Cascaded Vocoders?

[Lscott]

There has been some discussions of a cross-band digital repeater. Sort of buying one purpose built that leaves rolling your own. The easiest one is to cross couple two radios together using the microphone and speaker ports, with the necessary signal level conditioning. Some discussions on other boards seems to suggest trying to run digital through an analog repeater most likely won't work. However some have claimed they did manage to get it to work, sort of with certain types of digital voice modes.

In a real digital repeater the signal is not decoded to audio, just pure digital data, and then fed to the digital transmitter. The vocoder is not really used so no signal quality issues due to the conversion between analog audio and digital should be present.

The only "analog" repeater that has any chance of passing a pure digital signal I've run across is called a "linear translator", or "linear transponder". Since the signal is not converted to audio there should be no issue.

http://www.amalgamate2000.com/radio-hobbies/radio/dunedin linear.htm

http://www.amalgamate2000.com/radio-hobbies/radio/Rotorualinear.htm

The idea I'm thinking about implies transmitting to the repeater, with a digital radio as the receiver, decoded to analog audio, coupled to another digital radio to re transmitted in digital. The vocoder that is most likely to be found in the radio is the AMBE2+ from DVI. These vocoders do not digitize the voice but extracts the elements of speech the human ear recognizes as intelligence and transmits just that as a set of vocal track "model parameters". On reception the vocoder uses the info to "simulate" a human vocal tack to reproduce human speech.

The question is has anybody had any experience with using two vocoders in cascade? Since the second vocoder is not really trying to model real human speech I'm wondering just how much the quality degrades doing this?

 

[WROZ250]

Depends on the digital technology. 

DMR for example, would likely not work by simply cross connecting the speaker and microphone between two radios.  Indeed, pretty sure that is also the case for P25, Fusion, and D-Star.  Could it work? Possibly.  I suspect the audio quality would be less than optimum.  That sort of goes to your question about vocoders.  Nevertheless, virtually all digital systems convert the analog to digital before feeding it to the vocoder.  The comment about 'elements of human speech' is true for P25 but that is not necessarily true for other digital formats.

A better way would be to use one of the MMDVM modem cards (the ones without the onboard radios...i.e. not a hotspot modem).  This solution, however, typically means connecting directly to the modulator and discriminator of the radios.  That said, many radios such as the Motorola mobiles with the expanded accessory connector or, a stand alone repeater, provide these connections, so hookup can be simple.  The downside is that virtually all of these MMDVM modems require that the modulation be manually adjusted (A spectrum analyzer is a definite plus here for best results (but not strictly needed).

Good Luck!

[Lscott]

2 minutes ago, WROZ250 said:

The comment about 'elements of human speech' is true for P25 but that is not necessarily true for other digital formats.Good Luck!

I believe D-Star uses the AMBE while DMR, P25, and NXDN uses the AMBE2+ vocoder. At lest that's what the brochures for the radios I have claim.

The dPMR format popular in Europe I've seen has used several different vocoders while the cheap Chinese radios that claim to be dPMR used something like AMBE2+C which, apparently, is really ASLEP, or ACELP. 

https://radiosification.blogspot.com/2018/02/aselp-vocoder.html

I also know one of the major problems with AMBE type vocoders is background noise. It seriously mucks up the algorithms used to extract the model parameters. The manufactures mostly fixed that by including various types of signal filtering on the analog audio signal in the radio. With NXDN used by the railroads I read a lot of the older equipment is controlled using DTMF tones which the vocoder doesn't handle well to the point where they revert back to analog FM in those cases.

I've also been looking at the idea, where as you pointed out, connecting radio's discriminator output to the other radio's modulator input. I'm not sure that really works so well if at all. If it does I would like to see a good explanation why it does.

The Q4FSK modulation used by many digital formats requires the carrier frequency be shifted a couple of steps above and a couple below the unmodulated carrier frequency for a total 4 possible "discrete" frequencies. However with FM audio, applied to a modulator, results in a frequency deviation above AND below the carrier, not the step change above or below the carrier depending on the data you see with digital. The frequency    deviation depends on the signal amplitude while the harmonics present depends on the frequency content of the audio. 

[WROZ250]

2 hours ago, Lscott said:

The Q4FSK modulation used by many digital formats requires the carrier frequency be shifted a couple of steps above and a couple below the unmodulated carrier frequency for a total 4 possible "discrete" frequencies. However with FM audio, applied to a modulator, results in a frequency deviation above AND below the carrier, not the step change above or below the carrier depending on the data you see with digital. The frequency    deviation depends on the signal amplitude while the harmonics present depends on the frequency content of the audio. 

True, but FM itself is simply variations of phase.  You kinda have to see digital modulation on an analyzer.  Virtually all digital transmitters are just phase modulation. Quadrature modulation is basically created by the phase and, you can't really look at any FM modulation in just one dimension.  It's always just a given phase angle at a given time in space.  All of that of course is why you really need a spectrum analyzer to see and adjust the modulation of a digital transmitter (Yeah, I can see Randy stroking out from that last comment, but I just can't think of a simpler way of describing it).

 

[WROZ250]

The issue isn't so much the vocoder(s) involved, as it is how the desired base-band (audio) is received, processed, and where and how it is fed into the transmitter chain.
That said, there are some FM modulators that will, or can, have issues trying to generate a multi-phase digital modulation.  I seem to recall some of the early radios (I could be wrong, but Syntor comes to mind) having a compensation adjustment for digital modulation, because the digital modulation would screw with the PLL.

This isn't worth overthinking (IMHO) just look how others have done it and what their success level was.

