Ham Digital Voice Tech Notes

D-Star uses DVSI's AMBE codec. AMBE is slated to replace the IMBE codec in Phase 2 of APCO 25. That is not a coincidence or accident on the part of the JARL or Icom. I'm sure they looked at open source.

One, it puts us as ham radio operators into state-of-the-art in communications for the first time in about 10 years. It's so nice to say that. We appear to be Icom's lab rat for their rollout of commercial P-25 Phase-2 products.

Two, it gives us a common platform with professional public safety personal. When they catch up. Actually I think AMBE is backwards compatible capable with IMBE.

But yes, it would be great if someone came up with a open source codec that supplants AMBE. Which is, btw, starting to show its age.

If you look at M/A-Com's P25IP/OpenSky system it sure looks like D-Star. Of course M/A-Com is about 25 miles from M.I.T. where the codec was born. DVSI's founder is an M.I.T. alumni. And there has been bad blood between M.I.T, DVSI, and Lockheed (who acquired then sold Inmarsat) Inmarsat, Iridium, and XM-radio use AMBE.

While the radios may or may not be directly interoperable (time will tell) it sure seems to me that they would be able to gateway with a little glue-code.

There are some keys to consider when it comes to interoperability. The lowest common denominator (aside from frequency) would seem to be the modulation type. D-Star is GMSK, APCO-25 and pretty much everything else is 4FSK modulation. Upon that you have a MAC layer level access protocol, then the actual codec converting the voice to digital.

As long as there is a hardware common ground modulation type or ability to speak/receive multiple types (like they did with 802.11b & g - DSSS/OFDM), firmware should be able to be written to take care of any MAC layer differences. Transcoding can take care of the differences between the various codecs.

In summary, flashable user updatable firmware and or open source, and open codecs for software based transcoding.

I am concerned that the digital modes so far are being implemented as one might add a modulation technique- still point-to-point, single user at a time functionality.

I'd like to see a dynamic scheme that efficiently utilizes the spectrum to pack in multiple users. One could have either separate conversations going on simultaneously over a single repeater, or a roundtable affair, in multicast form, where multiple speakers could operate in true conference fashion. I have seen this implemented in several of the new combat radio systems in just the past couple of years. I believe this is the true path forward for amateur radio- a flexible communications format that can establish networks ad hoc, and route traffic through auto-discovery modes. As I read a lot of the DSTAR protocol, I see it has weaknesses in this area (I have just started reading about DSTAR in the last day or so- I've just discovered these forums, and it is a great way to get up to speed!!), and I'd like to see the amateur community avoid as many pitfalls as possible. Of course, we've all spent money on shiny new modes that tend to gather dust in five or ten years...Perhaps DSTAR is adaptable enough, but I have doubts at the moment.

One concern I have is DSTAR's selection of CODEC- the AMBE codec is not famous for voice quality. While I agree it is very bit rate/bandwidth efficient, I argue that one needs pleasing voice/intelligibility to be truly successful. I would like to see ham radio embrace an open standard like SPEEX. The minimum bandwidth-efficiency vs. voice quality seems to be right at 6 kbps. While this is more than AMBE's requirement, the voice quality is on par with FM, at a much narrower bandwidth requirement. I'd like to see amateur radio have a codec-agnostic transport scheme, where the endpoints can negotiate a codec based upon channel capacity, latency, and user quality.

I am experimenting with my own VoIP PBX (Private Branch Exchange) server. It has these qualities- it can negotiate with a distant endpoint to select a preferred codec, and can even transcode between codecs, if necessary. One can have endpoints with, say G729, or G711, AMBE, SPEEX (In a mind boggling number of bit rates), etc, and they can all play nicely with each other, as the servers on each end can perform any transcoding necessary.

All this infrastructure is open source (the PBX I'm using is ASTERISK based). Why not utilize the facilities already developed in ASTERISK to develop a ham radio communications network that gives near toll-quality audio performance? We could also do burst mode transmissions- wouldn't it be great to also use the system in a "meteor scatter" mode to get messages across during extreme transient events? Talk about useful features in emergency situations. Of course, there would need to be a lot of science performed to make such a system work, but it's not something that hasn't been tried before. But we do have considerable processing at our disposal that wasn't available throughout most of the prior research on the subject in years gone by.

Just some food for thought...

73,

Shane
KE7TR

I agree with "OPEN CODEC" this certainly would enhance experimentation and growth with D-STAR and other Digital modes.

