Re: [Freetel-codec2] Silly question regarding SM-1000 sample rate
On 20/09/15 03:43, Tomas Härdin wrote: > Couldn't you do further low-pass filtering in software, then decimate to > 8 kHz? (caveat: I haven't checked if the code actually does this) That's exactly what it does. It samples at 16kHz from the ADC, filters it to produce an 8kHz stream. Actually a thought just occurred, could we just do nearest-neighbour upsampling to get back to 16kHz? I'll admit I havent looked at what algorithm is used to do the resampling, however this is not a device that will be used by audiophiles (it doesn't do 192kHz 32-bit FLAC for starters) so linear interpolation or nearest-neighbour would probably work for getting it back to 16kHz. -- Stuart Longland (aka Redhatter, VK4MSL) I haven't lost my mind... ...it's backed up on a tape somewhere. -- ___ Freetel-codec2 mailing list Freetel-codec2@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/freetel-codec2
Re: [Freetel-codec2] Silly question regarding SM-1000 sample rate
I don't think a linear interpolator would work here. There's quite a bit of attenuation in the 0-Fs/4 band (~8dB by Fs/8) and only around -28dB on the aliases. -28dB might be enough for the aliases, but I don't think it's flat enough in the Fs/4 band to be useful. https://ccrma.stanford.edu/~jos/pasp/Linear_Interpolation_Frequency_Response.html On Sat, Sep 19, 2015 at 4:03 PM, Stuart Longlandwrote: > On 20/09/15 03:43, Tomas Härdin wrote: > > Couldn't you do further low-pass filtering in software, then decimate to > > 8 kHz? (caveat: I haven't checked if the code actually does this) > > That's exactly what it does. It samples at 16kHz from the ADC, filters > it to produce an 8kHz stream. > > Actually a thought just occurred, could we just do nearest-neighbour > upsampling to get back to 16kHz? > > I'll admit I havent looked at what algorithm is used to do the > resampling, however this is not a device that will be used by > audiophiles (it doesn't do 192kHz 32-bit FLAC for starters) so linear > interpolation or nearest-neighbour would probably work for getting it > back to 16kHz. > -- > Stuart Longland (aka Redhatter, VK4MSL) > > I haven't lost my mind... > ...it's backed up on a tape somewhere. > > > -- > ___ > Freetel-codec2 mailing list > Freetel-codec2@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/freetel-codec2 > -- ___ Freetel-codec2 mailing list Freetel-codec2@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/freetel-codec2
Re: [Freetel-codec2] Silly question regarding SM-1000 sample rate
Couldn't you do further low-pass filtering in software, then decimate to 8 kHz? (caveat: I haven't checked if the code actually does this) /Tomas On Fri, 2015-09-18 at 16:13 +1000, glen english wrote: > Hi Stuart > > An elaboration on David's reply. > > We want our audio BW to go to at least say 3.4 kHz . That is what > fixed line telephones offer. > > If we were to sample at 8kHz, our nyquist frequency is 4kHz of course. > Any spectral information above the nyquist rate will be aliased bay > into the baseband, IE below the nyquist rate. > > IF we assume we don't care about aliases between 3.4 and 4kHz, then > this allows for signals up to 4.0 - 3.4 = 0.6, and 0.6+Fyquist(4kHz) > = 4.6kHz. (as a 4.5kHz signal will get aliased to 3.5 kHz.) > > Assume we want our aliases 50dB down, so our aliasing filter must go > from 0dB down at 3.4kHz to 50 dB down at 4.6kHz > > Clearly alot of work in the analog department- probably needs an > elliptic filter. > a transition bandwidth factor of (4.6/3.4) = 1.35 > > Now, if the sample rate was 16kHz, now, Fnyquist is 8 kHz. again, we > dont care about aliases above 3.4kHz, > and Fnyquist, now 8kHz - 3.4kHz