Chest protectors, "new foil timings", different boxes & switch bounce

[TheArmourer]

There's a lot of talk about "good hits" not registering on chest protectors as the point "bounces".
Q1 Is this on all manufacturer's (new) boxes or on particular ones?

As far as I understand from the rules (m51 b2), scoring boxes shouldn't respond to anything less than 1ms.
I hooked up a foil to an osciloscope to measure the switch bounce, and even on hard surfaces, if the point was held long enough to register as a hit, the bounce always seemed to be in the order of microseconds.
If the "new timings" require that a hit is held for 13-15ms (IIRC), even with a massively long switch bounce, I don't see how this can really be an issue.
Q2 Can anyone explain why it seems to be? (Or am I missing something important?)

[remise]

Some people swear that the protector interrupts them getting a point. One fellow I know has always worn a protector for medical reasons, not for trying to 'get ahead'. I and others in the foil class have never noticed a difference in hitting him - although I will say that his lame is NOT skin tight (I heard that this can make a difference) and this fellow wears his protector LEGALLY - not beneath the lame, as I have read that others have done.

Hope this helps.

[DHCJr]

The rules for machines are designed for boxes, that do not exist. The rules are for boxes that have a continuous circuit and boxes now use sampling and programs to read the samples. Some of the early boxes with the new timing had a programming bug. If the blades came in contact after the touch but withing the timeout, it would start the clock over. Some boxes built their boxes on the rules as voted on by the FIE Congress, some did not and the Excutive Committee went with them. Since this is sampling, if one sample, which could be caused by many things, showed a break or a short to the lame, the light will not go on, depending on the sampling rate, which is not addressed by the FIE and the program.

[JEC]

According to the Nyquist theorem you need to sample at least at twice the frequency of the intended analog sample to be reproduced digitally. Do you know what is the sampling frequency of current model fencing boxes? It might be a specification to rate the quality of their electronics.

Fencing timings have been expressed in milliseconds. To reproduce a "on" signal for 12 msec (83.33 Hz), you need to sample at least at 166.7 Hz. That is not to bad. A lot of relatively cheap electronic equipment can do that. (At work, I use amplifiers with sampling rates of 30,000-90,000 Hz; those are fairly expensive.) I wouldn't be surprised that most new boxes are sampling at about 500-1,000 Hz. Meaning that they are able to reproduce a signal as brief as 2-4 ms. So, for some cheap boxes a change from 13-15 msec to 12 msec might not be noticeable. I guess that is what DHCJr might be alluding to.

Nyquist theorem explanation

[TheArmourer]

DHCJr & Remise, so is the whole switch bounce / chest protector thing officially an urban myth? (As I understand the whole thing, it must be.)

JEC, yep, farmiliar with Nyquist. (Got a degree in electronics.)
I was trying to figure out whether the commercially available boxes treated the timing and sample frequency seperately (by using independant timer registers and have a sample cycle based on the period of the actual decision making software) or whether they have the sample delay and timing subroutines inherantly linked.

Just as a point of interest, you refer to "cheap boxes". As far as I know, they're all relatively expensive. To buy a programable chip that will handle the task (and with a processor frequency of 4MHz) costs £0.50 (and that's without bulk discount). The two most expensive parts of a scoring box are the software and injection moulded cases. Software only costs what you choose to charge for it (and a lot of it is free) and appropriate cases (a lot tougher than some commercially available scoring boxes) can be bought for less than £10. So when you have to pay £200 / £300 / £400 etc, it all starts to seem a little expensive.

[KD5MDK]

What the software does with the signal once the sample is received matters, so errors could occur unrelated to sample rate.

[TheArmourer]

That was kind of my question - Is it the chest protector, the rules or the software that's the problem?

[remise]

It was once a factor, but not anymore. Refs now check to make sure the protectors are being worn legally.

[JEC]

This is the reply from one of the manufacturers of Fencing boxes:

"We do not give software or hardware information of our products."

