Just say NO! to Winlink expansion!

wa5ben:
"The real embarassment comes from those who deliberately distort every fact to twist the meaning into something that supports their limited and obvious agenda.
PACTOR III is NOT the only HF modem protocol.  ARQ is NOT the only available mode.  You know that.  Why do you insist upon continuing your nonsense?"

Have you USED Winlink, Larry? How do you know they use protocols and modes other than Pactor and ARQ? What other protocols and modes do you think they use?

wa5ben:
"Even for modified ARQ (e.g., GTOR, PACTOR, CLOVER), once the set-up is done (< 1 second), several protocols permit < 15% overhead."

ROFL. Larry, I have a 24/7 Pactor1/Winlink2.9 station on the air as part of the NTSD system. I can assure you that setup takes much longer than 1 second to accomplish. This is straight experience from sitting in front of the modem and tranceiver for many hours watching to see how things are going. It takes longer than a second for the two ends to even negotiate bumping speed from 100baud ascii to 100baud compressed. And, god forbid, some qrm comes on and you have to negotiate a downward bump in speed. It generally takes 5-10 seconds for the stations to even figure out a downward bump is needed, let alone get it negotiated and actual data to get started again. I can't tell you how many times it would have been faster for the message to be sent using 100baud ascii rather than the continual trying to bump up the speed to a higher value.

wa5ben:
"Simply put:   You have only one line, and it is false.  (One wonders why you invest so much energy in being negative and spreading false stories, and ZERO into looking after the future of ham radio.)"

Skip is using actual data from the Winlink site. You obviously see something wrong with the system operation because you believe the data can't possibly be correct. Well, THAT'S WHAT WE'VE BEEN SAYING!!!! Something is wrong with the system and that's what the data is showing!!!

Thanks very much for agreeing with us. Perhaps you can point this out to the Winlink system people - something is wrong with their system because the data shows it!!

tim ab0wr

 

first off, I use winlink2000 only on vhf/uhf as one of many message systems I use so I know winlink2000 and it's delays.

I have seen logs from the winlink2000 pmbos and know it has connection times as wel as the size of the messages transfered and the mode used.
So to put the whole debate of the correcness of the data to rest maybe some sysop is willing to provide us with a log of a week or so, that should give acurate data of the troughput once connection is established, the handshake time is probably known because that is the same nomatter what software is used.

73 de Andre PE1RDW

 

First, if you had "full duplex" on an HF radio, then you were using two HF frequencies simultaneously, each at 2400 or 3600 bps. That's quite a bit of spectrum, more than 3 kHz I suspect. However, that's just being a bit picky.

Secondly, near toll quality is pretty subjective. You don't state all the details surrounding the transmissions they use like SNR, power levels, frequencies, etc. Not knowing all the details of what the military has all I can go by is what commercial quality transmissions are using today.

Some GSM systems today use a bit rate of around 30 kps. Here are some rates I found on the internet and their opinion ratings of sound quality (1 - unsatisfactory, 5 - excellent)

System Codec Codec FEC Mean
Type Type Rate Rate Opinion
Kbps Kbps Score
(1 - 5)

Cellular
TDMA
Full Rate RPE-LTP 13 9.8 3.7

Cellular
TDMA
Half-Rate VSELP 5.6 5.8 3.6

Cellular TIA/EIA
CDMA IS-661 7.2 3.2 4

Dispatch
Project 25 IMBE 4.4 2.8 3.4

Dispatch
DIMRS
Canada VSELP 4.2 3.2 -

Digital
Cordless
Phones ADPCM 32 0 4

The lowest rate listed is a total of 7.2 Kbps. None are close to the 2.4 Kbps you say the military uses.

Jim
WA0LYK

 

Jim,

I worked with people doing voice at 2400bps in early 80's. The general concensus was that it could be understood, but there was no way to recognize the speaker to verify whether information content was valid. This work was done at M.I.T. Lincoln Laboratory as part of a project to develop a 2400bps voice link for military applications.

Ken

 

Again, you choose statements out of context.
Then you put 1 + 7 + 3 and claim that it equals 4.