is 4.6kHz. > > So, for freq up to Fnyquist + 4.6kHz - IE 12.6kHz will not alias into > problem areas. > > so, now our analog filter spec if relaxed > 0dB down at 3.4kHz, and 50dB down at 12.6kHz > a transition bandwidth factor of (12.6/3.4) = 3.7 > > > > QED > > > > > On 17/09/2015 7:53 PM, Stuart Longland wrote: > > > Hi all, > > > > Just looking at the source code, it hit me. When in analogue mode, we > > run 16kHz sample rate, fair enough. > > > > But in DV mode, we sample the ADC at 16kHz, downconvert to 8kHz, pass it > > through the modem and codec, then upconvert back to 16kHz for the DAC. > > > > We're dealing with voice frequencies, with SSB transmit bandwidths of > > less than 3kHz. > > > > Why not do the whole lot at 8kHz and save some CPU time? Maybe some > > rate-switching when in analogue or DV mode might be an option too, so we > > run ADC/DAC at full-rate in analogue (for higher fidelity), then switch > > to half-rate when we're doing DV. > > > > Regards, > > > > > > -- > > Monitor Your Dynamic Infrastructure at Any Scale With Datadog! > > Get real-time metrics from all of your servers, apps and tools > > in one place. > > SourceForge users - Click here to start your Free Trial of Datadog now! > > http://pubads.g.doubleclick.net/gampad/clk?id=241902991=/4140 > > > > > > ___ > > Freetel-codec2 mailing list > > Freetel-codec2@lists.sourceforge.net > > https://lists.sourceforge.net/lists/listinfo/freetel-codec2 > > -- > - > Glen English > RF Communications and Electronics Engineer > > CORTEX RF > & > Pacific Media Technologies Pty Ltd > > ABN 40 075 532 008 > > PO Box 5231 Lyneham ACT 2602, Australia. > au mobile : +61 (0)418 975077 > > -- > ___ > Freetel-codec2 mailing list > Freetel-codec2@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/freetel-codec2 signature.asc Description: This is a digitally signed message part -- ___ Freetel-codec2 mailing list Freetel-codec2@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/freetel-codec2
Re: [Freetel-codec2] Silly question regarding SM-1000 sample rate
Hi Stuart An elaboration on David's reply. We want our audio BW to go to at least say 3.4 kHz . That is what fixed line telephones offer. If we were to sample at 8kHz, our nyquist frequency is 4kHz of course. Any spectral information above the nyquist rate will be aliased bay into the baseband, IE below the nyquist rate. IF we assume we don't care about aliases between 3.4 and 4kHz, then this allows for signals up to 4.0 - 3.4 = 0.6, and 0.6+Fyquist(4kHz) = 4.6kHz. (as a 4.5kHz signal will get aliased to 3.5 kHz.) Assume we want our aliases 50dB down, so our aliasing filter must go from 0dB down at 3.4kHz to 50 dB down at 4.6kHz Clearly alot of work in the analog department- probably needs an elliptic filter. a transition bandwidth factor of (4.6/3.4) = 1.35 Now, if the sample rate was 16kHz, now, Fnyquist is 8 kHz. again, we dont care about aliases above 3.4kHz, and Fnyquist, now 8kHz - 3.4kHz is 4.6kHz. So, for freq up to Fnyquist + 4.6kHz - IE 12.6kHz will not alias into problem areas. so, now our analog filter spec if relaxed 0dB down at 3.4kHz, and 50dB down at 12.6kHz a transition bandwidth factor of (12.6/3.4) = 3.7 QED On 17/09/2015 7:53 PM, Stuart Longland wrote: Hi all, Just looking at the source code, it hit me. When in analogue mode, we run 16kHz sample rate, fair enough. But in DV mode, we sample the ADC at 16kHz, downconvert to 8kHz, pass it through the modem and codec, then upconvert back to 16kHz for the DAC. We're dealing with voice frequencies, with SSB transmit bandwidths of less than 3kHz. Why not do the whole lot at 8kHz and save some CPU time? Maybe some rate-switching when in analogue or DV mode might be an