Especially with the ability for any GOOD programmer/experimenter to add ideas easily. Think of the boundless future!

Ron, k1vsc

It's nice to see someone pushing open standards and license-friendly codecs. And by a guy who is in the industry and knows what he is talking about.

The voice quality with D-Star is not as intelligible as many would like. It would seem that the audio frequency response is primarily a function of the AMBE codec.

Fortunately there is no degradation when linked via the gateways, so at least it does not get any worse.

Open standards is one thing that the Win-Link people have overlooked, i.e that proprietary HF modem card. That's a deal-stopper, for many.

Will the digital hams listen to what this guy says? Will they grasp the open standards idea? This seems unlikely as long as there is a commercial interest involved (Icom). Seems to me that Icom is not completely open about the D-Star system as some people would like (undocumented protocol and other operational things). There is commercial inertia here, too. D-Star has some traction already.

Sad fact is, hams are thinking that they will be immune to this kind of thing because they are hobbyists. Tell that to DVD Jon, who just wanted to play DVDs on his Linux machine. He was a hobbyist. And so are a lot of other computer geek guys have gotten into scrapes because they were using a proprietary stuff in conflict with its licensing terms.

Real world D-Star experience is "it depends", and the biggest variable is the amount of multipath fade. When both ends are stationary, D-STAR has superb performance. However, if one station is mobile, D-STAR performance really falls off. The FM signal is often fluttery but copyable at levels that D-STAR is totally garbled. Deal with the multipath fade, and D-STAR is a winner in performance.

The processed signal modulates the VCO in GMSK direct frequency modulation using a GMSK-Modem. It is then transmitted as a 20F3E conventional FM signal.

In the receiving mode, the GMSK modulated RX-AF signal comes out of a demodulation circuit and is processed at the GMSK-Modem. This signal is again processed in the CPU, then decoded in Continuous Code Delta Codec (CVSD) to obtain the original analog voice signal (Digital-Analog conversion). As in a normal FM receiver, the signal is amplified for output to a speaker.

Spectral efficiency. The DV format of D-STAR has a bandwidth of 6 kHz, compared to 16 kHz for analog FM with 5-kHz deviation. This is Via Carson's Rule, BW = 2 × (Peak Deviation + Highest Modulating Frequency) = 2 (5 kHz +3 kHz) = 16 kHz.

Modulation methods: GMSK, QPSK, 4FSK
Data rate: Maximum of 4.8 Kbps
Voice encoding method: AMBE (2020) converting at 2.4 Kbps, FEC at 3.6 Kbps
Occupied bandwidth: Maximum of 6 KHz

D-star is a 4800 baud total data stream equivalent signal. Where: 2400 bps is reserved for actual digital voice, 1200 bps is reserved for FEC (forward error correction) on the digital voice. This is for callsign and short message data. And 1200 baud is reserved for serial data. This is for APRS, and text messages/text query's.

APCO P25 Phase I is the present version that is in used across the country for Digital Public Safety, the P25 “open” standard has been reworked by some manufacturers limiting some of the standardization that the P25 was hoped to present..

P25 Phase I repeaters have the ability to function as a analog system or digital system.

P25 Presently operates via FDMA (Frequency Division Multiple Access) with the plan for P25 Phase II to use TDMA (Time Division Multiple Access), P25 Phase II will also have the capability to “roll-back” to FDMA for “conventional emergency operations.”

Phase 1 radios use the IMBE vocoder and Continuous 4 level FM (C4FM) modulation for digital transmissions at 4800 baud and 2 bits per symbol, yielding 9600 bits per second total channel throughput. Where: 4400 baud are associated with the digital voice, 2800 baud are used for error correction on the voice signal, 2400 baud are devoted to signaling overhead.

Receivers designed for the C4FM standard can also demodulate the "Compatible quadrature phase shift keying" (CQPSK) standard, as the parameters of the CQPSK signal were chosen to yield the same signal deviation at symbol time as C4FM while using only 6.25 kHz of bandwidth.

Phase 2 is currently under development with concurrent work being done on 2-slot TDMA and FDMA (CQPSK) modulation schemes. Phase II will use the AMBE vocoder to reduce the needed bitrate so that one channel will only require 4800 bits per second.