I have seen no proposal to assign 20 kHz segments of 20 meters to any such service. It is, however, a viable option - and a legitimate usage - on 10 meters. We desperately need more usage of 10 meters.

The tenager's suggestion for interactive gaming (or any other networked activity) DOES NOT have to be on the internet. It just needs common access for all of the users. That can be a radio channel, a pair of wires, or fiber.

The statement about internet access concerns traffic when connected through an RF to internet connection - as would be required during emergency communications - and when no other channel was available. It applies to VHF, UHF, and microwave operations as well as HF operations.

1 + 7 + 3 does not equal 4.

 

While I cannot discuss actual frequencies used in any specific operation, I can say that we were within a range of 4 MHz to 18 MHz, with approximately 1 MHz frequency separation.

Typically, at the fixed end, we had a split site with a microwave link. At the other end, we used a mobile or portable installation. Power levels ranged between 1,000 Watts to 10,000 Watts (PEP) on the fixed end and normally ranged between 100 Watts and 400 Watts (PEP) on the mobile or portable end. (At least one aircraft could transmit 1,000 Watts PEP.)

In no case did transmit bandwidth exceed 2.8 kHz, and the primary energy was within a 2.5 kHz band (later verified by spectrum analyzer looking at the TX output).

Normally, a fairly wide (100 kHz to 300 kHz) pre-filter is used on the mobile or portable receiver and a post-filter (100 kHz to 300 kHz bandwidth) is used on the mobile or portable transmitter.

The filters are to eliminate the phase noise present in all transmitters. This noise is very strong at close proximity to any transmitter, with no modulation of the transmitter. (Frequently, the transmitter and receiver must also share an antenna via a duplexer.)

I was successful with a pair of separated dipoles and neither a pre-filter nor a post-filter in one instance - using a 400 Watt Russian transmitter.

"Near toll quality" implies that noise levels approach what a user would experience on a good telephone line, that the listener can correctly identify each of several speakers with similar voices, and that intelligibility is similar to that found on a telephone line.

The system was successfully used by native speakers of Chinese (meaning of words is highly inflection dependent), Japanese, Hindi, Urdu, Arabic, Hebrew, and several European languages. Repeats were never necessary and disgnostic rhyme tests were well into the upper portion of the 90% range.

The *commercial and military* system we were using was a vastly improved version of 2400 bps voice and a "superset" of LPC-10. It corrected most of the "buzz-saw" characteristics of LPC-10. (With LPC-10, speaker recognition is rather questionable.)

The primary problem with standard LPC 10 is that it samples and sends a frame, then samples and sends another frame. There is no transition between the frames, and the "smoothness" of voice is lost - as are the critical inflections. Our system did not suffer from these issues, as we used a mechanism to restore the "smoothness" and track inflections.

I have worked extensively with several types of digital voice in the 4800 bps to 64 kbps range, and am well acquainted with the methods used to test and pick out artifacts of the various algorithms. I am also familiar with the design trade-offs for algorithms.

There is MUCH that can be done to create a low-data rate, high-intelligibility digital voice algorithm with full toll-quality speaker recognition. The theoretical lower limit is 300 bps. I very firmly believe that a good and achievable target is 1200 bps for full specification, with the algorithm usable with some degradation down to 600 bps.

 

Your work sounds as if it significantly predated ours.

We had our system at AFCEA each year from 1984 or 1985 through at least 1989. We also had various personnel from the served agencies in our facility - and on tests to various sites. In addition, we had commercial customers on the equipment.

All were impressed by the speaker recognition not available on 1051/1015 equipment - and by the ability to offer a secure PBX link via an E&M interface. (I even put one on a plug-board PBX - with an E&M interface that I built from scratch in country!)

We also offered an 8 telephone secure PBX with intercom.

 

wa5ben:
"I have seen no proposal to assign 20 kHz segments of 20 meters to any such service. It is, however, a viable option - and a legitimate usage - on 10 meters. We desperately need more usage of 10 meters."

Just because you haven't seen it doesn't mean it doesn't exist. You've been given the URL for the proposal. I guess if you don't go look at it then it doesn't exist.

wa5ben:
"The tenager's suggestion for interactive gaming (or any other networked activity) DOES NOT have to be on the internet. It just needs common access for all of the users. That can be a radio channel, a pair of wires, or fiber."

