HD Audio

[trbonaut]

Hello,

There is something I've been looking forward to for a long while now, yet it doesn't seem to be happening. Namely: better sounding digital radios. Codecs, modulations, signal processors, processing power etc. have come a long way, yet we're stuck with DMR and the like, with prehistoric codecs. And my question is: why?

I am not aware of any narrowband radio system supporting HD audio (please correct me if I'm wrong), yet with modern codecs, higher tier modulation and robust error correction, this would seem like a possibility. I understand that there are bandwidth constraints to consider, but just merely looking around how audio codecs evolved over the years, I don't get why we don't have anything better. Wouldn't it make sense to have better intelligibility or it's merely about "good enough"?

Some Chinese POC radios, Zello and probably others already offer high fidelity audio, yet there doesn't seem to exist a simple walkie-talkie with such feature.

What's your take? I'd love to get some feedback.

 

[a417]

"HD audio" is marketing wank...pointed at the same people who buy gold plated Monster Cables, and the like.

Tonal range of entertainment/music & human voice vastly differ, and the codecs in place on the wide spectrum media side and the high ambient noise digital radio side should stay far far far away from each other.

With the public safety side, you're worried about human voice over a whooping PASS alarm or siren, and in the music/entertainment side you're worried about creating large asthetically pleasing sounds - polar opposites IMO.

This industry doesn't need "HD Audio"...my $0.02.
There are some highly intelligent people who are exceptionally knowlegable about this topic, here on RR, and I hope they chime in.

 

[mmckenna]

G.722 codec is pretty good. I've got our PBX set up to run that internally. It sounds much better than the lower rate CODECs (G.711 and G.729) and much better than POTS. When we first started using it, it was really impressive.

Just fine on a well designed IP network. Not so good if you have limited bandwidth.

Which is the issue, bandwidth. You can only compress things so much. LMR is (mostly) limited to 12.5KHz of channel width (you can run 20 and 25KHz on some other bands…) Trying to cram a fat CODEC into 12.5KHz won't work. NXDN can fit about 9.6Kb/s in a 12.5 channel, but that only gives you room for lower quality audio. Plus you add in FEC and the headers, you have even less room. You can only compress things so far. 6.25KHz channels, even less. P25 Phase 2 and DMR are also limited.

For radios, FEC becomes a bit more important out on the fringes. Being able to recover a usable signal out of the noise is more important than high rate audio codecs.

When you look at the human voice, POTS systems and the like, 300-3000Hz gets you most of the useful audio. Anything above/below that is kind of a waste since most human voices don't run much outside that (except my niece who can probably hit 20KHz easy).

It is available on the cellular/WiFi apps because they don't have the bandwidth limitations that LMR has.


And, cue the other old farts that will reminisce about 25KHz, 30KHz and wider FM analog. Which is a very valid point.

 

[BMDaug]

I want basic intelligibility… I don’t need rich, lush audio with lots of lots of bandwidth… it just lets the noise creep in! I’m not transmitting/receiving music and I’m using the equipment in high noise areas, where limited bandwidth is my friend.

Ya, in an office on a VoIP system, sure, the bandwidth is welcomed and quality is appreciated, but I have no issue with NFM from a 20+ year old HT750… I wish hams would just go narrow… the ‘old repeater our club uses doesn’t go narrow’ excuse is BS. If it’s THAT old, it probably needs replacing anyway and there would be a lot more space for repeaters in the limited if everything was 12.5KHz.

-B

 

[W8HDU]

As a TV broadcast guy, the term HD is often mis-used by the public and manufacturers. The bottom line is in communications there are two ways to efficiently transmit audio, and that is with low bit rate or reduced bandwidth. 8 kHz, 16 bit is about the bottom of the line, but there are some who call digital audio and use 6 kHz, 8 bit, and that is very noticeable.

 

[tunnelmot]

And, cue the other old farts that will reminisce about 25KHz, 30KHz and wider FM analog. Which is a very valid point.

Dude you ain't lying. I've been chasing that dragon since being around 110w X9000s on a public safety system as a 12 year old. I rode in take-home Chrysler K Cars where the 25khz Spectras had better audio than the factory AM/FM radio!
Ever hear an HT1000...on wide band? 🤣
I kid of course, but I guess I'm officially an old fart now LOL.

I wish hams would just go narrow… the ‘old repeater our club uses doesn’t go narrow’ excuse is BS. If it’s THAT old, it probably needs replacing anyway and there would be a lot more space for repeaters in the limited if everything was 12.5KHz.

Yuuuup. How many Micor and MasterII repeaters suffered in silence ID'ing into the empty nighttime ether with not so much as a kerchunk?😂
Imagine opening up spacing for all kinds of local P25, NXDN, etc. repeaters. I'd love to see a bandplan where legacy wideband FM machines keep their pairs, but narrowband channels opened up for narrow digital modes. One can wish I guess.

Forgive my humor. It's Friday and I'm enjoying my Yuengling's.

 

[speedway_navigator]

pointed at the same people who buy gold plated Monster Cables, and the like.

I walked into Radio Shack to get an audio cable and was 'helped' by a young kid wanting to earn his commission. He showed the gold plated cable and I asked if the whole cable was gold plated, he admitted it was just the connectors.

I asked him what the advantage was and he said the signal had better performance with gold.
So my smart assness kicked in and I said "so the signal starts out faster at the connector, then slows down in the cable and speeds up again at the other end?"

He walked away.

