Just say NO! to Winlink expansion!

Here is part of my problem with digital voice on amateur bands.

Most folks that are promoting digital voice on ham bands are doing so because we are "falling behind" what is being done on commercial radio. Other than cellular, which is going digital for different purposes, what is the driver for moving to digital voice? From what I have read, the biggest driver is homeland security. Digital radio systems for public safety organizations allow for easy encryption of both voice and data so there are secure communications. Since we can't encrypt anyway this reason doesn't apply to amateur radio.

Digital voice radio systems in the commercial arena are also channelized and mostly trunked type systems. In other words they don't have interference from nearby stations competing for spectrum and they do not operate in a multi-mode environment.

They use fm modulation and are local/regional in nature, i.e. no weak signal type work and basic immunity to atmospheric noise. Wider bandwidths than SSB are also required for adequate communications. With the need for maximum spectrum efficiency if they could adequately communicate in a bandwidth of less than 3 kHz, they would be doing so.

Maybe these are reasons to investigate digital voice on the ham bands. However, I suspect it will be found that digital voice using SSB is inferior to analog voice SSB for weak signal work and for operation in interference laden, both atmospheric and adjacent channel, spectrum.

I hope the folks that insist digital voice MUST be deployed in the ham bands don't also begin to insist that we must channelize and move to FM. HF just doesn't have the spectrum to allow for this without REDUCING THE NUMBER OF HAMS INSTEAD OF INCREASING THEM.

Another problem with digital voice that a lot of hams aren't aware of is the need for reducing the power output of amateur class equipment. Digital voice on SSB has a much higher average power requirement than analog voice SSB.

Most amateur class SSB transmitters and amplifiers have been designed for analog voice modulation, not digital modulation. I suspect power levels for 100 watt class transmitters will need to be reduced to 50 or even 25 watts when using digital voice. This won't make a lot of folks happy! My Collins 30L-1 specifically states "DO NOT use the 30L-1 in FSK, AM, or FM service." I'll also bet we will see a lot of transceivers and amplifiers on ebay and other sales forums with blown finals, power supplies, and tubes because of digital voice.

This is part of the 'technical cost' of assessing this mode. Again, I will bet analog voice SSB will come out ahead in the long run. Some will make the argument that hams are cheap and some are no doubt cheap. However, most are looking for the most bang for the buck and I don't believe digital voice will give it them, at least not for the forseeable future.

Consequently, allocation of space for digital voice at this time shouldn't include any more spectrum than is needed for experimentation. If there is the possibility that this mode will prove inferior to analog voice SSB on the amateur bands then minimum space should be allocated.

Jim
WA0LYK

 

Apparently the big appeal of digital voice (at least in the current version) is "broadcast quality". This seems to be inconsistent with the "quality necessary for communications". I think it is not tolerant of either QRM or QSB like analog SSB is, and the reference to "channelized" commercial digital voice systems is especially appropriate.

I know! The BEST place for digital voice is on 60m!

 

Anyone up for some hifi stereo drm on 30 meters?

 

LPC-10 can be implemented quite easily in a single DSP chip - probably no more powerful than the DSP chip in a modern sound card. You need only model the length and diameter of 5 pipes to model the human vocal tract. Linear Predictive Coding of 5 pipes with 2 coefficients each = LPC-10. The other two basic issues are extracting the fundamental pitch (80 Hz to around 200 Hz in most males, up to a maximum of 400 Hz for females), and determining voiced (vowel) or unvoiced (consonant) sounds.

BTW: A 400 MHz Pentium can be had for $100.00 or less - if you can find a PC that slow. You can probably get at least 800 MHz for $100.00 now. Sound cards are $5.00 to about $200.00 new (though I cannot fathom why anyone would spend $200.00 for ANY sound card.) A $20.00 sound card should be adequate for LPC-10.

The enhanced version of LPC-10 that we used was developed by Time and Space Processing (TSP) in CA. Against our stringent warnings and objections (we had a joint venture with TSP, and built a Strategic-level encryption board that fit into their hardware), TSP sold two "evaluation" units to the Israeli government.

Less than eight months later, at the AFCEA show in Washington, D.C., Tadiran (Israeli company - part owned by the Israeli government) suddenly had an LPC-10 box. All vendors, including TSP did interoperability testing at the show in the MIL-STD LPC-10 mode (MIL-STD-1051, if I remember correctly).

