|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
67.5.154.157
In Reply to: RE: Here's A More Direct Source posted by JURB on December 10, 2016 at 09:49:58
Actually, plain old analog FM stereo actually is based on sampled data -- the way it works (in the time domain) is the signal does a bit of Left channel, followed by a bit of Right channel, then Left again, then Right again, swapping between them 38,000 times a second. Then a 19kHz sinewave tone is played right on top of it, which tells the stereo decoder when to shift the signal toward the left or toward the right to separate the channels again. Hence, the sharp cutoff it has at 15kHz, to keep the 19kHz tone out of the speakers and to avoid images from the sampled signal. I know this because my High School science project was to make an FM stereo encoder (which managed to get used on a pirate station or two!).And, yes, that's ANALOG FM, as started back in the 1960's! Digital FM is of course even more convoluted.
_
Make super easy diffusors:--> http://www.diyaudio.com/forums/everything-else/269366-making-easy-diy-depot-sound-diffuser-panels-step-step.html#post4215464
Horn Design Spreadsheet:--> http://libinst.com/SynergyCalc/
Edits: 12/10/16Follow Ups:
There is no sampling in FM radio.
When broadcasting a stereo (L and R) signal, first the signals are band limited to 15 KHz. Then the modulator forms sum (L+R) and difference (L-R) signals. The L+R signal occupies the lower 15 KHz of the channel. Above that is the pilot tone at 19 KHz. The pilot tone is used to generate a 38 KHz subcarrier, and the L-R signal is amplitude modulated onto the subcarrier, which is suppressed for transmission. No sampling, just AM.
0-15 KHz - Sum (L+R) signal, baseband
19 KHz - pilot tone
23-53 KHz - Difference (L-R) signal, amplitude modulated on 38 KHz carrier
Most FM stations add other services in the remaining channel space between 53 KHz and 100 KHz such as radio for the blind, muzak, RDBS. And then the whole thing is frequency modulated onto the channel's carrier.
This approach allowed for simple & cheap monaural FM radio reception. All you have to do is the one FM demodulation and then filter out anything above 15 KHz.
For stereo FM, you also have to lock on the 19 KHz pilot tone, use it to generate the 38 KHz subcarrier (since the subcarrier was suppressed), demodulate the AM signal, and do another sum & difference to get the left and right signals back.
"There is no sampling in FM radio."
Technically debatable. I know how it works. I guess what people say is that if it isn't quantized it isn't sampled. I call that semantics because even the analog stereoplexers chopped up the signal into pieces 1/38,000 th of a second long.
The technique of the modulation BTW used to be called double sideband suppressed carrier.
It still has its limitations. With more modern techniques such as pilot cancelling rather than a notch filter they could do 17 KHz, maybe a tad more, but it simply is not transmitted.
And with all the digital junk they transmit with it they probably simply can't. In fact some older tuners have a problem with all that digital crap and it comes out as noise even on the strongest of signals with more than ample limiting. There are people out there who modify older FM tuners to fix that. I believe, IIRC, the filter goes in the IF strip. I'm thinking but not sure that if you have a wide/narrow IF bandwidth switch, using narrow should eliminate the noise but then you get the increased distortion. Like, take your pick.
In the old days the other subcarrier was SCA, for store cast allocation and FM tuners started having filters for that. It ran at 67 KHz and I think it was frequency modulated. But not by much, the quality was not all that great because it was used for things like Muzak and maybe voice.
Now of course they got more in there. Your car radio reads out the name of the song that is playing, that is something new. And of course that is a source of noise for older tuners. And to tell you what, the newer tuners I have heard do not sound as good. But then many of them are on a circuit board about the size of a USB thumbdrive. theyy run on a new process, I forgot what it is called but somehow one chip does iit all. there is no IF strip, some don't even have anything that looks like a detector alignment. And the same 18 pin chip also does AM.
Now compare that to an old Revox.
I maintain my position that in most cases FM is good for the car, but not for serious listening. If you get a station without all the digital garbage on it maybe, but where do you find that ? And still, the 15 KHz limitation is still there. Most of you wouldn't put up with that in a cassette deck. At least not for hifi use.
" Then the modulator forms sum (L+R) and difference (L-R) signals."
A comparator did that. In a digital stereoplexer it simply switches between the channels. Much cheaper, no phase problems generating the pilot because a simple flip flop (astable mulitvibrator) does that. The system did make it a bit simpler, but better ? I doubt it. Just like the old redbook CDs speced out better than vinyl, sometimes the vinyl just sounds better.
At first glance it seems like a semantic argument, but to me there is a key difference between square wave modulation and sampling. The former is continuous time process and the latter is discrete time. I suppose I've become a bit sensitive over this because so many people keep calling Class D "digital".
One reason why it's important is that discrete time information theory principles such as the Nyquist limit and Shannon-Hartley law apply to sampling but not square wave modulation. For example, you can theoretically use a switching modulator to amplitude modulate a carrier whose frequency is less than twice the maximum frequency of the input signal. Supposing the input signal bandwidth is 15 KHz, you can produce an AM signal whose carrier is 10 KHz. The bottom 5 KHz of the lower sideband will wrap onto itself, so the lower sideband needs to be suppressed, but it is possible to demodulate the AM signal and fully reconstruct the original input signal. If the input signal was sampled instead at 10 KHz, this would not be possible. Per Nyquist, the sampling frequency would have to be > = 30 KHz.
