Much lower. You have 96dB of S/N, which is the noise floor integrated over the 20kHz bandwidth.
I did a little research and it turns out that FM S/N ratio and distortion is related to the modulation index. For broadcast FM, this modulation ratio is 3.75 which gives a +/-75Khz frequency variation for a 20khz signal. This limitation is imposed by statute to allow up to 100 adjacent FM stations with minimal interference across the 88-108 Mhz broadcast band.
The actual ratio is SN/distortion ratio = 3*D(squared), where D is the modulation ratio.
So, if we assume that OTA FM can give 75 db S/N and 0.2% distortion under fairly optimal conditions, increasing the modulation ratio to 50 (+/- 1Mhz deviation for a 20khz signal, which should be readily achievable in a box player) would give 119db and 0.0011% distortion.
And, unlike PCM digital, FM's distortion level is not some unachievable minimum at 100% modulation level - it is essentially constant or lowered across the whole dynamic range. I suspect this is one reason that so many modern recordings have a single digit dynamic range - musicians and audio professionals are deathly tired of digital quantization crap. This characterisitic of OTA FM, btw, is why it sounds better than 16 bit PCM, even though it suffers somewhat in ultimate dynamic range by comparison.
The actual ratio is SN/distortion ratio = 3*D(squared), where D is the modulation ratio.
So, if we assume that OTA FM can give 75 db S/N and 0.2% distortion under fairly optimal conditions, increasing the modulation ratio to 50 (+/- 1Mhz deviation for a 20khz signal, which should be readily achievable in a box player) would give 119db and 0.0011% distortion.
And, unlike PCM digital, FM's distortion level is not some unachievable minimum at 100% modulation level - it is essentially constant or lowered across the whole dynamic range. I suspect this is one reason that so many modern recordings have a single digit dynamic range - musicians and audio professionals are deathly tired of digital quantization crap. This characterisitic of OTA FM, btw, is why it sounds better than 16 bit PCM, even though it suffers somewhat in ultimate dynamic range by comparison.
Last edited:
The snag is that FM gets harder to generate and receive with low distortion as the modulation gets wider - unless you use digital techniques! There is also the problem of needing accurately phase-linear filters, as filter dispersion creates distortion. Real life is always more complicated than people think. More research?
Base band response for FM stereo is band limited to 50Hz - 15kHz (45%) giving a theoretical modulation index of 5.. There is a pilot tone at 19kHz (10%) and the stereo subcarrier from 23kHz - 53kHz for the other 45%, other services like RBDS and SCA account for several % which results in slight over-modulation to about 102.7% of full modulation or 77kHz.. The base band and subcarrier amplitude may be slightly reduced to maintain 75kHz peak deviation.
http://www.silabs.com/Marcom Documents/Resources/FMTutorial.pdf
Something to bear in mind is modulation index is not a constant, it varies inversely with modulation frequency and can be quite low for subcarrier frequencies. (Perhaps more correctly I should have cited modulation ratio rather than index for a dynamically varying signal in the frequency domain.)
One could have two wideband FM modulators running at different frequencies which would avoid the need for multiplex and eliminate the SNR and separation problems inherent in stereo multiplex operation.
Most broadcast FM IMLE does not even approach redbook levels of performance due to various compression schemes in current use and limiting in the modulator signal chain, some public broadcasters do provide pretty high quality signals during live broadcasts. Most compressor limiter chains are set up to provide the highest possible average signal level in order to make the station stand out from the competition level wise.
http://www.bext.com/histproc.htm
http://www.silabs.com/Marcom Documents/Resources/FMTutorial.pdf
Something to bear in mind is modulation index is not a constant, it varies inversely with modulation frequency and can be quite low for subcarrier frequencies. (Perhaps more correctly I should have cited modulation ratio rather than index for a dynamically varying signal in the frequency domain.)
One could have two wideband FM modulators running at different frequencies which would avoid the need for multiplex and eliminate the SNR and separation problems inherent in stereo multiplex operation.