To TSP's amazement, when they connected to Tadiran's box, the box synced up in TSP's PROPRIETARY ALGORITHM. TSP's management threw a fit, and demanded that the Tadiran box be removed from the show. Tadiran claimed that the algorithm was "invented by a professor at Haifa University", and that TSP "must have stolen it from Tadiran". (Bear in mind that we had been on the market for almost FIVE years at that time! -- and that Tadiran had NEVER had an LPC-10 implementation.)

The U.S.G. did nothing (as always when Israel steals something), and TSP almost went broke trying to recover their intellectual property. (They never did.)

The TSP algorithm averages over two frames (highly simplified explanation), while standard LPC-10 "jumps" from one 54 bit frame to the next. The TSP algorithm is also vastly superior in decoding fundamental pitch, and in handling inflections. The difference is dramatic.

Chinese speakers had to employ what they termed "best guess" when using standard LPC-10, but they absolutely marveled that they could both understand all words AND identify the speaker on the TSP algorithm. (The Chinese language is highly inflection dependent. Example; "ma" may be either your mother or a horse - depending upon the inflection.)

Our "high powered" processor was a 16 bit operating at something on the order of 20 MHz, with a TRW multiplier-accumulator to pull the coefficients. If I remember correctly, we had "lots" of memory - something less than 128 kb in a dual port set-up.

A 400 MHz processor should be loafing - or waiting on I/O - most of the time! There is no reason that an average PC cannot run LPC-10.

The 300 bps is based upon the information rate of typical speech. My data comes from converations with Dr. Barry Morgan and several other researchers/practitioners. Barry was our Cryptographer - and an advisor and crypto algorithm developer working with NSA.

Speech is highly redundant. If we can "peel away" the redundancies - while retaining all of the nuances - we can reach the theoretical limit. (Of course, the devil is in the details!)

I would be VERY interested in the work done in the 70's to get to 300 bps. Nine processors then would equate to about 4 wrist watches - or one school calculator - today! It would be interesting to see if my ideas for getting there match the methods - or any sub-set of them.

I am aware of algorithms that can get there - but they use phonem look-up tables (brain - that spelling doesn't look right - but it's late) and have ZERO speaker characteristics. The goal is to get there WITH full speaker recognition.

LPC-10 is NOT the best way to get there. It is ONE way to get to a relatively low bit rate. It tends to fall apart rather rapidly below 2000 bps. (The TSP algorithm is still pretty decent down to almost 1200 bps -- it noticeably degrades at around 1400 bps.)

When some fringe clown comes in and calls me "The fable guy" - essentially, a liar - that IS an outburst. Of course, the fact that he lacks the manhood to apologize - or even to offer rational thought - proves more than any words could convey.

As a matter of fact, I work in the wireless industry - for a company with over 50% market share of cell phone chips. The *worldwide* trend in "cellular" (including GSM) is to use *wider* bandwidth - not narrow band. The 3G system is rather wideband. It can carry full-motion video PLUS stereo audio, as was demonstrated at a recent show in Las Vegas. High data rates are the norm - not the exception. The critical factors: How much power does it need? How much reals estate does it occupy? How much "sex appeal" does it have? How can it set my customer's company apart?

There is essentially zero desire for low-bit rate voice for 3G. For older systems, the CELP (Code Exited Linear Prediction) algorithm was less computationally intensive than LPC (of which it is a form), and gives reasonable results at 4.8 kbps. A good CELP algorithm is really, really NICE at 9.6kbps - and compares quite well with ADPCM at 7.2 kbps.

As I said, "Nine processors then would equate to about 4 wrist watches - or one school calculator - today!"

 

Here are facts from a long history of digital voice usage on HF, VHF, UHF, microwave, and SATCOM.

Digital signals can occupy less space than an analog signal carrying the same information. They can also occupy more. The system is designed to interface with the medium. A 2500 Hz bandwidth is quite adequate for HF, and can support 2400 bps digital voice with built-in error detection and correction.

As bandwidth drops, the noise within the passband also drops. Also, coherent detection may be used to pull signals from below the noise. In essence, knowing the arrival time allows the receiver to correllate, and accept only the good signal.

Codes are available to permit recovery of correct information even in the face of significant loss of bits. These codes greatly reduce "hearable" errors.

We successfully achieved crypto synchronization in bit error rates (BER) of 10^-2, and usable voice in bit error rates (BER) of 10^-3. With appropriate modem algorithms - or post-processing - or internal detection and correction, the digital voice device or software can operate successfully in a VERY noisy channel.

The goal is to REDUCE the bit rate. Reduced bit rate means reduced bandwidth and slower baud rate, which means lower error rates. It also means lower noise, which means our minimum detectable signal is lower - meaning greater range.