Your reading ability is showing again. Go take another look at who was giving the talk. Then tell us all how Everquest works. Have YOU played it?

wa5ben:
"The statement about internet access concerns traffic when connected through an RF to internet connection - as would be required during emergency communications - and when no other channel was available. It applies to VHF, UHF, and microwave operations as well as HF operations."

And your point is?

BTW, playing Everquest over an RF to Internet connection is NOT "emergency communications" - even if no other channel is available.

"1 + 7 + 3 does not equal 4."

Since you obviously don't read the posts on here it is no wonder you can't get the numbers right.

tim ab0wr

 

Now, add spectral noise to a 4.8kb digital voice signal. Now you're back to sounding like "Darth Vader". We called this "digital pipe dream bull$#it" . And how much spectrum will you need for one 4.8kb voice quality conversation. My guess is 5 to 6 kHz once the real world kicks in and you add enough error correction. What about lost range digital versus analog voice?

So don't do it! LOL.

There is one simple fact. Digital users account for about 5% of all use and Winlink less than 1%. You may have the greatest idea since sliced bread, but efforts to supplant voice and CW with digital signals is not supported by the use numbers.

And if anyone had ANY doubts about Winlink's intentions to take over the bands, just see how they keep expanding their requests on an almost weekly basis.

Just say "no"!

And by the way, one of the reasons given for doing this band plan is so that new modes are easier to create. Didn't I hear somewhere that most of these already have STAs from the FCC? Isn't the STA process till pretty simple?

 

Charlie,

I'll restate for you again. The FCC has indicated that it feels our regulations are too restrictive and that they wish for us to take more responsibility for managing our spectrum.

Remeber, this is a proposal in the making. Get off the stick and send in constructive comments to the director in you division. If you feel that the status quo is preferred policy to the proposal on the table then it's your responsibility to make this known. Now, if you are not a member, then you will have to wait until the RM process begins to have your shot at this idea. Then you have the opportunity to exercise your democratic obligation to participate in the process.

Ken

 

wow 500 replies, you would almost think this is a hot topic

 

And here is the "guy" who is neither man enough to apologize nor man enough to be civil.

You, "sir", obviously lack any level of either knowledge or decency.

You have never used digital voice on HF, and have probably never used it even on a landline. Yet you post highly offensive messages that contradict what those of us with EXPERIENCE AND KNOWLEDGE are willing to share!

In actuality, modems capable of significantly greater than 4.8 kbps within a 3 kHz passband are readily available.

AND, since you are incapable of reading, my point was NOT about digital voice on HF at 4800 bps - it was about the LOWER LIMITS of practical digital voice. Currently, 2400 bps is quite feasible well within the bandwidth of a single SSB voice channel.

You STILL OWE ME AN APOLOGY for your last idiotic outburst. You now OWE ME ANOTHER.

 

wa5ben:
"AND, since you are incapable of reading, my point was NOT about digital voice on HF at 4800 bps - it was about the LOWER LIMITS of practical digital voice.  Currently, 2400 bps is quite feasible well within the bandwidth of a single SSB voice channel.  "

Everything I can find about the LPC-10 protocol you speak of, and there are several implementations of it freely available for use with VoIP, requires at least a 400Mhz Pentium PC with floating point co-processor in order to properly handle the calculations for full duplex operation - with faster processors very preferable.

Now, as a military contractor you may have been able to get either dedicated processors on one-off setups capable of handling this or you may have had access to other arrangements but it is very doubtful that very many hams are going to pay the kind of money necessary for a peripheral with this kind of processing power merely to do digital voice.

wa5ben:
"There is MUCH that can be done to create a low-data rate, high-intelligibility digital voice algorithm with full toll-quality speaker recognition.  The theoretical lower limit is 300 bps.  I very firmly believe that a good and achievable target is 1200 bps for full specification, with the algorithm usable with some degradation down to 600 bps."

I would be very interested in how you calculated this theoretical low limit for toll-quality speaker recognition. Could you please give us the mathematical relation you used to find this limit? Or give us a reference we can look up in the library or on the web?