option too, so we run ADC/DAC at full-rate in analogue (for higher fidelity), then switch to half-rate when we're doing DV. Regards, -- Monitor Your Dynamic Infrastructure at Any Scale With Datadog! Get real-time metrics from all of your servers, apps and tools in one place. SourceForge users - Click here to start your Free Trial of Datadog now! http://pubads.g.doubleclick.net/gampad/clk?id=241902991=/4140 ___ Freetel-codec2 mailing list Freetel-codec2@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/freetel-codec2 -- - Glen English RF Communications and Electronics Engineer CORTEX RF & Pacific Media Technologies Pty Ltd ABN 40 075 532 008 PO Box 5231 Lyneham ACT 2602, Australia. au mobile : +61 (0)418 975077 -- ___ Freetel-codec2 mailing list Freetel-codec2@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/freetel-codec2
Re: [Freetel-codec2] Silly question regarding SM-1000 sample rate
Hi David & Glen, On 18/09/15 16:13, glen english wrote: > If we were to sample at 8kHz, our nyquist frequency is 4kHz of course. > Any spectral information above the nyquist rate will be aliased bay into > the baseband, IE below the nyquist rate. > > IF we assume we don't care about aliases between 3.4 and 4kHz, then this > allows for signals up to 4.0 - 3.4 = 0.6, and 0.6+Fyquist(4kHz) = > 4.6kHz. (as a 4.5kHz signal will get aliased to 3.5 kHz.) Fair enough. I just wondered, seemed that we could save some CPU processing power if we used a lower rate, but good point on the analogue filtering, I wasn't thinking of that side of things. Definitely we want to avoid nasty aliasing, and having a higher rate does make the filter cutoff less critical. -- Stuart Longland (aka Redhatter, VK4MSL) I haven't lost my mind... ...it's backed up on a tape somewhere. signature.asc Description: OpenPGP digital signature -- ___ Freetel-codec2 mailing list Freetel-codec2@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/freetel-codec2
Re: [Freetel-codec2] Silly question regarding SM-1000 sample rate
Also there is very little speech energy out past 12 kHz, i.e. its already 50dB down, and this is communications quality speech so a little aliasing is lost in the codec artefacts. On the radios interface side the audio is band limited by the radio's frequency response, ie the xtal filter BW. - David On 18/09/15 15:43, glen english wrote: > Hi Stuart > > An elaboration on David's reply. > > We want our audio BW to go to at least say 3.4 kHz . That is what fixed > line telephones offer. > > If we were to sample at 8kHz, our nyquist frequency is 4kHz of course. > Any spectral information above the nyquist rate will be aliased bay into > the baseband, IE below the nyquist rate. > > IF we assume we don't care about aliases between 3.4 and 4kHz, then this > allows for signals up to 4.0 - 3.4 = 0.6, and 0.6+Fyquist(4kHz) = > 4.6kHz. (as a 4.5kHz signal will get aliased to 3.5 kHz.) > > Assume we want our aliases 50dB down, so our aliasing filter must go > from 0dB down at 3.4kHz to 50 dB down at 4.6kHz > > Clearly alot of work in the analog department- probably needs an > elliptic filter. > a transition bandwidth factor of (4.6/3.4) = 1.35 > > Now, if the sample rate was 16kHz, now, Fnyquist is 8 kHz. again, we > dont care about aliases above 3.4kHz, > and Fnyquist, now 8kHz - 3.4kHz is 4.6kHz. > > So, for freq up to Fnyquist + 4.6kHz - IE 12.6kHz will not alias into > problem areas. > > so, now our analog filter spec if relaxed > 0dB down at 3.4kHz, and 50dB down at 12.6kHz > a transition bandwidth factor of (12.6/3.4) = 3.7 > > > > QED > > > > > On 17/09/2015 7:53 PM, Stuart Longland wrote: >> Hi all, >> >> Just looking at the source code, it hit me. When in analogue mode, we >> run 16kHz sample rate, fair enough. >> >> But in DV mode, we sample the ADC at 16kHz, downconvert to 8kHz, pass it >> through the modem and codec, then upconvert back to 16kHz for the DAC. >> >> We're dealing with voice frequencies, with SSB transmit bandwidths of >> less than 3kHz. >> >> Why not do the whole lot at 8kHz and save some CPU time? Maybe some >> rate-switching when in analogue or DV mode might be an option too, so we >> run ADC/DAC at full-rate in analogue (for higher fidelity), then switch >> to half-rate when we're doing DV. >> >> Regards, >> >> >> -- >> Monitor Your Dynamic Infrastructure at Any Scale With Datadog! >> Get real-time metrics from all of your servers, apps and tools >> in one place. >> SourceForge users - Click here to start your Free Trial of Datadog now! >> http://pubads.g.doubleclick.net/gampad/clk?id=241902991=/4140 >> >> >> ___ >> Freetel-codec2 mailing list >> Freetel-codec2@lists.sourceforge.net >> https://lists.sourceforge.net/lists/listinfo/freetel-codec2 > > -- > - > Glen English > RF Communications and Electronics Engineer > > CORTEX RF > & > Pacific Media Technologies Pty Ltd > > ABN 40 075 532 008 > > PO Box 5231 Lyneham ACT 2602, Australia. > au mobile : +61 (0)418 975077 > > > > -- > > > > ___ > Freetel-codec2 mailing list > Freetel-codec2@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/freetel-codec2 > -- ___ Freetel-codec2 mailing list Freetel-codec2@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/freetel-codec2
Re: [Freetel-codec2] Silly question regarding SM-1000 sample rate
yeah you can pretty much get away with NO aliasing filter considering the source ( a band limited microphone) and the required voiceband SN. still, good reason to have at least a single RC- from say switchmode power supply whistles etc getting back into the input, and aliasing down into the baseband (from say 50-300kHz) , or even the transmitter. Yes-on 160m, I have had the RF aliasing back into the passband from the 32kHz sampled mic audio. g On 18/09/2015 7:35 PM, David Rowe wrote: > Also there is very little speech energy out past 12 kHz, i.e. its > already 50dB down, and this is communications quality speech so a little > aliasing is lost in the codec artefacts. > > On the radios interface side the audio is band limited by the radio's > frequency response, ie the xtal filter BW. > > - David > > On 18/09/15 15:43, glen english wrote: >> Hi Stuart >> >> An elaboration on David's reply. >> >> We want our audio BW to go to at least say 3.4 kHz . That is what fixed >> line telephones offer. >> >> If we were to sample at 8kHz, our nyquist frequency is 4kHz of course. >> Any spectral information above the nyquist rate will be aliased bay into >> the baseband, IE below the nyquist rate. >> >> IF we assume we don't care about aliases between 3.4 and 4kHz, then this >> allows for signals up to 4.0 - 3.4 = 0.6, and 0.6+Fyquist(4kHz) = >> 4.6kHz. (as a 4.5kHz signal will get aliased to 3.5 kHz.) >> >> Assume we want our aliases 50dB down, so our aliasing filter must go >> from 0dB down at 3.4kHz to 50 dB down at 4.6kHz >> >> Clearly alot of work in the analog department- probably needs an >> elliptic filter. >> a transition bandwidth factor of (4.6/3.4) = 1.35 >> >> Now, if the sample rate was 16kHz, now, Fnyquist is 8 kHz. again, we >> dont care about aliases above 3.4kHz, >> and Fnyquist, now 8kHz - 3.4kHz is 4.6kHz. >> >> So, for freq up to Fnyquist + 4.6kHz - IE 12.6kHz will not alias into >> problem areas. >> >> so, now our analog filter spec if relaxed >> 0dB down at 3.4kHz, and 50dB down at 12.6kHz >> a transition bandwidth factor of (12.6/3.4) = 3.7 >> >> >> >> QED >> >> >> >> >> On 17/09/2015 7:53 PM, Stuart Longland