With a relatively small market, they are able to take a stand like this. It would be useful to do comparative tests independently among the different fencing boxes. Anyone interested?

[DHCJr]

First, it is not an urban myth and as JEC stated, the manufacturers are not going to tell. This is not new. When boxes were circuit boards, the chip numbers would be filed off, painted over or some other way obscured.

It is interesting to note that SEMI recommended no more than 10ms debounce time after their rounds of testing. Anything more would cause direct attacks to not work. This information was sent of course to the Executive Committee to be written up and passed on to the Congress. Somehow in the process it got changed to 15ms when it got to the Congress.

[TheArmourer]

DHCJr I'm curious that they deal with debounce times in milli-seconds. My tests put them in 10s of micro-seconds, ie 1000 times less. Baffled.

[DHCJr]

As I said before, the new boxes do not send a continuous DC current. It takes samples, which are hopefully in the microseconds range. It sends a test, recieves it, does some processing, make some decisions and then starts this over again.

[JEC]

The minimal digital sampling for the circuit in a fencing box is either 0 or 1. It is either a closed or an open circuit. Like any electronic piece there is a threshold for one or the other. This is why the cables have a maximum resistance. If there is a high resistance, the circuit might not be able to detect the change in the state (closed/open) or just detect an open state. Theoretically, you could have an interruption during the debounce time that would not be detected if it occurred between two sampled times.

[TheArmourer]

OK guys, thanks for your replies but unfortunately none of this is really answering my questions. Maybe I'm not asking them clearly enough, but not to worry. I shall see if I can find the answers another way.
Thanks again.

[JEC]

Current fencing boxes do not read continuously in an analog fashion. They have digital equipment. Digital equipment by its nature has to sample at various times (sampling frequency). The manufacturers by economics use the cheapest electrical components that are able to reproduce the signal that the FIE requires them. It means that they use equipment that samples at some higher frequency than the minimum time that the FIE asks them to do. Due to some mathematical constraints that sampling frequency is twice as fast as the minimum frequency of the time they need to reproduced.

I suspect that higher level machines (FIE machines) probably sample more times than lower level (non-FIE machines). The more times you sample, they more opportunities exist to demonstrate a signal that was not sustained for the entire debounce time.

[rsy]

Originally Posted by JEC

Current fencing boxes do not read continuously in an analog fashion. They have digital equipment. Digital equipment by its nature has to sample at various times (sampling frequency). The manufacturers by economics use the cheapest electrical components that are able to reproduce the signal that the FIE requires them. It means that they use equipment that samples at some higher frequency than the minimum time that the FIE asks them to do. Due to some mathematical constraints that sampling frequency is twice as fast as the minimum frequency of the time they need to reproduced.

I suspect that higher level machines (FIE machines) probably sample more times than lower level (non-FIE machines). The more times you sample, they more opportunities exist to demonstrate a signal that was not sustained for the entire debounce time.

JEC,

I have no training in this so I am not sure I am getting this right. Are you & DHCJr saying that the higher the sampling rate, the more likely that the box will detect a change in the signal from a micro break and not record a hit? And if so, is that something that needs to be addresed by a technical fix that can adjust for this?

-r

[DHCJr]

In a sense, yes but it could be a short (B-C) or a break (B-A) that would stop a touch. But if the box was made, 'according to the rules', in other words using a continuous DC current, it would even more likely not signal.

If the sampling was slower, it might not signal also for another reason. Since we have talked about it needs to be at least half, let us use a sampling rate of 5ms. 1ms after a sample there is a touch, 18ms later there is a break. Will there be a touch? No, because there has to be 4 samples, 0,5,10,15 that show the hit to be a touch. At 4ms, the first sample shows a touch, a second at 9ms, a third at 14ms, but at 19ms there is no touch.

[JEC]

Agree with DHCJr. That is one of the problems that "WE" consumers lack the technical specifications of the boxes we buy from manufacturers. Furthermore, results of systematic test/rating of Fencing Scoring boxes are not available to the public (consumers) to make educated choices when buying scoring boxes.

