Will a system using 1000 non-regenerating repeaters have any chance of working?

[MichaelMontgomery]

I am having a debate with a colleague. He is engineering a system that uses 1000 non-regenerating repeaters. I am try to explain to him why this approach has a very high risk of failure, and why he should use regenerating repeaters instead. The protocol he wants to use does not lend itself to regenerating repeaters, and he does not want to change. I explained that since he must use regenerating repeaters, he will just have to accept that the protocol will change.

I am trying to muster support and arguments to show that any system with 1000 non-regenerating repeaters is doomed to failure.

I realize on this board that this it is like trying to muster arguments to prove that water is wet. But I would really appreciate it if as many people as possible would chime in on this topic, even if just to state the obvious: that a system with 1000 repeaters obviously must use regenerating repeaters to avoid noise and distortion stacking.

Thank you for your help, and my apologies for having to introduce such a topic!

 

[BradtheRad]

If a situation does not make sense...

Look for a reason that does not make sense.

Could your colleague have the hope that he will be called back, to install 1,000 workarounds? Thus guaranteeing job security? This is a motivator for countless people.

 

[MichaelMontgomery]

I think in this case it is more a matter of not wanting to reengineer the protocol, rather than wanting long term job security. I think he honestly believes that the 1000 non-regenerating repeaters will actually work. He measured the noise and distortion from a few prototypes in a lab environment, and scales that up to 1000 repeaters, and thinks it will work. This does not take into account field conditions, such as temperature, shock, and random noise. Although frankly, I think one might have trouble with 1000 non-regenerating repeaters even under ideal lab conditions.

So I am trying to muster arguments for why a system with 1000 non-regenerating repeaters is an unworkable plan, or it least get a body of support that it is indeed an unworkable plan.

 

[FvM]

I would prefer an exact problem description.

 

[betwixt]

This could be anything from a chain of 1,000 mechanical relays to a chain of 1,000 orbital satellites. Please be more specific.

Actually - mechanical relays probably would work, if you didn't mind waiting a while until the last one operates!

Brian.

 

[MichaelMontgomery]

I can give some information, but it would compromise my company to give too much.

We have up to 1000 linear communication links, each one with losses averaging 20 dB or so, but which could be more (up to double) under some circumstances. Currently, PPM is used to minimize power. The repeaters detect the PPM pulse, and generate an output pulse that corresponds to the detection time. Noise on each link is expected to be very low, almost inconsequential (except maybe when being cascaded 1000 times without regeneration, which is my main point). Channel distortion is normally low, but it is difficult to know how much field conditions will impact this. My concern is that even if noise or distortion affects the detection time by only 1 ns or so, with 1000 repeaters, the pulse could easily be pushed to the next detection slot (the PPM slots are only 80ns apart.)

I realize this is not a ton to go on, but what I am looking for is people who can state that this is a no-brainer: even with low noise and distortion, a good engineer just does not use 1000 repeaters with no regeneration anywhere in the chain, it is just too risky to let the noise and distortion from 1000 links add up.

 

[BradtheRad]

At the core, this issue may be technological, but to your colleague it will become more personal the more he is pressured.

Questions:

* Who is responsible for the success of the installation? If he is responsible, does he think he'll lose face by doing what one or more co-workers recommend?

* Does he have a track record of success most of the time? Does his track record outweigh yours? (Not that I know anything about either of your track records.)

* Does his proposal keep costs down? This may be the all-important criterion to the boss (whoever he may be). Your colleague may not let on to you that he has to follow orders.

* Technology advances. Is it possible he expects technical improvements to validate his approach?

* When a failure occurs, what is the most that will be lost? What is the minimum amount of material and labor that will be needed to fix it? He may see this as an acceptable scenario.

 

[MichaelMontgomery]

My colleague is actually working on the project. I was a reviewer.

He has a track record for success in the field of Electromagnetics, but not communications. I have a track record in communications, with over a dozen patents in the field of communications, and over 40 patents total.

His proposal does keep costs down, since a non-regenerating repeater is cheaper than a regenerating one.

A failure would be catestophic. If we go down this path, and it fails, the minimum we would lose is $10 million, but that does not even factor in the cost of lost opportunity. It is unlikely that anyone would be given a shot at a redo.

I think the fundamental problem is that he does not understand the risks. I was hoping that members of this board would help clarify the risks for him.

The risk is that his scheme depends on detecting the PPM slot with no more than 80ns signal skew (which would put it into an adjacent slot). I do not know of any real word system on any kind of links that can be depended upon to not stack up more than 80ns skew over 1000 non-regenerating repeaters.

