Advantages of digital over FM

[Project25_MASTR]

You're contradicting Shannon's theorem. How much data can you send in a 3 kHz channel? Approx. 30 kbit/s using Trellis coded modulation. https://en.wikipedia.org/wiki/Modem#V.34.2F28.8.C2.A0kbit.2Fs_and_33.6.C2.A0kbit.2Fs
That's ignoring today's advances in audio codecs (re. CD quality vs lossy encoding).

Not really. You have to bring all of Hartley's, Shannon's, and Nyquists theorems and laws into play.

If you have a bandwidth of 3000 Hz, the Nyquist rate is 6000 baud (not to be confused with bps). Applying Hartley's law (which defines the number of definable bits per symbol) with a standard 4FSK modulation scheme, then 12,000 bps is the maximum throughput you can get. Now, if you go to something like 64QAM, that number increases by a factor of 6 (so the limit would be 72,000 bps). Then you have the Shannon-Hartley theorem which actually states the channel capacity is based upon the SNR (limit as you approach an SNR of 0 dB tells you the maximum amount of packet loss you can handle before you are under the noise floor).

Then you are starting to get into modulation indexes (which actually measure the deviation versus bandwidth or deviation versus data throughput)...

 

[jonwienke]

Actually, you are. You're ignoring the fact that the maximum data rate that can be encoded in a channel with a given bandwidth is directly proportional to bandwith of the channel AND the the signal/noise ratio of the channel.

"When calculated, the Shannon capacity of a narrowband line is log2 (1+Pu / Pn), with Pu / Pn the (linear) signal-to-noise ratio. Narrowband phone lines have a bandwidth of 3,000 Hz so using Pu / Pn = 1000 (SNR = 30 dB), the capacity is approximately 30 kbit/s."

Real-world digital radio modulation schemes have to be able to work well with far less than 30dB signal/noise ratio in the channel. And that limits the usable bit rate to far less than 30Kb/sec. Current digital modulation schemes are actually pretty close to the Shannon limit, given the design criteria of matching or beating the intelligibility of analog FM at comparable RF signal/noise ratios. There are still incremental improvements to be made in improved error correction and compression algorithms, but that's it.

 

[quad_track]

Well, I was reacting to the statement that you can't fit lossless audio in 12.5 kHz. You absolutely can.
You didn't mention anything about SNRs but since we're talking that, fact is: you can have spectral components up to 48 kHz in modern 10 kbit/s codecs, and you can fit that codec in 12.5 kHz with lots of room to spare at an adequate Eb/N0 comparable to the 4FSK curves (which are for coherent detection btw., something which none of the above mentioned radios do).
That audio will sound perfect at SNRs at which analog FM is noisy nevermind reduced audio spectrum.

So digital beats analog both on weak signal and high quality.
I'd love to chat more about this, but I think I'd better point you to some papers by Jean Marc Valin who's more reputable than me

 

[jonwienke]

The context here is comparing digital and analog performance when both are using a 12.5KHz channel with a S/N ratio such that intelligibility is borderline. In that context, your claims are ludicrous because the throughput figures you're throwing around require a far higher S/N ratio than would be practical for a real-world radio link. Yes, you could increase throughput and increase overall audio quality, but at the cost of significantly reducing the range at which the bitstream could be demodulated with a reasonable error rate. Better audio quality is useless if it only works at 1/10 the range of analog.

 

[quad_track]

I don't understand what's with the 12.5 kHz figure and why you think you have to use the same bitrates for both weak signal and high quality.
Let me state this again: for weak signal you can use bitrates of up to 700 bits/sec and throw a rate 1/2 code, fit it in 2.4 kHz and decode at a SNR of 2 dB. Take that, SSB.
High quality, 10 kbit/s codec rate, 7/8 rate code and decode at 12 dB SNR. So much for FM

Lots of people don't believe in a round Earth or ghosts, but that doesn't change the fact that the Earth is flat and ghosts exist.

 

[jonwienke]

I don't understand what's with the 12.5 kHz figure

Because that's the channel width involved in the comparison in your original post. Apples to apples and all that.

for weak signal you can use bitrates of up to 700 bits/sec

So you acknowledge that real-world S/N ratios are relevant for practical radio comms, and that the bit rates you were tossing around earlier are not practical for real-world radio comms where usable range is an important factor, and the system has to work even if S/N ratio is low.