I have not seen a range reduction on HF, and digital has some factors that permit BETTER weak signal reception. Your statement about the "superiority" of SSB in weak signal environments cannot be supported by any technical data - but it can be disproved.

The comparison with public safety systems is like comparing a tractor with a pick-up. Both do their jobs well - and both have vastly different jobs.

Bandwidth is NOT an issue in public safety. Interoperability IS an issue. We CAN use less than 3 kHz on VHF/UHF - but it interoperates with nothing! (In point of fact, the first CVSD DES encrypted digital voice on VHF/UHF FM required TWO channels - and REDUCED range by a factor of 2 to 3.) At 4.8 or 9.6 kbps with a modern radio, encrypted digital voice has no impact on range - because it operates in a single channel.

Trunked or non-trunked systems have equal range for equal power and antenna height. Trunking neither adds nor subtracts range.

Finally, the power issue is bogus for a number of reasons: 1) Higher AVERAGE power in a MORE NARROW bandwidth means greater SNR. 2) Lowered transmitter output power DOES NOT mean lower signal strength for the digital signal versus analog voice - in fact, if one reduces the power by 75%, the AVERAGE level of the digital signal is about equal to the AVERAGE level of the analog voice signal. 3) Coherent detection allows reception of digital signals below the noise. 4) Error detection and correction may be built into the algorithm, permitting efficient communication over very bad channels.

And, again, the transmitter DOES NOT have to be SSB - but it CAN be SSB. (Cooling fans are NOT that expensive, and thermal engineering - or re-engineering - is not rocket science.)

As my Granddaddy used to say: "'Can't' never did anything." Anyone who is unwilling to learn is not a model for any child. If you are unwilling to TRY a mode, HOW can you know whether it is superior or inferior?

Lead, follow, or get out of the way! The train HAS left the station. We can all be passengers, engineers, facilitators, watchers - or the cow on the tracks.

We DO NOT want ALL of the HF spectrum to be digital - until that is the ONLY mode (hopefully in the FAR distant future). We DO desperately NEED to move TOWARD digital - and begin to make the advances hams are famous for doing. We CAN'T do it sitting on our butts!

 

Observations by a "digital dummy".

I have been interested in swapping pictures on HF radio for quite some time. Although I rarely transmit, I do quite a bit of listening. I've enjoyed analog SSTV using MMSSTV and I've enjoyed watching the new digital picture/file modes work (e.g. WinDrm, etc). Here's what I've observed. The "digital" file transfer modes are easily affected by QRM, QRN, and QSB. If the signal on the receiving end is not very close to S9 or better and "in the clear", errors occur and an imcomplete file/picture is received ... requiring "bad block" requests and retransmission of the bad blocks. In comparison and under like conditions, with analog SSTV you get "something" even if it's not "closed circuit" quality. Which is better? I guess that depends on the quality desired. With digital it appears to be a matter of all or nothing. I must say that the picture quality of digital is far better when compared to analog.
I've also observed that folks running digital modes always make contact via voice (i.e. SSB) before exchanging files. While the analog folks can send a CQ by transmitting a picture with CQ imprinted on it. Don't know if this is possible with digital but, I don't recall ever seeing it done.
Now, it appears to me that digital voice QSOs would need to be initiated using an analog method (e.g. SSB, CW, etc) or perhaps a prearranged schedule. It will be adversely affected by QRM, QRN, and QSB and propagation. In which case, SSB would be superior to digital when it comes to "making the contact" (e.g CW vs. SSB).
These are just my ignorant and unscientific observations ... which leave me wondering what all of the fuss is about? I can't see how anyone would consider using digital voice in lieu of SSB when it appears to me to be the inferior mode. I think it is a passing fad ... a new toy to spend money on that we'll all look back on and laugh at. OR ... is it, as someone mentioned, an outgrowth from the "broadcast quality" audio crowd to attempt to improve their audio even further?

 

I think Jim has kept the focus of this thread, which is that "expansion" of the frequencies needed for "development" of new digital modes is NOT needed (yet!).

I think we all agree that the world is slowly going digital (just try to buy a new vinyl record!).

However, the current digital voice implementation simply does not meet the needs of the amateur radio environment. That is not to say it will NEVER meet those needs, but that it currently does not. I believe you cannot breakin (and I am not even sure about roundtables), it drops out suddenly, not fades out like analog SSB, and has problems with QRN and QRM. The jury is still out on whether or not digital voice will ever displace analog voice, but, of course, it probably will, but probably not in the early implementation we see today. Rule changes are just not needed, or safe (yet!).

