SPDIF output. I kind of laughed when I found it and sent Dustin a picture of the eval board. That was probably 22 years ago when Dustin called me and asked if I was interested in looking at it. I was just finished designing a USB controller for another company then and this was before Async was even thought of. So I was kind of interested in removing the jitter that way for the company I was designing for. But I don't think that ever really became a product. Anyway TI screwed up and so we ended up with Async USB so that was best for everyone.
Thanks,
Gordon
There is even a statement in datasheet to see the examples filters for more information???
So stage 2 did you guys just blast 0s into those?
So I was cleaning out my office the other day and came across the first ESS product the ES1204. This was an upsampler with SPDIF input and I2S & SPDIF output. I kind of laughed when I found it and sent Dustin a picture of the eval board. That was probably 22 years ago when Dustin called me and asked if I was interested in looking at it. I was just finished designing a USB controller for another company then and this was before Async was even thought of. So I was kind of interested in removing the jitter that way for the company I was designing for. But I don't think that ever really became a product. Anyway TI screwed up and so we ended up with Async USB so that was best for everyone.
Thanks,
Gordon
BTW just found the filter doc from 2015. It includes references to some text files which of course are not included.
Maybe I will cattle prod them again.
Thanks,
Gordon
Maybe I will cattle prod them again.
Thanks,
Gordon
about stage 2 : first set all koeff to zero, to be shure all is zero, then put a set of coeff. for desired filter;
to calculate a filter, i used a tool to just see effect of one filter stage:
http://t-filter.engineerjs.com/
to proof it, make a sound file with repeated single pulse and check the output with a DSO : puls shape looks like the simulated? then program is ok.
to calculate a filter, i used a tool to just see effect of one filter stage:
http://t-filter.engineerjs.com/
to proof it, make a sound file with repeated single pulse and check the output with a DSO : puls shape looks like the simulated? then program is ok.
Thanks I will try this out. A friend has QED2000 as well so I should be able to plug in the same stuff and get similar results.
Gordon