High quality, 10 kbit/s codec rate, 7/8 rate code and decode at 12 dB SNR

12db RF S/N ratio != 12dB SINAD.

 

[prcguy]

If a radio digitizes the signal at some point and uses computer simulation of various functions like demodulation, IF BW, AGC, etc, then its simply an SDR. Direct conversion and superhet hybrids are all SDRs and it doesn't matter if its run from a desktop computer or if its a box with display and knobs.

There are even some direct sample SDRs that emulate dual conversion superhetrodyne completely in software. I think its best just to call an SDR an SDR and elaborate on its topology separately like direct conversion or superhet. I don't think the word "full" fits in anywhere when discussing SDR radios because you can't pin down a specific type to a pure breed and the future will bring even more simpler designs like the entire radio from antenna input to speaker output on a single chip.

With all that said I think the first SDR handheld ever marketed was the Racal MSHR released in about 1999 and the subsequent MBITR in about 2000. The MSHR was VHF only and analog/LPC10 digital with encryption. The MBITR is 30-512MHz with various analog and digital modes upconverting to an IF around 1700MHz then being digitized at some point.
prcguy

Careful with the terminology. Is it a full SDR in that the incoming RF is direct sampled? Or, is it a hybrid SDR using superheterodyne techniques to down-convert the RF to an IF that is then sampled? I'm not aware of many full SDR VHF-UHF receivers in the hobbyist market.

 

[paulears]

If you can squeeze broadcast audio into 12.5KHz channels, then why don't they? Because they can't.

Back a bit you said you weren't talking about DMR or dPMR - what are you talking about then? The topic started talking about VHF upwards.

If I'm reading it correctly, people are trying to explain that with comms we do not waste bandwidth transmitting pointless data, and that frequency response is irrelevant in this context. Somehow we got an example of HF digital thrown into the mix. As far as I can tell, there is some serious misunderstanding about digital transmission techniques going on and none of the attempts to explain this have convinced quad track he's got a bit confused. If we're talking audio, then it's simple to see how quality and digital data are conflicts - You only have to look at the DAB spec to see how even gentle mangling spoils the sound. We don't want to get sidetracked into full audio bandwidth digital techniques - we're talking comms, and clearly 48K sample rate with 24K frequency response is NOT what we're talking about.

 

[prcguy]

I agree and the type of hand held digital radios being discussed are the farthest thing from high fidelity you can get. I use an amount of digital radios in DMR, P25 and military LPC-10 modes and in many cases the audio fidelity of these radios is so bad I have no clue who is transmitting because there is too much information missing to recognize a voice.
prcguy

If you can squeeze broadcast audio into 12.5KHz channels, then why don't they? Because they can't.

Back a bit you said you weren't talking about DMR or dPMR - what are you talking about then? The topic started talking about VHF upwards.

If I'm reading it correctly, people are trying to explain that with comms we do not waste bandwidth transmitting pointless data, and that frequency response is irrelevant in this context. Somehow we got an example of HF digital thrown into the mix. As far as I can tell, there is some serious misunderstanding about digital transmission techniques going on and none of the attempts to explain this have convinced quad track he's got a bit confused. If we're talking audio, then it's simple to see how quality and digital data are conflicts - You only have to look at the DAB spec to see how even gentle mangling spoils the sound. We don't want to get sidetracked into full audio bandwidth digital techniques - we're talking comms, and clearly 48K sample rate with 24K frequency response is NOT what we're talking about.

 

[Ed_Seedhouse]

I think you are confusing audio frequencies with bit rates. They are not the same thing.
You can't simply say that we can transmit a 5000 kHz signal using analog, but 48000 Hz using digital. It's not a direct conversion.

No one can hear 48,000 Hz.
Human hearing gives out above 20,000 Hz. and 5,000 Hz is plenty for intelligibility.

There are clearly advantages to digital, but frequency response is not one of them for the ham bands.

 

[Ed_Seedhouse]

If you can squeeze broadcast audio into 12.5KHz channels, then why don't they? Because they can't.

Strange then that Broadcast band A.M. radio limited to +/- 5,000 Hz is still in wide use nearly a hundred years after it first began.