There is nothing wrong with postulating whether digital or analog will win out - and if it is analog, it is simply a matter of opinion, and should not be criticized as being anti-progress.

The point that Jim makes is that it is premature to alter all the rules with the expectation that digital voice will replace analog voice, until it is happening on the merits of digital voice alone. In their attempt to take away spectrum from analog users and dedicate it to promote digital development, some overlook the fact that the FCC already provides a STA, or Special Temporary Authority, to permit experimental use of new modes until they prove worthy of a permanent place in the amateur spectrum, or are worthy of sharing places with other modes.

Digital voice will stand on its own merits the way SSB did when it displaced AM with a huge 9db power advantage, half the bandwidth, and elimination of annoying hetrodynes. Until digital voice also provides more benefits other than just studio quality audio, it SHOULD be relegated to experimental status. The current ARRL proposed petition to prematurely change all the rules to make it easy to bypass the STA (and the small justification needed to get the STA) is just not in our best interest. Suppose spread-spectrum on HF were allowed under new rules and turned out to be disasterous to the hobby. Would that have been a huge mistake? Of course (because it turned out to be disasterous), and it could have been averted by keeping to the STA process to first PROVE it was a good idea, and get feedback from the radio amateur community.

This thread was started to question the need for "enhancing" Winlink 2000 so large email attachments, such as pictures, could be transmitted like they are on the Internet. Well, after more study, it turns out that even Pactor-3, which uses five times the bandwidth of Pactor-2, but provides only 30% or less faster text email transfer, would waste five times the bandwidth of Pactor-2, even when only text messages were being sent! Winlink tries to sell us on this "need" in the name of "modernization" and "progress" and not inhibiting new developments, but just duplicating the Internet capabilities for the benefit of a few users, and using five times the bandwidth ALL THE TIME, just to allow large attachemets SOME OF THE TIME is hardly "progress"! Progress might be coming up with a protocol that did not need five times the bandwidth to send that attachment, not just using more spectrum (which is limited anyway) to do the job.

If "progress" is desired through the development of new digital modes, the STA process makes it possible with very little delay or hassle - usually only a letter to the FCC justifying the reasons a STA should be issued.

We are granted a very limited spectrum by the FCC, to be used, and fairly shared by 660,000 FCC-licensed hams for all interests, analog and digital, and "progress" comes in using that limited spectrum in ways to accomodate more users in the same space, not in ways that reduce the number of users that can be accomodated.

 

wa5ben:
"LPC-10 can be implemented quite easily in a single DSP chip - probably no more powerful than the DSP chip in a modern sound card. You need only model the length and diameter of 5 pipes to model the human vocal tract. Linear Predictive Coding of 5 pipes with 2 coefficients each = LPC-10. The other two basic issues are extracting the fundamental pitch (80 Hz to around 200 Hz in most males, up to a maximum of 400 Hz for females), and determining voiced (vowel) or unvoiced (consonant) sounds."

I've heard this "easily" and "cheaply" music before. I'm still waiting for a parts list that would let a home experimenter play with a DSP/DDS transmitter cheaply. Can you give me a parts list for a DSP chip that would let me experiment with LPC-10 coding?

wa5ben:
"BTW: A 400 MHz Pentium can be had for $100.00 or less - if you can find a PC that slow."

Maybe you have infinite desk space, I don't. I don't want to plug a whole PC and monitor into my mic jack to get digital voice.

wa5ben:
" You can probably get at least 800 MHz for $100.00 now."

Maybe you can. Not where I live. I can't even find any advertised in the latest Nuts & Volts magazine.

"Sound cards are $5.00 to about $200.00 new (though I cannot fathom why anyone would spend $200.00 for ANY sound card.) A $20.00 sound card should be adequate for LPC-10."

It might be adequate to encode the speech but it won't do the compression and it won't do simultaneous output to the microphone in a format capable of transmission over HF frequencies.

I don't think this will be nearly as cheap and easy as you make it out to be. So far we need a dedicated 400Mhz PC and at least two sound cards, assuming you can find a HF transmission protocol you can interface the output of the compression software into. Since all of the lpc-10 software routines I can find on the internet indicate at least a 2400bps data stream is needed (and most indicate a 4800bps stream is needed for adequate quality voice) my guess is that a Pactor III setup would be needed at the very least, sound card generated tones through a microphone just won't cut it. So far, we are up to about $1200 by my calculation.

Can you give me a parts list to do it any cheaper?

wa5ben:
"Our "high powered" processor was a 16 bit operating at something on the order of 20 MHz, with a TRW multiplier-accumulator to pull the coefficients. If I remember correctly, we had "lots" of memory - something less than 128 kb in a dual port set-up."