 

[speedway_navigator]

I've often wondered about the science behind codecs that remove parts of the audio spectrum someone has decided we don't need to comprehend what we are hearing. I've seen pictures of Styrofoam heads with headsets over mics in the ears to show us what the data minders think we hear.

 

[mmckenna]

So my smart assness kicked in and I said "so the signal starts out faster at the connector, then slows down in the cable and speeds up again at the other end?"

"Conductor induced wow and flutter".

 

[dlwtrunked]

"HD audio" is marketing wank...pointed at the same people who buy gold plated Monster Cables, and the like.

Tonal range of entertainment/music & human voice vastly differ, and the codecs in place on the wide spectrum media side and the high ambient noise digital radio side should stay far far far away from each other.

With the public safety side, you're worried about human voice over a whooping PASS alarm or siren, and in the music/entertainment side you're worried about creating large asthetically pleasing sounds - polar opposites IMO.

This industry doesn't need "HD Audio"...my $0.02.
There are some highly intelligent people who are exceptionally knowlegable about this topic, here on RR, and I hope they chime in.

Exactly right. Most people do nut understand that human voice intelligence is in the higher frequencies and bass characteristics, though good for music, does not helo in understandabilty of voice; in fact, it has the opposite effect.

 

[speedway_navigator]

When I was with Bearcom and first went with DMR we had NBA users ask to remain on analog because they could recognize users by voice, the digital channels sounded too different.

 

[mmckenna]

When I was with Bearcom and first went with DMR we had NBA users ask to remain on analog because they could recognize users by voice, the digital channels sounded too different.

One of the reasons I chose NXDN.

 

[ElroyJetson]

Two things that always strike me about even the most modern digital audio codecs: The total lack of background noise under normal conditions. And the much-worse-than-analog quality of voices. At best the voice quality is different, and it always takes an adjustment period where you learn how to hear digital voices clearly.

You can do a fun little experiment by checking with your local car audio store and getting any 4" full range speaker, even a used take-out out of a car, and put that in your mobile radio's speaker housing and see how much it improves the audio quality. It should make a significant difference, but nothing is going to make AMBE 2 sound like analog voice. Or any other bandwidth limited codec used in commercial and public safety voice communications.

Personally I think that this drive toward digital everything is being powered by marketing, and not by need. Analog narrowband voice does the job as well as digital in the same footprint and you don't have to re-learn how to listen to it.

In any event we're only looking at bandwidth limited audio, 300 to 3000 Hz, with a narrow dynamic range. There's not much that can be done to make it sound "HD" but going to analog is a good start.

I don't believe in making communications more complex and expensive than is strictly necessary. Digital, trunking, encrypted, there's three layers of cost and complexity that stands in the way of simple voice communications. But it sure is more profitable for the salesdroids and manufacturers!

 

[AM909]

Another data point: Icom's LTE radios' implementations of G.726 @ 32 kbps sound absolutely incredible.

 

[hp8920]

Look at the bitrate that digital LMR systems use. P25 FDMA and NXDN 12.5 kHz are 4400 bps of voice data, and DMR, P25 TDMA, and NXDN 6.25 kHz run 2450 bps. G.722 runs at 64 kbps, and AMR/EVS WB runs at typically 13.2 kbps, more than 5x what LMR can do, particularly with 6.25 kHz regulatory requirements.

I mentioned on another thread that the tiny bandwidths of LMR are a severe hinderance to performance. If it was up to me, I'd make LMR channels 10 MHz wide and force all users to do spread spectrum. Then you could up the bitrate to something reasonable.

 

[mmckenna]

I mentioned on another thread that the tiny bandwidths of LMR are a severe hinderance to performance. If it was up to me, I'd make LMR channels 10 MHz wide and force all users to do spread spectrum. Then you could up the bitrate to something reasonable.

You just described LTE.

 

[hp8920]

You just described LTE.

And IS-95 CDMA, and WCDMA, and Wi-Fi, and Bluetooth.

What you'd do is license the spectrum based on power. Within this footprint and this channel, you can contribute x dBm out of a noise floor limit of y dBm. Spectrum efficiency is increased because if you aren't talking, you aren't transmitting and adding to the noise floor, and you can license more users, instead of reserving some number of fixed channels forever.

That's how we regulate Wi-Fi today. At 5 GHz, you're allowed 17 dBm max in any 1 MHz bandwidth.

 

[speedway_navigator]

Do you think the reason G.722 and other higher quality codecs are not used is do to licensing fees or chip cost/implementation costs?
Better marketing by the lower quality marketing groups?
(that's a silly question, no company would put profits over public safety)

 

[hp8920]

It's the opposite. G.722 is so old, originally approved 1988, that all patents have expired. AMR-NB, released 1999, should be off patent. AMBE/IMBE traditionally has huge patent licensing fees, DVSI have been pains in the butt about it, requiring their own codec chips, even though they promised FRAND.

Again, look at the bitrates. G.722 takes 64 kbps without error correction, whereas DMR, NXDN 6.25 and P25 TDMA provide 3.6 kbps total and 2450 bps after FEC. It's a consequence of narrow channels and the weaker signal (high-site architecture) of LMR.

 

[mmckenna]

Do you think the reason G.722 and other higher quality codecs are not used is do to licensing fees or chip cost/implementation costs?
Better marketing by the lower quality marketing groups?
(that's a silly question, no company would put profits over public safety)

I think at one point, when they still charged me, G.722 feature on our PBX was $0.74 per phone.
If I was paying 74¢, I'm sure they were paying much less.