Actually this talk of FM is getting out of the domain of thread drift and into a hijack, which I don't like. the similarity is there though, that like 128K MP3s are good for the car or portable use, while the higher bitrate ones should be used on your good system. And I consider anything less than 128 unlistenable.
But how much of this is attributable to dynamic range compression rather than the data compression ? In analog recording, say on tape, you could go beyond 0 dB. It would create some distortion. On tape there would be some third harmonic, on vinyl it risks fast groove wear or even overcutting, which makes the master useless.
But with digital it has to be hard limited because you cannot exceed a certain, absolutely defined number. It has to be clipped like an overdriven amp. If not, in most cases it will make a terrible noise if over"modulated".
Dave_K, you're not wrong, and that's the FREQUENCY DOMAIN explanation. In the time domain (where the actual electronic process is, or at least was, done in practice) it is as I described. Filter out everything above 15kHz, you're left with L+R, no matter what domain you describe it in.
_
Make super easy diffusors:--> http://www.diyaudio.com/forums/everything-else/269366-making-easy-diy-depot-sound-diffuser-panels-step-step.html#post4215464
Horn Design Spreadsheet:--> http://libinst.com/SynergyCalc/
"Dave_K, you're not wrong, and that's the FREQUENCY DOMAIN explanation. In the time domain (where the actual electronic process is, or at least was, done in practice) it is as I described. Filter out everything above 15kHz, you're left with L+R, no matter what domain you describe it in."
Once you put the FM signal in the discriminator, the decoding of the L+R and the L-R happens simultaneously (Also the recovery of the RDS info). It doesn't switch back and forth in a standard analog demodulation the way you describe.
====
"We have met the enemy and he is us" - Pogo
Sorry, it does. I've designed and built such encoders long before whatever internet article you maybe read existed! It is done with an analog multiplexer (switch), a 38kHz square wave source, and a div/2 ff and some phase shift networks to get the synchronized 19kHz pilot tone. If you can find a block diagram of the old National Semiconductor PLL FM stereo decoder chips you'll see the decoding done in the (reversed) same way.
_
Make super easy diffusors:--> http://www.diyaudio.com/forums/everything-else/269366-making-easy-diy-depot-sound-diffuser-panels-step-step.html#post4215464
Horn Design Spreadsheet:--> http://libinst.com/SynergyCalc/
Be careful.I have worked in RF and Microwave in TV, Radio, and Wireless for over 20 (Almost 30!) years. Mostly on infrastructure and transmitters.
I just posted up the block diagram for clarity since I did not know your background.
Peace. I'll assume you are correct as I have not got into the guts of chips, and have mostly been on the transmission side of it.
====
"We have met the enemy and he is us" - Pogo
Edits: 12/12/16
It reminds me a little of my early days working on PA's for TV transmitters - the modulation for color TV was a very complicated one with AM, FM and all of that mixed up at once.
I got into it when they were working on the HDTV standards, and back when they were thinking of doing it analog. Quickly they started using QAM signals in Europe (where I was working at the time) when the US went their own way with 8VSB (developed by Zenith before being consumed by, I think, LG or Samsung - can't remember).
Good times.
Glad to see an old RF guy like myself here. Too often people do "engineering by google" so you never know where the other fellow is standing.
Didn't know about the commutating time domain sampling method. SOunds like a clever way to re-use circuity or save some costs. Wouldn't occur to me to do it that way. But as a high power RF PA guy, I am all about KISS - and reliability. Since when you are making a 30kW RF system, you really don't want anything to fail!
====
"We have met the enemy and he is us" - Pogo
"the modulation for color TV was a very complicated one with AM, FM and all of that mixed up at once."
Not exactly, unless you are including the sound carrier.
The chroma signal is like the FM L-R channel in that it is carrier suppressed. Also, the detectors in an FM tuner for L-R are actually synchronous detectors. The detectors in the NTSC system had two synchronous detectors, one running 90 degrees out of phase thus yielding two discrete signals.
As with FM, it is not that different on the transmitting end. But it is a bit more complicated than that. The I signal had a bit more bandwidth but only a few TVs could use it. The RCA CTC111 I believe had what was called "wide I" demodulation but then they used a crap CRT. It was the early days of the inline gun CRTs and the picture would have been much better on a delta gun CRT. The pitch of those early inline gun jobs was so poor they shouldn't have bothered. Wide I demodulation also required another delay line. That chassis also had one of the earliest digital COMB filters on the market.
The I and Q signals were the difference signals for red and blue, I forget which is which. The green was derived from a matrix circuit, when the red and/or blue went down, it made the green go up and vice versa.
There is some complex math involved as they found that the green carried the most detail so it was fairly predominant in the monochrome signal, which underlies those difference signals.
But there was no FM in there, that was phase modulation and was not even intentional because when you mix two waveforms that are 90 degrees out of phase you get different phases which range from the instantaneous to the quadrature.