In other words, you had dedicated hardware. That tends to be more efficient.

wa5ben:
"A 400 MHz processor should be loafing - or waiting on I/O - most of the time! There is no reason that an average PC cannot run LPC-10."

Not according to any site I can find on the internet. If you are calling an average PC one that meets today's standards, I might agree. But the one I have is tied up doing other things. I just don't want or need a second one anywhere around the operating position.

wa5ben:
"The 300 bps is based upon the information rate of typical speech. My data comes from converations with Dr. Barry Morgan and several other researchers/practitioners. Barry was our Cryptographer - and an advisor and crypto algorithm developer working with NSA.

Speech is highly redundant. If we can "peel away" the redundancies - while retaining all of the nuances - we can reach the theoretical limit. (Of course, the devil is in the details!)"

Almost everything I can find on the internet indicates the information rate of typical speech varies from 100 to 500 words per minute. I suppose with the proper assumptions you could convert this to bits per second. None of the documentation I can find on this, however, indicates that you would still keep speaker recognition at this rate.

"I would be VERY interested in the work done in the 70's to get to 300 bps. Nine processors then would equate to about 4 wrist watches - or one school calculator - today! It would be interesting to see if my ideas for getting there match the methods - or any sub-set of them."

Remember, he said 300 *baud*. You don't know the bit rate till you know the transmission scheme used. 300 baud could be anything from 300 bps to 4800 bps.

I did some work in the very late 70's and early 80's with Z80 processors doing vt100 emulators. Nine Z80 processors would have been a LOT of processing power. A packet TNC of that era would have used a single Z80 and in the used market they still run $100 to $200 even today. Something with 9 of them would probably run $500 even used. Still a hefty price for digital voice.

Ask AOR why their digital voice box is so high priced.

tim ab0wr

 

There are many, many factors to consider when designing an HF modem and an HF protocol. I have done both. (I have designed HF tactical modems, and am soley responsible for the concept that became known as G-TOR.)

These factors are critical:

The tone spacings/phasing must be optimized for the baud (not bit) rate

Error detection and correction coding must be incorporated into the protocol. This means the receiver must be capable of correcting reasonable errors without requesting re-transmission.

Memory ARQ is required for point-to point links. This means that the receiver throws away nothing. It keeps up to three copies of an errored block, and uses the two or three copies to build one good block. This allows "partial block" reception, reducing the necessity for re-transmission.

A Forward Error Correction "broadcast" mode is required for "point to multi-point" links (i.e., nets). This means that (normally) two or (sometimes) three copies of the information are sent displaced from each other in time. They are used (along with the error detection and correction coding) to build a correct copy.

Proper speed control is required. There must be a threshhold for speed up and speed down. The receiver must tell the transmitter (in point - to - point links) its error rate at each turn-around. Link must be automatically adjusted based upon the threshhold.

Recognition and negotiation of capabilities is required for point - to point links. The first transmission from an originating station (ISS) must state which capabilities it has (and its preferences). The first reply from the called station must state the desired settings, based upon the ISS capabilities. The second transmission from an originating station must state the parameters to be used and a training mesage to allow the called station determine link capabilities. The reply from the called station must state error rate and fall up/fall down.

Block sizes must be optimized to the link capabilities. Using the same error rate information used for speed, the block size must be adjusted rapidly upward for good links and modestly incremented downward for poor links.

Data compression must be designed in. It is not an optional requirement.

In short, the HF modem must be designed to work WITH the Baud rate, WITH the protocol, and WITH the link. The protocol must be designed to adjust for actual link performance. The data must be properly packaged for the actual link performance.

If these principals are followed, QRN/QRN/QSB will have minimal impact on link performance. If they are not followed, ANY factor will impact the link.

That sounds like HF packet. Using Bell 103 tones with a 200 Hz shift at 300 bps/300 Baud (bps and Baud are the same for FSK) is just plain stupid! If you widen the tone spacing to 600 Hz, you will improve the error rate by at least a factor of 5, and probably run mostly error free on any reasonable (6 dB SNR) channel. The throughput will rise dramatically. (This reflects actual results obtained on the Rockwell/Collins HF simulator at Cedar Rapids.)

SSTV users normally use voice to initiate a transfer. It is not required, and it is done in no other mode.

There is no advantage of "SSB voice" over digital voice. There is a realizable communication advantage for a digital voice signal over the analog signal.

 

It's called a sound card.