Significant attention is also paid to interoperability with legacy equipment, interfacing between repeaters and other subsystems, roaming capacity and spectral efficiency/channel reuse. In addition, Phase 2 work involves console interfacing between repeaters and other subsystems, and man-machine interfaces for console operators that would facilitate centralized training, equipment transitions and personnel movement.

Uses AMBE+2 codec and 4FSK (4 level FSK) / FDMA (frequency-division multiple access scheme) digital modulation

Transmission rate: 4800 bps, with a codec rate: 3600 (voice 2,450 + error correction 1,150 bps)

The is a brand-new digital format being co-designed by Kenwood & ICOM that is geared towards the business sector. It is designed for those that want to meet the up-coming FCC mandate for 6.25 KHz channel spacing, but that can't (or don't) want to move to the APCO P25 Phase-II equipment that will soon come to market. The format is based on the AMBE+2 voice codec (similar to ICOM's D-STAR), but uses a 4-level FSK modulation (FDMA). The radios are capable of narrowband analog, along with 12.5 KHz & 6.25 KHz digital emissions. Kenwood is offering the system under the name NEXEDGE, and the radios are capable of both conventional & trunking operation. The attached sound file contains all of the formats the system is capable of producing, including the raw data streams of both digital formats.

D-Star, developed by ICOM, is the forerunner to the commercial counterpart of the technology we now know as IDAS (ICOM Digital Advanced System).

IDAS, also known as FDMA is the system generally best suited for commercial use since it meets all FCC technical standards through 2018 and is backwards compatible with 25 kHz, 12.5 kHz analog systems plus capable of operating in the digital mode on 25, 12.5, and 6.25 kHz channel spacing.

Icom and Kenwood have jointly developed a very narrowband 6.25kHz digital communications technology using an FDMA 4-level FSK modulation method

The best overview I have found:
http://www.mygmrs.com/nxdn/index.php

NexEdge/IDAS was built off D-Star. Un like D-Star the repeaters can repeat analog or digital so you can make a smooth migration.

It supports unit ID auto-roaming/registration much like how D-Star works.

What I find interesting is that it uses AMBE+2 codec which is supposedly backwards compatible with IMBE.

MotoTRBO does not have analog capability, thus migration from analog to digital requires of separate repeaters.

It is not Motorola proprietary format, but rather it is ETSI (European Technical Standards Institute) (standard number ETSI TS102 361-1). It is a DQPSK (Differential Quadrature Phase Shift Keying) modulation scheme. This is a published standard, as every bit as much an open standard as P25. DQPSK is a digital modulation technique commonly used with cellular systems, so it resembles TDMA in many ways.

It uses TDMA technology to split a single 12.5 kHz (narrowband) channel into two "virtual" 6.25 kHz channels. The result is twice the calling capacity. A TRBO repeater is analog OR digital, but not both on the same channel at the same time. There is no "mixed mode" operation. It could not work as the digital modulation is TDMA and is capable of two time slots of information (voice + voice or voice + data). Typical TRBO digital sensitivity is 0.3 uV for 5% BER.

VSLEP (vectorsum excited linear prediction encoding) has not been produced and supported by Motorola since late 1995. V SLEP was Motorola's offer into the digital game before the IMBE codec was selected. The audio quality between VSLEP and APCO 25 is night and day. VSLEP digitizes the voice's essential elements. It's used with digital sound processing techniques, along with proprietary algorithms owned by the chip maker. VSLEP is the codec of choice for Motorola's iDEN (Integrated Digital Enhanced Network) systems, and Nextel (iDEN). iDEN uses TDMA technology to split a 25 KHz frequency into six separate time slots. iDen is a derivative of Motorola's older VSELP digital trunking transmitted over a standard 6:1 TDMA cell format.

Voice data goes into the codec as 8000 8-bit samples per second. The total input data rate is 64kbps. the output of the codec is 7950 bits per second. this reduction out the output is achieved by removing redundancies that are inherent in human speech. The codec breaks the incoming data into 20ms blocks called speech frames. Each speech frame consists of 160 samples of voice data at 8 bits per sample, there are 1280 bits per frame. Frame energy is sent only once per frame but the accuracy of the frame energy has a large impact on the perceived audio quality. If the frame energy is corrupted the received audio will sound garbled. Wave pattern describes the main features of the signal over one frame. The wave pattern describes the main peaks and valleys which define the pitch of the signal. Residuals are like the comfort noise because the low frequencies have been taken out.

VHF/UHF Digital Voice Samples and Info