Most broadcast FM IMLE does not even approach redbook levels of performance due to various compression schemes in current use and limiting in the modulator signal chain, some public broadcasters do provide pretty high quality signals during live broadcasts. Most compressor limiter chains are set up to provide the highest possible average signal level in order to make the station stand out from the competition level wise.
http://www.bext.com/histproc.htm
Of course, I was talking about the original mono FM - stereo FM with its subcarriers complicates the argument I was trying to make
OTOH, OTA FM standards were readily approachable with the old Dynaco FM3 budget tuner from the late '50'
s, so I wouldn't be too worried about doing significantly better than that, even with an all analog approach.
Btw, I am in no way advocating going back to LD video! I think high definition and SHD digital video is the way to go - digital can do notably better than analog if your total dynamic range is less than 50 db, and I don't think anybody has found a marketable way to compress digital video to 1db dynamic range yet
OTOH, OTA FM standards were readily approachable with the old Dynaco FM3 budget tuner from the late '50'
s, so I wouldn't be too worried about doing significantly better than that, even with an all analog approach.
Btw, I am in no way advocating going back to LD video! I think high definition and SHD digital video is the way to go - digital can do notably better than analog if your total dynamic range is less than 50 db, and I don't think anybody has found a marketable way to compress digital video to 1db dynamic range yet
Last edited:
The best sounding recordings I have of 16 bit 44.1kHz PCM generally limit their peak amplitude to something less than -10dBFS, and those recordings have fairly wide dynamic ranges and sound pretty good actually. My flawed understanding is the limited dynamic range of many modern recordings is due to a desire to maximize the average level for oddly enough FM broadcast and conversion to MP3 for portable players - the so called loudness wars. I have a fair amount of wide dynamic range classical music on CD and while they aren't the ultimate in sound quality in most respects they aren't generally terrible either.
I am however listening to vinyl as I write this...
Sy didn't mention the advantage of psychoacoustic weighted noise shaped dither - available on CD for a least a decade - linearity, perceived audio noise are way better than the 16 bit*6.02dB calc today
and practically delivered by now ubiquitous "24-bit" audio DAC with 20+ bit dynamic performance in leading parts
and practically delivered by now ubiquitous "24-bit" audio DAC with 20+ bit dynamic performance in leading parts
I'd love to see a format that would readily surpass LP in all regards, not just in some, and also be recordable and as resistant to noise as digital.
Is that too much to ask
I've been critical of LP sound in the past (like whining why not more 45RPM LPs), but as far as transferring music, it is among the best that has been offered and the only one still somewhat available to consumers.
Is that too much to ask
I've been critical of LP sound in the past (like whining why not more 45RPM LPs), but as far as transferring music, it is among the best that has been offered and the only one still somewhat available to consumers.
To achieve the maximum 'perceived' 'increase' which is still usually less than 10db, noise floors above 3khz have to be increased quite a bit - sometimes becoming obtrusively perceptible.
I was a fan of high definition digital audio - often describing its sound as 'analog like', but where is it now, except on Blu Ray?
Mostly, people are still at the bottom of the acoustic barrel with MP3 which sounds best in a high noise environment like a subway.
There is one- digital. You may not like what recording engineers are doing with the tool (I sure don't!), but the tool surpasses LP in every respect. Once I compared the output of my D/A with the direct mike feed, something I did regularly with tape, I never looked back.
There's just no reason to do development work on analog recording/playback except for non-technical motivations, e.g., playback of old recordings only available on vinyl or tape, nostalgia, or fashion. Not very compelling for serious researchers.
The end of open reel tape came abruptly here when I did some A/B comparisons of recordings from vinyl to tape at 7 1/2 ips and vinyl to 2496 digital captured by my modest but surprisingly decent sounding antique M Audio Audiophile 2496 sound card.. I don't have the kind of money required for really good tape, and the ReVox G36 MKIII with new half track heads was not pulling its weight unfortunately. If I still had my 15/30ips Otari I might have felt differently, but I can't afford Tape Project recordings so I figured what's the point. The Otari had a bridge on it, was huge and took up a lot of space, was not the best nor worst sounding half track machine I have heard. I cannot remember the model for the life of me.