I'm not going to hold my breath waiting for this since I died while holding my breath and waiting for the parts list from you for a cheap DSP and/or DDS transmitter. But I keep hoping!

BTW, the only research I can find using LPC-10 type coding to get to 1200bps required throwing away totally any error-correction and depending entirely upon error-free physical media for transmission. That's going to be pretty hard to do using RF as the media. You have to have error correcting somewhere and the minute you add it the *effective* bit rate goes right back up.

wa5ben:
"You STILL OWE ME AN APOLOGY for your last idiotic outburst.  You now OWE ME ANOTHER. "

I think you are the only one on here seeing outbursts. What is being pointed out to you is if the level of speech compression you are speaking of was commercially viable, the cellular companies would be abandoning existing systems by the trainload. I don't see that happening. Not only that, but  remember, the changeout from analog to digital came well AFTER the late-1980's. They had their chance to use your extremely narrow-band compression and apparently didn't find it usable. What's usable and affordable for the military doesn't always translate into usable and affordable for the commercial industry.

For the curious, here is another table of compression types and codec rates.

Compression Method  Company  Bit Rate (kbps)  MOS Score

MASC ACM Codec      Vianix    from 5.0          4.0

G.726 ADPCM        Lucent          32             3.85

G.728 LD-CELP      Lucent          16             3.61

G.729 CS-ACELP    Sherbrooke      8             3.92

G.729 x 2 Enc.       Lucent            8             3.27

G.729a CS-ACELP   Sherbrooke      8             3.7

G.723.1 ACELP       Sherbrooke      5.3           3.65

GSM HR                 ETSI               5.6           3.5

IS 54/136              Sherbrooke      7.95          3.5

GSM EFR                Sherbrooke     12.2           4.0

G.711 PCM              ---------         64             4.1

I don't see any getting down to even 4800bps let alone 2400bps or 1200bps. The only one that does better than toll quality requires 64kbps Perhaps Larry needs to hire on as an advisor to all these companies.

tim ab0wr

 

Tim,

In the 70's I worked with a brilliant mathematician who had just such an algorithm. It was implemented in hardware using 9 microprocessors and did, in fact, achieve voice recognition at 300 baud.

Now, I know that you will want the algorithm, details, schematics, etc. All you have to do is get the USN to declassify these items and they are yours.

Dennis KG4RUL

 

ab0wr:
"I would be very interested in how you calculated this theoretical low limit for toll-quality speaker recognition. Could you please give us the mathematical relation you used to find this limit? Or give us a reference we can look up in the library or on the web?"

kg4rul:
"In the 70's I worked with a brilliant mathematician who had just such an algorithm. "

So you are saying that the theoretical minimum is algorithm dependent? That's an interesting thing to know all by itself.

Did you and Larry work with the same algorithm? Guess not, since LPC-10 is not classified. Interesting you both come up with 300 (even if his is bit/sec and yours is baud).

Typically, it is a violation to even let the public know that classified material exists, let alone know what it is for. It's also illegal to give out any of the information in the classified material - of which letting me know that the algorithm encodes speech at 300baud would be part of the material.

Are you going to have to kill me now?

I would still like to know how he came up with 300 b/s as the theoretical minimum. I wonder if he will claim military secrecy as an excuse for not answering?

ab0wr:
" but it is very doubtful that very many hams are going to pay the kind of money necessary for a peripheral with this kind of processing power merely to do digital voice. "

kg4rul"
"It was implemented in hardware using 9 microprocessors and did, in fact, achieve voice recognition at 300 baud."

Btw, 300baud could relate to almost any bit rate. I would need to know the bit rate that is associated with the algorithm to know if, in fact, your classified algorithm is really very special.

But the use of 9 processors to do the encoding already tells me that it puts it outside the realm of a feasible retail product for ham radio. That makes it a very un-useful area for experimentation on the ham bands.


kg4rul:
"Now, I know that you will want the algorithm, details, schematics, etc. All you have to do is get the USN to declassify these items and they are yours."

Actually, I would need to know the actual bit rate to know whether it would be worth my time.

How about it, Dennis? What was the bit rate?

tim ab0wr