Favero on the New Timings

[Craig]

Dan,

The problem is that code that works on one chipset and set of hardware will not work on another, or will behave differently. I know this from working with the different manufacturers for hardware in my real life.

What is clear to me is that the full functional spec was not developed. The FIE provided a broad product requirement: "hits should be 15ms to register" but did not fully specify what they meant by that: "non-valid hits should be x, valid hits should be y, 15ms specifies time for the circuit to be in Y state, requires z ms contact with conductive surface".

It's really easy to code to a true technical specification. The problem is that the FIE delivered a requirement that left a lot of the specific functionality undefined so the manufacturers had to best interpret what the FIE wanted.

It looks like the manufacturers are sharing this knowledge and some of the results so hopefully the FIE will have SEMI issue a true technical requirements document.

Craig

[darius]

The FIE isn't in the code business.

They provide specifications for what the boxes need to do (poorly, given the off-target debounce issue), and the implementation is left up to the manufacturer.

If I want to create a box from relays, breadboards, and alligator clips that implements the FIE spec, then it can be used in competition. Forcing people to make a specific implementation will stifle creativity -- we never would have had user-upgradeable boxes if that had originally been mandated.

darius

[oso97]

I'm still waiting for a fully programable box that allows one to vary the various parameters...

[DanInMI]

Originally Posted by darius

The FIE isn't in the code business.

They provide specifications for what the boxes need to do (poorly, given the off-target debounce issue), and the implementation is left up to the manufacturer.

If I want to create a box from relays, breadboards, and alligator clips that implements the FIE spec, then it can be used in competition. Forcing people to make a specific implementation will stifle creativity -- we never would have had user-upgradeable boxes if that had originally been mandated.

darius

Baloney, you can create a jacket that conforms to the FIE standards but if it doesn't have the FIE stamp of approval then you can't use it in an FIE competition. That is what i am recomending here.

[DanInMI]

Originally Posted by Craig

Dan,

The problem is that code that works on one chipset and set of hardware will not work on another, or will behave differently. I know this from working with the different manufacturers for hardware in my real life.


Craig

Well, yes I understand that the FIE would have to mandate that the code is approved to be used in a particular way. The point is that in order to avoid one company having a monopoly on scoring machines they should require that the technology be licensable. Whether the FIE does it themselves or they require that the creator of the design do that it should be done.

[jBirch]

Originally Posted by DanInMI

Baloney, you can create a jacket that conforms to the FIE standards but if it doesn't have the FIE stamp of approval then you can't use it in an FIE competition. That is what i am recomending here.

No, you're talking a different issue. The FIE stamp is there to ensure that the item has been tested as conforming to FIE specifications, not to specify HOW it's been made to conform.

The code issue is the same.

As Craig mentioned, all that needs to be done is for the particulars of the technical requirement to be spelled out and then tested and enforced. It would be very plausible in my mind for the FIE to say, on this particular issue, that the acceptable "microbreak" length is no greater then X. They could also approve a 15ms debounce time as having a duration of 15ms and an average signal strength of Y over that time. Lots of ways to do it though the fact remains that 15ms debounce is too loose a term to capture the whole requirement.

James.

[DanInMI]

Originally Posted by jBirch

No, you're talking a different issue. The FIE stamp is there to ensure that the item has been tested as conforming to FIE specifications, not to specify HOW it's been made to conform.

The code issue is the same.

As Craig mentioned, all that needs to be done is for the particulars of the technical requirement to be spelled out and then tested and enforced. It would be very plausible in my mind for the FIE to say, on this particular issue, that the acceptable "microbreak" length is no greater then X. They could also approve a 15ms debounce time as having a duration of 15ms and an average signal strength of Y over that time. Lots of ways to do it though the fact remains that 15ms debounce is too loose a term to capture the whole requirement.

James.

My point is that it should really be the same code and chipset used in all the boxes to ensure that it is the same on all.

A comparison to such a practice is one design sailboat racing. Many one-design class organizations (MC Scows, for example) own the master mold for the hulls of thier boats. Multiple boat manufacturers may build boats, but they must make thier molds from the master mold so that they are all alike.

That is what I think should happen with scoring machines.

[DHCJr]

Consider, this before these boxes each manufacturer made their own boxes. St. George went ahead for awhile, because they had inovations that were not mandated. For example, if you had a short in your Epee you could still get a touch. This was not true for other boxes.

If requiring the same code is the way to go, why not require the same computer language for all govenment computers? I prefer having a minimum and those who go beyond the minimum will get my business.

[jBirch]

Originally Posted by DanInMI

My point is that it should really be the same code and chipset used in all the boxes to ensure that it is the same on all.

A comparison to such a practice is one design sailboat racing. Many one-design class organizations (MC Scows, for example) own the master mold for the hulls of thier boats. Multiple boat manufacturers may build boats, but they must make thier molds from the master mold so that they are all alike.

That is what I think should happen with scoring machines.

I see what you're saying. I disagree with it, but I see what you're saying.

The reason that the boats are mandated is the same reason that cars are mandated in motorsport. Particularly, it is because each individual competitor is competing both technically and technologically. In fencing, and with the boxes in particular, it's not so. Each competitor is equally susceptable to the vagaries of the box so mandating that they conform to a certain set pattern doesn't do much except stifle innovation and lock down changes.

I'm with DHCJr and Craig in that the parameters need to be outlined rigourously but the implementation of those parameters is free.

Hope this helps.

James.