1. Why would you need a monitor on a dedicated digital voice computer? Or a keyboard?

2. You don't plug it into the mic jack - you put it through a modem.

Computers of that (lack of) power are thrown away every day. People don't pay to advertise junk. Look at flea markets or garage sales.

The entire LPC-10 algorithm should run on a single sound card. That IS the "compression".

1. You do not need a dedicated computer. --- $0.00

2. One sound card for the LPC-10 -- $20.00

3. One DSP for the modem. -- Full manufacturers Development kits - with DSP and PCB - start at about $200.00.

4. Open Source modem / DSP code. $0.00

In REALITY BASED words, we had less than the processing power of an 12 Mhz 80286 / 80287 combination.

Why on earth would you need a second PC?

1. Absolutely NOTHING related to digital voice requires nine processors.

2. I have designed on the Z-80 platform. Paying $100.00 for a Z-80 based design on the used market makes no sense.

3. Again, NOTHING related to digital voice requires NINE of today's processors.

 

wa5ben:
"The goal is to REDUCE the bit rate.  Reduced bit rate means reduced bandwidth and slower baud rate, which means lower error rates.  It also means lower noise, which means our minimum detectable signal is lower - meaning greater range."

As I've found in my recent research, this also causes bigger holes in the voice when an error DOES occur. That causes problems all of its own.

You are also making the assumption that there is or will be equipment available very soon that can encode higher amounts of information in the same or lower number of bits that we use today.

I can find nothing that indicates there is anything on the horizon that will allow this. The only other option is to decrease the bandwidth used for the same number of bits which, according to the Hartley-Shannon law, requires an increased signal-to-noise ratio because the reciever has to distinguish between an increased number of transmission states. My guess is that you will find that the increase in (S+N)/N is more than the decrease in N you will get from decreased bandwidth.

wa5ben:
"I have not seen a range reduction on HF, and digital has some factors that permit BETTER weak signal reception.  Your statement about the "superiority" of SSB in weak signal environments cannot be supported by any technical data - but it can be disproved."

You apparently have not read the QST review of the AOR digital voice equipment - which is about the only option available to amateurs. It has a much worse signal-to-noise ratio than SSB.

wa5ben:
"Finally, the power issue is bogus for a number of reasons:  1) Higher AVERAGE power in a MORE NARROW bandwidth means greater SNR.  2) Lowered transmitter output power DOES NOT mean lower signal strength for the digital signal versus analog voice - in fact, if one reduces the power by 75%, the AVERAGE level of the digital signal is about equal to the AVERAGE level of the analog voice signal. "

First, the AOR modem uses the same bandwidth as a voice signal. So there is no increase in (S+N)/N due to decreased bandwidth.

If you can point me to a commercially available digital radio peripheral affordable to the typical amateur radio operator, I will agree with you but I am not going to hold my breath. As usual, you are making fantastic claims for something that doesn't actually exist. If someone would just build a $100 linear running 1500 watts output, every ham could have one. They just don't exist, Larry.

As far as your statement No. 2 - exactly what do you think WA0LYK just said? You just repeated what he said!

wa5ben:
"And, again, the transmitter DOES NOT have to be SSB - but it CAN be SSB.  (Cooling fans are NOT that expensive, and thermal engineering - or re-engineering - is not rocket science.)"

For Pete's Sake, Larry, just how many hams are going to take their Icom 751a's or Kenwood TS-570's or or Yasue's apart, add bigger heat sinks and bigger cooling fans.

As usual, you are making claims that just don't bear any resemblance to reality. Buying a heat sink that will mount on the back of a TS-570 or an Icom 706 that will dissapate 4 times as much heat, even with a bigger cooling fan just isn't realistic.


tim ab0wr

 

wa5ben:
"It's called a sound card."

Really? A sound card can become a SSB transmitter? Exactly what frequency IF will the sound card put out?

wa5ben:
"1. Why would you need a monitor on a dedicated digital voice computer? Or a keyboard?"

Larry, *YOU* are the one that suggested getting a used computer for $100. wa5ben:"BTW: A 400 MHz Pentium can be had for $100.00 or less - if you can find a PC that slow."

Exactly what kind of computer were you thinking of?

All 486/586/686 motherboards today that I am aware of require a keyboard to be attached to finish the boot process. If you don't have a monitor hooked up you won't even know the boot process didn't complete!

If you are running a linux operating system of course you can access it remotely. That doesn't work so well with Microsoft products. If you are running Windows you need a monitor to be able to tell when something goes wrong!

As usual, you seem to be somewhat disconnected from the real world.

wa5ben:
"2. You don't plug it into the mic jack - you put it through a modem."

