[sdiy] 1-quadrant multiplier with 2164

Eric Brombaugh ebrombaugh at gmail.com
Thu Jan 20 20:43:09 CET 2022


I've used a phase vocoder to do it. Similar to the STFT approach that 
Mike described:

* Chop the incoming signal into FFT-sized records with overlapping.
* Window and FFT.
* Convert the frequency domain data to polar notation and and 
differentiate phase from the previous pass.
* Scale phase delta and magnitude to new bins for the shift.
* Accumulate the scaled per-bin phase deltas in output accumulators.
* Convert phase accumulators and magnitude back to cartesian.
* Inverse FFT and overlap Add into the output time domain stream.

Works pretty well but has a fairly long latency and a bit of modulation 
artifacting due to the FFT rate.

Eric

On 1/20/22 11:07, Mike Bryant wrote:
> Can be done various ways.
> 
> You could use a ShortTerm Fourier Transform, then simply reconstruct the signal by summing DFTs at different sampling rates for each frequency band.  This will have a few problems, such as discontinuities at the joins between the transforms which need filtering before feeding to the DFT bank.
> 
> Another way is to feed the signal to a wavelet filter bank to produce a Continuous Wavelet Transform, transform the mother transform reconstruction coefficients according to whatever pitch raise you want by frequency, then reform the audio signal from the new coefficients.
> 
> 
> 
> -----Original Message-----
> From: cheater cheater [mailto:cheater00social at gmail.com]
> Sent: 20 January 2022 17:35
> To: Mike Bryant
> Cc: David G Dixon; synth-diy at synth-diy.org
> Subject: Re: [sdiy] 1-quadrant multiplier with 2164
> 
> How would you do it?
> 
> On Thu, Jan 20, 2022 at 6:17 PM Mike Bryant <mbryant at futurehorizons.com> wrote:
>>
>> Yes :-)
>>
>> But infinitely easier if you go digital
>>
>> -----Original Message-----
>> From: Synth-diy [mailto:synth-diy-bounces at synth-diy.org] On Behalf Of
>> cheater cheater via Synth-diy
>> Sent: 20 January 2022 15:25
>> To: David G Dixon
>> Cc: synth-diy
>> Subject: Re: [sdiy] 1-quadrant multiplier with 2164
>>
>> I wonder if it's possible to build a frequency shifter that shifts higher harmonics more than lower harmonics.




More information about the Synth-diy mailing list