wrote: >>> Hi all, >>> >>> Just looking at the source code, it hit me. When in analogue mode, we >>> run 16kHz sample rate, fair enough. >>> >>> But in DV mode, we sample the ADC at 16kHz, downconvert to 8kHz, pass it >>> through the modem and codec, then upconvert back to 16kHz for the DAC. >>> >>> We're dealing with voice frequencies, with SSB transmit bandwidths of >>> less than 3kHz. >>> >>> Why not do the whole lot at 8kHz and save some CPU time? Maybe some >>> rate-switching when in analogue or DV mode might be an option too, so we >>> run ADC/DAC at full-rate in analogue (for higher fidelity), then switch >>> to half-rate when we're doing DV. >>> >>> Regards, >>> >>> >>> -- >>> Monitor Your Dynamic Infrastructure at Any Scale With Datadog! >>> Get real-time metrics from all of your servers, apps and tools >>> in one place. >>> SourceForge users - Click here to start your Free Trial of Datadog now! >>> http://pubads.g.doubleclick.net/gampad/clk?id=241902991=/4140 >>> >>> >>> ___ >>> Freetel-codec2 mailing list >>> Freetel-codec2@lists.sourceforge.net >>> https://lists.sourceforge.net/lists/listinfo/freetel-codec2 >> -- >> - >> Glen English >> RF Communications and Electronics Engineer >> >> CORTEX RF >> & >> Pacific Media Technologies Pty Ltd >> >> ABN 40 075 532 008 >> >> PO Box 5231 Lyneham ACT 2602, Australia. >> au mobile : +61 (0)418 975077 >> >> >> >> -- >> >> >> >> ___ >> Freetel-codec2 mailing list >> Freetel-codec2@lists.sourceforge.net >> https://lists.sourceforge.net/lists/listinfo/freetel-codec2 >> > -- > ___ > Freetel-codec2 mailing list > Freetel-codec2@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/freetel-codec2 > -- - Glen English RF Communications and Electronics Engineer CORTEX RF & Pacific Media Technologies Pty Ltd ABN 40 075 532 008 PO Box 5231 Lyneham ACT 2602, Australia. au mobile : +61 (0)418 975077 -- ___ Freetel-codec2 mailing list Freetel-codec2@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/freetel-codec2
Re: [Freetel-codec2] Silly question regarding SM-1000 sample rate
Hi Stuart, Running the ADC/DAC at 16 kHz largely removes the need for analog anti-aliasing/reconstruction filters, significantly simplifying the hardware. - David On 17/09/15 19:23, Stuart Longland wrote: > Hi all, > > Just looking at the source code, it hit me. When in analogue mode, we > run 16kHz sample rate, fair enough. > > But in DV mode, we sample the ADC at 16kHz, downconvert to 8kHz, pass it > through the modem and codec, then upconvert back to 16kHz for the DAC. > > We're dealing with voice frequencies, with SSB transmit bandwidths of > less than 3kHz. > > Why not do the whole lot at 8kHz and save some CPU time? Maybe some > rate-switching when in analogue or DV mode might be an option too, so we > run ADC/DAC at full-rate in analogue (for higher fidelity), then switch > to half-rate when we're doing DV. > > Regards, > > > > -- > Monitor Your Dynamic Infrastructure at Any Scale With Datadog! > Get real-time metrics from all of your servers, apps and tools > in one place. > SourceForge users - Click here to start your Free Trial of Datadog now! > http://pubads.g.doubleclick.net/gampad/clk?id=241902991=/4140 > > > > ___ > Freetel-codec2 mailing list > Freetel-codec2@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/freetel-codec2 > -- Monitor Your Dynamic Infrastructure at Any Scale With Datadog! Get real-time metrics from all of your servers, apps and tools in one place. SourceForge users - Click here to start your Free Trial of Datadog now! http://pubads.g.doubleclick.net/gampad/clk?id=241902991=/4140 ___ Freetel-codec2 mailing list Freetel-codec2@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/freetel-codec2