 

[ko6jw_2]

Digital vs FM? They are different types of signals, but both are FM. Your radio does not transmit 1's and 0's.The conversion is complicated by the number of digital modes. I am the trustee of a repeater with dual mode analog and C4FM. In the year and a half that it has been on the air I can count on the fingers of one hand the number of times it has been used in the digital mode. The analog users hate it because it sounds like white noise to them. I predict that it will take several years to convert people. I have a digital radio that does C4FM, but not DStar, DMR, P25 etc etc etc. The demise of analog is a long way off.

 

[jonwienke]

Strange then that Broadcast band A.M. radio limited to +/- 5,000 Hz is still in wide use nearly a hundred years after it first began.

Which is not anywhere the same audio quality as broadcast stereo FM, which is what was referenced...

 

[ko6jw_2]

AM frequency response is a function of bandwidth. That is, the upper and lower sidebands will appear on each side of the carrier separated by the frequency being transmitted. If you wanted to transmit 20Khz you would need a 40Khz wide channel.

 

[bill4long]

If you're using signal strength variations vs position changes to DF, you're doing it wrong. The linked unit will calculate an accurate vector to the transmitter within a second of keyup.

Not "wrong." Just more time consuming, and less convenient, because it usually means getting out of your vehicle to take a measurement.

I had a doppler for 2 meters back on the 80s. Worked great and found a lot of jammers and had fun with fox hunts (which spanned many hundred square miles.) You still have to do position changes with a Doppler, in order to triangulate. The doppler is just a convenient way to show the direction from your current position from the comfort of the inside of your vehicle. Of course, if you have a friend or two out there with dopplers (or any sort of directional antenna) than triangulation can happen in seconds.

 

[N4KVE]

Unscientifically, a full quieting FM signal will sound better than digital. But when that analog signal is very noisey, the digital signal will still be crystal clear.

 

[bill4long]

No, I'm not confusing bitrate with audio sample rates. An analog radio will only have an audio frequency response of approx. 300 - 5000 Hz. But with digital, you can sample audio at whatever rate you want, encode it and then send it. Usually the low bitrate codecs use 8000 Hz as an input sample rate because they're not interested in fidelity, while the high bitrate codecs use 48000. The output bitrate depends on the codec's capacity of compression. Modern codecs will have great frequency response for low bitrate. A 10 kbit/s codec can carry audio of up to 48000 Hz in this case.

The AMBE 2 codec that DMR tier 2 (and Mototurbo which is a superset) and C4FM (which Yaesu uses) has a max sampling rate of 8khz, which means it has an effective 4khz response.

 

[bill4long]

You can sample at 48KHz and transmit CD-quality audio digitally, you just can't do it through a 12.5KHz channel.

Right. DMR tier 2 (Mototurbo is a superset) use a two-slot (6.25khz each) in a 12.5 khz space. NXDN uses a 6.25khz space. These use the AMBE 2 codec, as I previously mentioned. Audio frequency response approximately 4khz of decent audio.

 

[bill4long]

Strange then that Broadcast band A.M. radio limited to +/- 5,000 Hz is still in wide use nearly a hundred years after it first began.

Not quite correct. Most of that time, in modern broadcasting in the USA, AM have a transmitted bandwidth limit of 15khz, with total occupied b/w of 30khz. That changed in 1989 when the transmit b/w was tightened to 10.2khz with a max of 20.4 occupied b/w. I suspect this change was made because music broadcasts on AM were largely abandoned for FM. There is a significant audible different between 15khz and 10.2khz when it comes to music, but irrelevant when it comes to talk radio.

 

[bill4long]

Unscientifically, a full quieting FM signal will sound better than digital. But when that analog signal is very noisey, the digital signal will still be crystal clear.

Totally subjective. Some people love digital. Some hate it. And everything in between.

Personally, I don't care for DStar much. It has the old AMBE 1 codec and sounds much more quantized than the AMBE 2 codec which DMR (and Mototurbo) and C4FM (and Yaesu) uses. I rather like the way the AMBE 2 codec handles the compression. Quite acceptable, very clear, with excellent frequency response. Fantastic S/N. I still prefer FM for extended ham radio ragchewing, and probably will, because it sounds warm and totally unquantized. For commercial or public safety, then AMBE 2 digital would definitely be the choice, because the "cut through" clarity is maximal. Different tools for different purposes.