Right. And which modem would that be? I don't believe my Timewave or MFJ TNC's will work with compressed audio as inputs. And I'm not sure how to interface compressed audio to the telephone modems I have.

As usual, you seem to be somewhat disconnected from the real world.

wa5ben:
"Computers of that (lack of) power are thrown away every day. People don't pay to advertise junk. Look at flea markets or garage sales."

Your point is? I already said I don't have the space for another computer at my station.

wa5ben:
"The entire LPC-10 algorithm should run on a single sound card. That IS the "compression"."

Can you point me to a place on the web with a working copy of software that runs the encoding and compression in the sound card? I can't find one

 

wa5ben:
"It's called a sound card."

Really? And what IF frequency does this sound card transmit SSB at?

wa5ben:
"1.  Why would you need a monitor on a dedicated digital voice computer?  Or a keyboard?"

I guess you are unaware that almost all 486/586/686/pentium/amd motherboards require a keyboard to be attached to finish the motherboard bootup? If you don't have a monitor attached you won't even know the bootup is hung. If you are running linux you don't need a monitor - but you do if you are running a Windows product. Otherwise you'll never be able to tell if something has gone wrong.

Your disconnect from reality is showing again.

wa5ben:
"2.  You don't plug it into the mic jack - you put it through a modem."

Right. The only problem is that my Timewave and MFJ TNC's won't take encoded speech as inputs and I'm not sure how to interface the encoded speech to my telephone modem. So the same problem exists.

Not everyone has such modems laying around, Larry. Your disconnect from reality is showing again.

And just where does the modem plug in, Larry.

wa5ben:
"Computers of that (lack of) power are thrown away every day.  People don't pay to advertise junk.  Look at flea markets or garage sales."

I already told you I don't have space for another computer at the station. Your disconnect from reality is showing again.

wa5ben:
"The entire LPC-10 algorithm should run on a single sound card.  That IS the "compression"."

Perhaps you can point me to someplace on the web that has this software running in the sound card. All the ones I could find use the sound card for encoding but then use a software algorithm running in the processor/floating point coprocessor to do the compression.

wa5ben:
"1.  You do not need a dedicated computer.  --- $0.00"

It's nice to know that you can tell from such a distance what I have sitting here, that you can tell what it is running, and that you can tell what capability the PC has.

FWIW, the computer I have is loaded up running as a Winlink Classic hub over HF and VHF (which means it also has a serial port connected to a HF transceiver), as a PIC programming station, as a printer hub, and as the X-window station for three linux computers located throughout the house. It doesn't have any capacity to do anything else.

So sorry, but a dedicated processor of some kind will be needed.

wa5ben:
"2.  One sound card for the LPC-10 -- $20.00"

Well, I'll check and see what is available but the last time I was at the computer store a $20 sound card wouldn't even handle duplex voip. But I'll take your word for this.

wa5ben:
"3.  One DSP for the modem.  -- Full manufacturers Development kits - with DSP and PCB - start at about $200.00."

Ok, so far we are up to about $320, assuming I can find space for another PC.

wa5ben:
"4.  Open Source modem / DSP code.  $0.00"

BTW, even if my PC had the processing power, it doesn't have enough slots. You'll need a dedicated PC just to get enough slots and interrupts.

wa5ben:
:Why on earth would you need a second PC?"

Because some of us live in the real world.

So we are now up to about $350 (including connecting cables, etc) for a set of equipment to let us do digital voice. For that kind of money I can buy a spare Icom 751a or Kenwood ts-430 and put it in my car to do mobile SSB.

Perhaps your opportunity cost is such that spending $350 on a peripheral is a valid choice. It's not for me. And I would venture to say that you aren't going to find a lot of hams shelling out $350 for a setup that they have to program, put in a case, and basically manufacture just so they can experiment with digital voice. *AND* then hope they can find someone running compatible software on the other end (slim and none).

Get your cost down into the $75 range and it will be much more likely to become a viable product on the amateur bands.

$350 costs is one of the big reason I don't see digital voice overtaking SSB for a long time - not when a good used SSB transceiver costs just about the same!

wa5ben:
:In REALITY BASED words, we had less than the processing power of an 12 Mhz 80286 / 80287 combination."

Well, Larry, not a single web site I could find with LPC-10 recommended less than a 400mhz PC. So more power to you. Perhaps you need to bring out a commercial product - you could make a killing if you could do it so inexpensively where no one else can.

wa5ben:
"2.  I have designed on the Z-80 platform.  Paying $100.00 for a Z-80 based design on the used market makes no sense."