I suspect that if DHCJr and I were in my lab with 15 scoring boxes and we test them for sampling rate, we could find significant differences among them. Probably greater differences among the non-FIE low cost boxes, but we could also be surprised. Perhaps, there might be uniformity as the SEMI might have tested them, forced manufacturers to comply to a "secret minimum sampling rate" but chose not to disclose results at the request of the manufacturers.

[TheArmourer]

What follows will be very boring for most people. You have been warned.
(And I appolgise if I'm telling anyone how to suck eggs ...)


There seems to be some missunderstanding about the capabilities of programable chips how a "sampling program" works in respect to a fencing box.

1. (Although I've not actually tested one) I can pretty much garantee there will be a constant DC voltage across the weapon, just as with the old boxes. It's the easiest way to do it ... and the rules say yo have to do it that way. Whether or not the software looks at it costantly should be irrelavent. The sampling itself does not affect the external foil circuit, in the same way that sampling the output from a microphone capsual for a digital recording does not affect the whether or not the diaphragm is oscilating. (As a point of interest even digital signals are usually continuous: the difference is that the wave from looks like stairs no a hill.)

2. Even the cheapest programable chips sample at frequencies far above what is neccessary to sample in milliseconds. The instruction cycles (the bits of the program that make decisions: fetch information form inputs, send information to outputs, write to or read from memory registers, make decisions, etc etc) take at the most microseconds. (In the the slow microcontrollers I am using, an instruction takes a MAXIMUM of 2 micro seconds.) If you only want to sample miliseconds, a chip will spend about 95% of it's time "counting on its fingers" just to waste time. (This is in no way an exageration. I can show you the machine code to prove it.)

3. It is possible use digial circuits that do monitor for breaks constantly. Such aplications will use what are called "interupts". eq if there is a change in an external circuit, it can cause a voltage to go high (ie provide a logic value of 1), which in turn can interupt what the processor is doing and tell the CPU to make a note of the timing generated by the chips internal timer. Such methods do not require "sampling" of inputs at all. Chips are readily avaiable that can respond to this in as little as nano seconds ie virtually instantaneous. And they cost significantly less than a fencing box!

Sampling is a completly different way of dealing with external signals.

4. As for FIE complient boxes using more expensive processors, this (IMHO) is highly unlikely. Even the cheapest microcontrollers are incredibly powerful devices. The are in effect complete computes in a lump only a couple of centemeters square. (Obviously they don't have a mouse / keyboard / screen, but these have little bearing on what the microcontroller does - they are merely input and output devices, analgous to the foil, lame, bulbs and buzzers in a fencing scoring box.) In terms of number-crunching-ability, Microcontrollers that cost only 10s of cents / pence are capable of doing all the processing for several pistes, all at the same time, and 1000 times faster than is needed. The thing that stops them is most likely to be the restiction in the number of physical inputs and output. If you want to use a more expensive chip, you can increase the overcapacity by a factor of 1000. (It's called a Pentium and you're probably sitting in front of one.) Telling a chip to do something that is FIE complient is no more expensive than non FIE complient (if you discount the time to rewrite the code.
It is more likely that increased costs for FIE complient boxes are due to more components in the boxes (other than the procerssor its self, and that would only be likely to cost a dollar or two more), highter development costs, more hassle getting the boxes certified FIE complient, or (and I am not accusing anyone of anything here) possibly even manufacturers taking the **** and charging people more because the customer doesn't understand.

There is a huge amount of information on microcontrollers, their capabilities, and how to use them all over the internet. A good place to start is www.piclist.com



I hope this helps.
The Armourer BSc (Electronics)

[JEC]

Some apparatus of some manufacturers, which do have microbreaks problems are currently under study.

Page 8 of SEMI June 5-6 meeting report
http://www.fie.ch/download/letters/2...202005-ANG.pdf

Please notice that it read "some" apparatus of "some" manufacturers. This report is confirming that there are differences among scoring boxes as it pertains on the detection of debounce time.