[JLeikhim]

You can test the effects of double vocoding rather easily by connecting a simplex DMR or P25 radio to an Argent Data Systems simplex repeater. The simplex repeater will record demodulated digital audio (analog) and then repeat the analog signal through the transmit vocoder. In choosing a simplex repeater, choose the highest sampling rate as not to introduce distortion .

While one can make a duplex P25 repeater from two radios, making a duplex DMR repeater is a huge challenge as in the duplex mode, subscribers require synchronizing with the repeater transmitter.

Sent from my SM-T350 using Tapatalk

[gman1971]

19 hours ago, Lscott said:

There has been some discussions of a cross-band digital repeater. Sort of buying one purpose built that leaves rolling your own. The easiest one is to cross couple two radios together using the microphone and speaker ports, with the necessary signal level conditioning. Some discussions on other boards seems to suggest trying to run digital through an analog repeater most likely won't work. However some have claimed they did manage to get it to work, sort of with certain types of digital voice modes.

In a real digital repeater the signal is not decoded to audio, just pure digital data, and then fed to the digital transmitter. The vocoder is not really used so no signal quality issues due to the conversion between analog audio and digital should be present.

The only "analog" repeater that has any chance of passing a pure digital signal I've run across is called a "linear translator", or "linear transponder". Since the signal is not converted to audio there should be no issue.

http://www.amalgamate2000.com/radio-hobbies/radio/dunedin linear.htm

http://www.amalgamate2000.com/radio-hobbies/radio/Rotorualinear.htm

The idea I'm thinking about implies transmitting to the repeater, with a digital radio as the receiver, decoded to analog audio, coupled to another digital radio to re transmitted in digital. The vocoder that is most likely to be found in the radio is the AMBE2+ from DVI. These vocoders do not digitize the voice but extracts the elements of speech the human ear recognizes as intelligence and transmits just that as a set of vocal track "model parameters". On reception the vocoder uses the info to "simulate" a human vocal tack to reproduce human speech.

The question is has anybody had any experience with using two vocoders in cascade? Since the second vocoder is not really trying to model real human speech I'm wondering just how much the quality degrades doing this?

 

Double vocoding nullifies all the benefits of a digital repeater. You lose the ability to send data, radio checks, use ARS services, text messages, etc. Double vocoding is probably fine for a short range vehicular repeater, but IMO, its not useful for anything else... it should be a last ditch effort to get a repeater going.

G.

[Lscott]

18 hours ago, WROZ250 said:

True, but FM itself is simply variations of phase.  You kinda have to see digital modulation on an analyzer.  Virtually all digital transmitters are just phase modulation. Quadrature modulation is basically created by the phase and, you can't really look at any FM modulation in just one dimension.  It's always just a given phase angle at a given time in space.  All of that of course is why you really need a spectrum analyzer to see and adjust the modulation of a digital transmitter (Yeah, I can see Randy stroking out from that last comment, but I just can't think of a simpler way of describing it).

 

Well for at least NXDN Q4FSK the technical docs have the following.

Now the screwy part is for analog voice we talk about just the “Deviation” nothing about positive or negative. 

Now am I correct in thinking the frequency out of the demodulator is one of those in the table. It’s not a pure carrier frequency shift. But when they show a negative frequency it’s the same FM modulation but the demodulated frequency is reversed in phase compared to the positive one. That makes more sense to me.

If so that might explain why some have claimed success sending digital through an analog repeater. It does impose some tough requirements on the phase response of the audio circuits. I know human hearing is not sensitive to phase from what I read some time back so the designers likely wouldn’t care about it in the audio path in the radio.

[WROZ250]

I need to clarify something...

First, this specific subject (at this level) is a bit out of the scope of the site (IMHO).

Moving on...

The difference between phase modulation and 'True FM' is, fundamentally, how the information is imposed on the RF carrier.  That said, all FM signals, be it Voice, Data, or Combined Voice&Data is still just 'FM', but more typical these days, it is generated in the radio using a phase modulator.  Technically speaking, even Analog Voice (generated by a 'true FM modulator') is phase modulation, which is why an FM discriminator will recover information from either 'True FM" or Phase Modulation.

Correct, the human ear is not sensitive to phase (as far as I know).  It is sensitive to pressure which, after the radio has done it's job (regardless of the modulation/demodulation), is what the speaker reproduces after any internal processing.  At that point, other than perhaps fidelity, what goes on inside the radio really has nothing to do with what one hears coming from the speaker (assuming a properly working radio). 

That however, is not necessarily the same as the information (prior to processing) within the radio and, internal to a repeater, that gets transferred between the transmitter and receiver portions.

As gman1971 alluded to, even if one manages to pass voice information by connecting two radios 'back to back' there is other 'digital' information that would never be transferred through 'base-band' (audio), and so many 'digital features' of the particular technology in question would not function.   Basically, you have de-rated (degraded) a digital system to analog, that the voice is still digitized is irrelevant.

IMHO, all of the above said, this discussion has gone way too far beyond the original question.

[Lscott]

1 hour ago, WROZ250 said:

IMHO, all of the above said, this discussion has gone way too far beyond the original question.

Yes it has, but for me still very interesting and useful. I’ll have to did through some books I have in my personal engineering library to see what I’ve got on digital modulation methods. It’s got my interest up a bit. One never quits learning until you’re dead.

[Lscott]

If anybody has the interest there is some info in this PDF on digital modulation.

https://ee.eng.usm.my/eeacad/mandeep/EEE436/CHAPTER2.pdf