Welcome to the real world, Larry. Not all of us have circuit board manufacturing facilities that will let us design and build our own TNC's. We have to live with what is available. Go look on the qrz and eham seller pages and see what used Z80 based TNC's are going for.

wa5ben:
"3.  Again, NOTHING related to digital voice requires NINE of today's processors."

ROFL!!! Nobody said anything about NINE of today's processors. Go back and read the message again.

tim ab0wr

 

wa5ben:
"There is no advantage of "SSB voice" over digital voice.  There is a realizable communication advantage for a digital voice signal over the analog signal. "

There is a huge advantage, Larry. One that digital advocates always seem to forget, rather conveniently in my opinion.

Your own claims for digital modes being better able to pick signals out of the air comes back to haunt you when spectrum efficiency is actually laid out for the modes.

Call it propagation stacking or whatever you want, but at 2330z this evening on 3920khz there were three different SSB conversations going on that could be heard here in the central US. A ragchew in the SE United States, a SSB traffic net in the Central Plains, and something going on in the sourthern Mountain states ( I didn't listen closely enough to tell what it was).

All three were going on at the same time and were doing so successfully. As far as spectrum efficiency goes there were at least SIX amateurs using the frequency at any one time (and actually more when you consider the ragchew roundtable and the net participants).

This is possible because of the low power density SSB has because of analog voice characteristics.

If this were attempted with digital voice or data, the spectrum efficiency would actually go down because of the mutual interference the digital signals with higher average power would incur. You MIGHT get two conversation going on at once. More than likely you would only get ONE.

So your spectrum efficiency just went from 3 for analog voice to perhaps 2 for digital voice but, more likely, 1.

I have yet to find anyone who can dispute this. The only digital mode that I have found that would allow the same efficiency is Packet. Packet was designed to allow sharing a frequency. None of the other digital modes that I have tried have that capability.

This should have been part of the studies the ARRL performed before coming up with a new bandplan based on the "efficiencies" of digital. When digital voice gets to where it can routinely operate in a bandwidth of less than 900hz, then perhaps it will reach the same efficiency level. I wonder how long that will be?

tim ab0wr

 

You missed my whole point! I said "Maybe these are reasons to investigate digital voice on the ham bands. However, I suspect it will be found that digital voice using SSB is inferior to analog voice SSB for weak signal work and for operation in interference laden, both atmospheric and adjacent channel, spectrum."

There are only two digital voice systems available to the amateur radio operator right now, AOR's and DRM. I don't know what is coming but I haven't seen any announcements for new implementations by any manufacturer. Both of these implementations require a significantly higher SNR than SSB. Neither have been proven to operate well in an interference laden environment where signals overlap each other.

Your comments do not address the issue I stated. All these systems operate in an environment where there are discreet channels for carrying the signal, i.e. they are channelized, and adjacent channel interference is minimal. Trying to say that HF amateur operations are "falling behind" in digital voice implementations ignores the difference in operational characteristics. Amateur operators on HF DO HAVE to contend with adjacent channel interference, lots of it. Commercial applications and their success do not translate well into our operations. Do you know of any HF implementations of digital voice that have been implemented in an environment that has the adjacent channel interference levels of 20m or 40m phone and are successful? Will digital voice implementations work when digital signals overlap each other? I suspect not very well. Consequently my conclusion was "I hope the folks that insist digital voice MUST be deployed in the ham bands don't also begin to insist that we must channelize and move to FM. HF just doesn't have the spectrum to allow for this without REDUCING THE NUMBER OF HAMS INSTEAD OF INCREASING THEM." While the FM part may not come to pass, I suspect one thing the digital at all costs folks will ask for is CHANNELIZATION of the HF bands so they can operate in an environment that is free of adjacent channel interference.

You again ignore my point and go your merry way saying I don't know what I am talking about. Here is what I said "Another problem with digital voice that a lot of hams aren't aware of is the need for reducing the power output of amateur class equipment. " I'll also bet we will see a lot of transceivers and amplifiers on ebay and other sales forums with blown finals, power supplies, and tubes because of digital voice."

Do you see anything in here about SNR? Do you see anything in here about detection or error correction? My only point was that digital voice implementaions will stress transmitters more than analog voice. I only have to look at PSK31 and the number of hams that don't understand the need to reduce power levels to know that there are a lot of hams that won't do so with digital voice implementations either.

Your reply

is a real great solution after the final transistors or tubes have blown!

No one here is saying digital shouldn't be allowed on HF. What we are trying to point out are some of the pitfalls that digital voice has and don't be surprised if analog SSB may continue to be the better mode for quite some time to come!

Jim
WA0LYK