To provide a bit more detail, each link is a coaxial cable with the cable and connectors giving 20 to 40 dB loss (over each link). The prototype test system in the lab shows less than 1 ns pulse skew over 10 repeaters, so he felts confident that that 1000 repeaters will give less than 80ns skew. I think that only Winnie the Poo exceeds his optimism. In the field, noise, temperture, shock, etc. will join forces to make a high risk of failure. I have no personal stake in this, only to save my company from wasting a lot of money, and my colleague from suffering the reputation hit from a failed project.

But I guess I myself have an unrealistic expectation of this board. Since I cannot give any priopriatary information, I cannot expect someone to definitively agree that 1000 non-regenerating repeaters has a high risk of more than 80 ns skew.

 

[BradtheRad]

cost of lost opportunity

I get the idea this is a prototype of something new.

It's better if a prototype is over-engineered so it will perform as claimed, and impress prospective buyers.
Or else the prototype may be 'pardoned as an experimental model', and you will 'get the bugs worked out' before delivery.

In a chain of 1000 units, it's a high likelihood there will be at least one weak link. A component whose performance is degenerating. A tarnished connector. Etc.

Your colleague is banking that a digital signal will always be stronger than the noise level.

What is really needed is a large-scale preliminary test. In the field. Maybe 50 or 100 repeaters. In bad weather, a lightningstorm, etc. In fact he may be planning to do this. He knows what is at stake.

What are the chances an interested customer will ask to see such a test?

Do you have a figure in mind, as to the number of repeaters that will fail to carry the signal? You might request that such test be done at that point. Prior to an all-out commitment of energy and material. Expect to be overruled, however.

Even if you show him a boatload of expert testimony, he will no doubt overrule you.

Until one day he will 'decide on his own' that the larger-scale test has to be done. You may have to watch and wait until the project is partially constructed. It has to be 'his idea.' People do this sort of thing all the time. They backtrack. It is human nature.

The day that (field) test is done, you can expect the results to confirm your expectations. Looking at it straight on, however, today the (lab) test confirms >his< expectations. For the moment, he is winning this dispute.

My aim is not to question your line of reasoning or your persuasiveness. My aim is only to make you feel better about a situation that you cannot prevent.

And granted you are correct...

Could a 'middle ground' design be found? Say, to use a limited number of regenerating repeaters?

Could he have 'something up his sleeve'? Now or in the future?

Could he decide to rent/buy an operating network that is presently in place?
Or to switch to fiber optic cables?
Etc.

Can there be an improvement no one has considered yet? Maybe better than any current plan?

 

[betwixt]

Intuitively, I agree that the chances of it working are slim but not impossible. The propagation delay may not be a problem. You mention the accumulated delay may overflow the time slot but that would only happen if the time slot itself was independant of the data. If the defined time slot is determined by the data itself or by using a known time interval synchronized to the data, it would be delayed by the same propagation processes and still be aligned correctly.

I would have long term reliability concerns, this sounds like a data link covering a large area and obviously involves a lot of electronics, even with best construction practise and good quality components the chances of a breakdown in so many units is high and in a repeater system, the weakest link can kill the whole chain.

Brian.

 

[rcdaniels]

I can give some information, but it would compromise my company to give too much.

We have up to 1000 linear communication links, each one with losses averaging 20 dB or so, but which could be more (up to double) under some circumstances. Currently, PPM is used to minimize power. The repeaters detect the PPM pulse, and generate an output pulse that corresponds to the detection time. Noise on each link is expected to be very low, almost inconsequential (except maybe when being cascaded 1000 times without regeneration, which is my main point). Channel distortion is normally low, but it is difficult to know how much field conditions will impact this. My concern is that even if noise or distortion affects the detection time by only 1 ns or so, with 1000 repeaters, the pulse could easily be pushed to the next detection slot (the PPM slots are only 80ns apart.)

I realize this is not a ton to go on, but what I am looking for is people who can state that this is a no-brainer: even with low noise and distortion, a good engineer just does not use 1000 repeaters with no regeneration anywhere in the chain, it is just too risky to let the noise and distortion from 1000 links add up.

I can't give you this analysis without more information, but it seems like you could easily make your point through symbol/bit error calculations. Sorry I can't help more, but my definition of a (non)regenerative repeater and your definition are slightly different. Does your non-regenerative repeater restrict its transmission pulse width to the same pulse widths of the original communication protocol? Does the regenerative repeater have integrity (error) checking?