Home > On-Demand Archives > Talks >

The Quest for the Perfect Asynchronous Sample Rate Converter

Wessel Lubberhuizen - Watch Now - DSP Online Conference 2022 - Duration: 37:13

The Quest for the Perfect Asynchronous Sample Rate Converter
Wessel Lubberhuizen
In audio applications you often need to convert audio streams from one sampling domain to another. When the clocks of these domains are unrelated and possible drifting, you’ll need an asynchronous sample rate converter. In this talk we'll discuss ways to implement these with a minimal number of multiplications per sample, using multistage, multirate IIR filters based on all-pass filters and halfband filters, and polynomial interpolation using normal, transposed and partially transposed Farrow structures.
M↓ MARKDOWN HELP
italicssurround text with 
*asterisks*
boldsurround text with 
**two asterisks**
hyperlink
[hyperlink](https://example.com)
or just a bare URL
code
surround text with 
`backticks`
strikethroughsurround text with 
~~two tilde characters~~
quote
prefix with 
>
gretzteam
Score: 0 | 2 years ago | 1 reply

Fantastic presentation thanks lot.
Two questions...

  1. The interpolating filter has a 100db stopband yet the overall response of the rate converter only shows 100db for a tiny band near 24k and then goes down much lower, likely because both the interpolator and decimator combine for much of the response. If the rate converter was going from 48k to 192k, the decimator wouldn't combine with the interpolator so is it fair to say you'd get 100db attenuation for ~26k to 96k?
  2. In the case of the interpolating asrc (say 48k to 192k) do you need to decimate at all or the transposed farrow could just run at the output rate?
wesselSpeaker
Score: 0 | 2 years ago | no reply

Thanks!

  1. Yes, that is correct. However, it is possible to increase the stopband attenuation to ~200 dB by running the last output lowpass filter without decimation, e.g. 192 kHz input and 192 kHz output.
  2. In this case you can skip part of the decimation chain. You still need some decimation, otherwise you will get droop from the transposed farrow response, because it has zeros are at multiples of the output sample rate. But you can skip last output lowpass filter if you are happy with 100 dB image suppression.
napierm
Score: 0 | 2 years ago | no reply
This post has been deleted by the author
wesselSpeaker
Score: 0 | 2 years ago | no reply

I have used scripts from this website to compute coefficients for the polyphase IIR up/downsampling filters:
http://vadkudr.org/Algorithms/EMQFdemo/EMQFdemo.html
The scripts can be downloaded from this link:
http://vadkudr.org/Algorithms/EMQFdemo/EMQF.zip

OUR PARTNERS