Bob Stuart gave 18 dB improvement as practical in 2004, the dither "color" only perceptible when you turn up the gain to levels that would deeply clip properly mastered music peaks http://www.ece.rochester.edu/course...ntries/2009/1/15_Week_1_files/Stuart_2004.pdf
today some have built psychoacoustic masking noise shapers that dynamically change the noise shaping filter/dither spectrum to hide in the psychoacoustic masked frequency bands - the S/N numbers for the perceived noise floor then become meaninglessly large - and admittedly dependent on masking model accuracy
Stuart also likes showing the numbers for dither applied to much higher sample rates - 8/192 anyone?
today some have built psychoacoustic masking noise shapers that dynamically change the noise shaping filter/dither spectrum to hide in the psychoacoustic masked frequency bands - the S/N numbers for the perceived noise floor then become meaninglessly large - and admittedly dependent on masking model accuracy
Stuart also likes showing the numbers for dither applied to much higher sample rates - 8/192 anyone?
Last edited:
I have never understood where this idea comes from, a correctly dithered quantiser is LINEAR all the way down, there are no steps in the reconstructed output or anything daft like that, it simply has an uncorrelated noise floor (Just the same as any analogue circuit does).
Now an undithered (Broken) quantiser has all sorts of issues of course, so don't do that, but I have never figured out where this idea that a normally functioning audio ADC/DAC chain has distortion that inherently increases at lower input level comes from (Undithered quantisers are NOT used for audio by essentially anyone these days (There were a few around in the early days, possibly persisting as late as the mid 80's)).
I blame daft books with silly diagrams showing stair steps overlaid on a sine wave and hard of thinking lecturers. :headbang:
On the subject of FM, building a really good generator for ultrawideband FM is **HARD** (The best of the broadcast stuff these days use a DDS chip and do the work digitally, outputting RF directly from a numerically controlled oscilator), older rigs use a phase locked loop (Digital divider to set the TX frequency and a varicap diode to do both the tuning and the modulation, not really **THAT** linear, but it works well enough for the purpose).
FM discriminators are another area where it is hard to build a really linear one, pulse counting is about as good as it gets, and of course the output of the limiting amplifier is essentially digital anyway.
It is interesting to note that the broadcasters are finally starting to do something about the 'everything louder then everything else' phenomoma with the R.128 loudness metering becoming ever more commonplace in the (Particularly TV) production technical standards.
We might finally start to see the music production croud lay off with the massive compression (Which, actually radio stations are quite capable of applying themselves (and will anyway) thank you very much), the stations transmit chain processing is well able to spindle and mutilate audio with no help from the original audio producers.
Regards, Dan (Who used to do FM transmitters for a living).
Or maybe not. I was using a two track 15ips Ampex 351 with rebuilt electronics. A great machine. The little outboard 24/96 converter I use, which is not much bigger than the Ampex's input transformers, is leagues better.
but today's car audio system's electronics could do a fine job too - and no need for anyone to broadcast mangled source
of course the future of broadcast is now digital psychoacoustic lossy compression - often to voice quality so you can squeeze more talk channels into your spectrum allotment
Last edited:
err no - you really don't grasp dither theory if you say that
you do need "all the bits" of accuracy/linearity in the playback device to see the dithered signal's linearity - why I mentioned the 20 bit performance of todays better audio DACs - but the 16 bit properly dithered source doesn't have correlated quantization noise - you can easily see/hear sine/music with fractional 16 lsb amplitude
you do need "all the bits" of accuracy/linearity in the playback device to see the dithered signal's linearity - why I mentioned the 20 bit performance of todays better audio DACs - but the 16 bit properly dithered source doesn't have correlated quantization noise - you can easily see/hear sine/music with fractional 16 lsb amplitude
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.