More useless knowledge, I have tons of it.
Getting back to the OP here, these ultra good digital formats will happen when not only the media is invented, but when the sources are available. I listen to some music from the 1950s and find a 128K MP3 to be quite adequate, though even on that material I have sometimes noticed a difference with a higher bitrate file. When I downloaded I would get several copies of everything and delete the inferior ones. It is actually somewhat surprising how good some of those old recordings are, considering the times. Others not so good.
There is now the holographic disk. It makes blu-ray obsolete. It has so much capacity nobody can use it. Perhaps that is the medium of choice. The ultra high quality formats would take downloading back to the 20th century, like it was on dialup. Imagine a five minute song being 800 MB.
Here's a wiki on those disks :
https://en.wikipedia.org/wiki/Holographic_Versatile_Disc
If you wanted to send someone a song, depending on your internet speed you might be better off just burning one of those and snail mailing it. However the technology has not been perfected. They've apparently done it but it is not quite market ready I guess. But they're talking capacities up in the terabytes.
They get that ready, THEN you can throw away the vinyl.
A bit of background info...
One method of implementing amplitude modulation is to multiply the input signal by a square wave at the carrier frequency, which can be implemented using diodes as switches. Because a square wave is just the superposition of an infinite series of cosines at the fundamental frequency of the square wave and its harmonics, the result of multiplying the square wave by the input signal is an infinite series of amplitude modulated signals. The fundamental and all of its harmonics are amplitude modulated by the input signal. Then you use a bandpass filter to select the one that matches the desired carrier frequency, usually the fundamental or third harmonic. This "switching" method of amplitude modulation is cheap and easy to implement in low level solid state circuits and is more linear than a square law modulator. So it's well suited for implementation on a chip and thus common.
When the FCC was soliciting proposals for FM stereo, two of the companies (sorry, I don't remember which) proposed transmitting sum (L+R) and difference (L-R) signals, with the sum transmitted in the 15 KHz baseband just like FM mono and the difference signal AM modulated into the channel space above that. If I remember correctly, one of the proposed implementations involved explicitly forming the difference signal and then using amplitude modulation, and the other proposal involved using a square wave to switch between the L signal and an inverted R signal. These two approaches are mathematically equivalent.
By the way, my background is originally RF as well (and computational electromagnetics). But I'm far from that now.
Not to do a peeing contest, but was in RF design (primarily for military and govt agencies) 26years as an Engineer, about 6 before that as a tech. 7 ears of electronic repair before that. Not god's gift to electronics but I do know my way around a signal chain.
Here's a link to an (old) article from when this stuff was new, see figure 2 which shows the waveform (without pilot) when one channel is an audio sinewave and the other is silent.
_
Make super easy diffusors:--> http://www.diyaudio.com/forums/everything-else/269366-making-easy-diy-depot-sound-diffuser-panels-step-step.html#post4215464
Horn Design Spreadsheet:--> http://libinst.com/SynergyCalc/
What I objected to was calling it sampling.
You can use a switching modulator to perform amplitude modulation, but that's not the same thing as sampling.
Mathematically, a switching modulator can be shown to produce the desired AM wave plus the square wave and a bunch of harmonics. The square wave and the harmonics can be filtered out. When you do that, you get the proper, continuous-time amplitude modulated signal. It is not a discrete-time signal, it is not sampled.
"And, yes, that's ANALOG FM, as started back in the 1960's! Digital FM is of course even more convoluted. "
Zzactly. I have been in arguments about whether that is actually quantizing or not. It is in one dimension but not the other. It is quantized in time but not amplitude, there are still an infinite number of amplitudes. But there is still that time thing.
Now when we quantize the amplitude a few other things happen. First of all if the source is completely noise free they have to add noise to it, it is called dither. It is just enough to trigger the ADCs and it sort of biases them like an amplifier. In fact on the old AAD CDs made from the old master tapes they had enough noise on them from the tape to not need dither and in some cases the quantization process subdued the noise in a very unobtrusive way, no Dolby, DNR or DBX or anything of the sort.
I actually was thinking of building a device that would do that outboard, and user calibratable. Set up a nice high bitrate, but then the lower bits close to digital zero are expanded just a bit to ignore most (but not all) of the hiss from the tape or whatever. Turned out that really the demand would never be enough and I would need too much help on it to really make money. To do it right you have to modify an ADC and really, the AAD CD did it well enough anyway.
But when they got to the DDD CD they had to add he noise, which you never hear. People, I mean alot of people REALLY could hear the difference, not just the golden ears. The dither, which is basically white noise actually biases the system just like an audio amplifier. you don't hear it, but without it, it just sucks. In fact even though they are extremely rare, there have been designs of amps with AC bias or even noise bias. Long abandoned designs, and I am not sure why. It almost seems more logical because it would tend to linearize the outputs' gain at lower levels better than DC bias. Or maybe not. Whatever. Since we do not hear of this now it obviously failed. Maybe it was too hard to get the thermal tracking right, I dunno.
But the fact remains that regular people could hear the difference, not just audiophiles, so they put the dither in.
Post a Followup:
FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: