Any reason you can't just use the Speex resampler? It's implemented in
both fixed-point and float, has reasonable complexity (with a quality vs
CPU knob) and it's easy to rip out of the Speex source code. The
resampling ratio can be any rational number, so 7500 Hz is no problem.
Cheers,
Jean-Marc
On 08/02/2015 08:33 PM, glen english wrote:
> I'll do this. easy.
>
> Interpolation is usually cubic. Leave that up to me.
>
> There wouldnt be any implementation loss. just a few more MIPS.
>
> platform ?
>
> though, I would STRENUOUSLY suggest changing the modem symbol rate.
>
> 7500 Hz is .. well, a pain ....and will be a pain forever unless it is
> nipped in the bud now.
>
>
>
> On 3/08/2015 10:03 AM, David Rowe wrote:
>> Thanks Glen,
>>
>> We are stuck with the 7500Hz for the moment (an integer ratio of the
>> modem symbol rate, details in octave/cohpsk_frame_design.ods).
>>
>> Yes I had in mind something like your latter suggestion, up sample to 16
>> kHz, then linear interpolation. It would need to be tested in terms of
>> modem implementation loss.
>>
>> However I know what needs to be done, brainstorming that is the fun
>> part. What would be wonderful is someone stepping forward to carefully
>> code and test.
>>
>> Any volunteers? I'll work with you to help code, review, and test. You
>> need C coding skills with an interest in DSP, some time, and determination.
>>
>> Thanks,
>>
>> David
>>
>> On 03/08/15 09:09, glen english wrote:
>>> David
>>> for resampling...
>>>
>>> so you want to go from 8000 to 7500 Hz ?
>>>
>>> I've always picked my modems to be integer divisors of 48k . makes life
>>> easy.
>>> However, if we are stuck with 7500...
>>>
>>> 1) SRC using 120kHz as the common sample rate (polyphase resampler
>>> interpolator/decimator)
>>> cleanest, highest MIPS usage, easiest to understand.
>>>
>>> 2) Farrow interpolator.
>>> Best if passband is a small fraction of the sample rate.
>>> Most used for fractional sample rate conversion for mismatched clocks.
>>>
>>> A counter is used counting samples in and out, the fraction determined,
>>> and the resampling FIR coefficients are calculated on the fly.
>>>
>>> given a big chunk of the Nyquist BW is used, perhaps upsample to 16k ,
>>> convert to 15k,
>>> and then crash-downsample/undersample to 7500 (cheap).
>>>
>>>
>>> --
>>>
>>> I can do this.
>>> ---
>>>
>>>
>>>
>>> On 3/08/2015 9:20 AM, David Rowe wrote:
>>>> Further to this thread (started on the digitalvoice list) James has
>>>> provided a patch to implement his suggested changes to the FreeDV API,
>>>> which is now checked into codec2-dev.
>>>>
>>>> So wonderful for me to have some one step up and contribute some code.
>>>>
>>>> Thanks James!
>>>>
>>>> Now I need to go to work on the FreeDV GUI program to support these
>>>> changes.
>>>>
>>>> Down the track I'd like to see the FreeDV API accept a 8000 Hz sample
>>>> rate for the 700 mode (its currently 7500Hz). I have some ideas on how
>>>> to do this if anyone would like to work on the C code. But right now my
>>>> focus is on tuning the FreeDV 700 mode.
>>>>
>>>> On that subject I'm getting some great reports from people working
>>>> across the Atlantic on FreeDV 700 on as little as 5W - thanks Gerhard,
>>>> Walter, and Mel! The new modem is really working well on the HF fading
>>>> channel, and we have several examples of 80% DV copy when SSB was
>>>> unusable. I feel the speech codec needs some work, but we will get there.
>>>>
>>>> Exciting times for HF DV!
>>>>
>>>> - David
>>>>
>>>> ------------------------------------------------------------------------------
>>>> _______________________________________________
>>>> Freetel-codec2 mailing list
>>>> [email protected]
>>>> https://lists.sourceforge.net/lists/listinfo/freetel-codec2
>>>>
>> ------------------------------------------------------------------------------
>> _______________________________________________
>> Freetel-codec2 mailing list
>> [email protected]
>> https://lists.sourceforge.net/lists/listinfo/freetel-codec2
>>
>
------------------------------------------------------------------------------
_______________________________________________
Freetel-codec2 mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