[DieterS]

The majority here can see the problem.

Just to clarify, the term "micro-break" is actually a misnomer. As most of you probably know, in foil a touch or off-target is registered not by making but by breaking contact in the point. So, the "micro-breaks" that Favero is talking about are actually "micro-contacts."

No, the FIE rules do not allow for any "filtering." As it says in Annex B.A.1.b.1 (amended to include the experimental timings):

"Any hit must cause a signal whatever the resistance of the cicuits external to the apparatus. The duration of the break of contact which must always ensure that a signal is registered must be 15ms."

And in b.2 it continues:

"An apparatus will be rejected if it is possible for a valid hit to be registered as a result of a break in contact of less than 13ms."

However, until the FIE fixes these timings, here's something to try. Before taping your blade, paint the uppermost 1/4 inch (.6 cm) of your barrel with something like nail polish. Then tape your blade. Let us know if you notice a difference.

Dieter
Eigertek

[DanInMI]

Originally Posted by DieterS

T
However, until the FIE fixes these timings, here's something to try. Before taping your blade, paint the uppermost 1/4 inch (.6 cm) of your barrel with something like nail polish. Then tape your blade. Let us know if you notice a difference.

Dieter
Eigertek

great idea, thanks

[R. Exnicios]

okay, I'll bite what should we see happen?

[C.J.]

I'd assume the letter was written by an engineer, so the term micro-breaks indicates the breaks are measured in microseconds. So that means whenever a machine detects a break like that, it could be nothing else(e.g. it is humanly impossible for a remise to be made in such a short period of time.) So there is no need or reason to make rule changes.

There, however, is still the problem of the insulation tape on the barrel being pushed away on impact and resulting a short. Rigid insulation will be needed for the barrel as well.

[keith]

Originally Posted by C.J.

I'd assume the letter was written by an engineer, so the term micro-breaks indicates the breaks are measured in microseconds. So that means whenever a machine detects a break like that, it could be nothing else(e.g. it is humanly impossible for a remise to be made in such a short period of time.) So there is no need or reason to make rule changes.

um no, as DieterS pointed out these are either;

1) microshorts between the barrel of the tip (if this is metallic) and the base (the bit of the tip that screws onto the blade) via the lame.

Or

2) microbreaks where the 'constantly depressed' tip loses electrical contact with the lame.

Either of these events will result in the failure of the hit to be recorded if the current FIE spec is followed. Working on DieterS's observation that the current spec allows for no filtering.

Originally Posted by C.J.

There, however, is still the problem of the insulation tape on the barrel being pushed away on impact and resulting a short. Rigid insulation will be needed for the barrel as well.

This is a subset of problem 1 above. The real issue is whether or not there is a difference in how the manufacturers have implimented the current rules. Leaving aside that it appears even the FIE has no bloody idea how the current spec should be implemented.

Do any of them have a filter in place, and more importantly what is it? See the other timings thread for a description of the filter used by the spanish ATLAS machine

(Also we need to know what the various sampling frequencies are since there is a difference between filtering via sampling rate and applying a post sampling filter.)

[DanInMI]

Originally Posted by DieterS

Just to clarify, the term "micro-break" is actually a misnomer. As most of you probably know, in foil a touch or off-target is registered not by making but by breaking contact in the point. So, the "micro-breaks" that Favero is talking about are actually "micro-contacts."


Dieter
Eigertek

Ummmm... I am not an electrical engineer, so please forgive my stupidity...I understand that there is a connection broken when the tip is depressed, but isn't there a simultaneous connection made between the point and the lame? Isn't there one connection that is broken when the tip is depressed, and one that is made when the tip is depressed and touching the lame?

I believe it is the connection between the tip and lame that is in question. This is where the issue lies. The tip maintains depression, but the connection between the lame and the tip is broken (or weak) very briefly. Or the lame comes in contact with the barrel very briefly, grounding out the connection.

If we can filter out these brief interuptions so that the boxe recognizes that the tip has stayed depressed and on target for the duration of the debounce timing shouldn't that be done?

[jBirch]

Originally Posted by DanInMI

Ummmm... I am not an electrical engineer, so please forgive my stupidity...I understand that there is a connection broken when the tip is depressed, but isn't there a simultaneous connection made between the point and the lame? Isn't there one connection that is broken when the tip is depressed, and one that is made when the tip is depressed and touching the lame?

I believe it is the connection between the tip and lame that is in question. This is where the issue lies. The tip maintains depression, but the connection between the lame and the tip is broken (or weak) very briefly. Or the lame comes in contact with the barrel very briefly, grounding out the connection.

If we can filter out these brief interuptions so that the boxe recognizes that the tip has stayed depressed and on target for the duration of the debounce timing shouldn't that be done?

Tip touches lame. Connection one. Base of the tip touches contact inside the barrell that is kept apart by a spring that is fully compressed when a force of 500g is applied to it. Connection two. Both connections need to be complete for 15ms (debounce time).

So, two things can happen.

1) The lame material touches the barell causing the circuit to short out so that the return path of the current goes through the lame instead of the tip. This looks to the electronics like a hit to the guard. This is called a micro-short (or micro-ground).

2) The spring uncompresses and recompresses in a fraction of a second (due to the tip skipping off folds in the fabric of the lame or bouncing off the surface) looking like the current is broken and then reestablished. This is called a micro break.

If we digitise the circuit with a 1 ms sampling rate we may get a data set like this:

111101111101111

Now, with the 15ms debounce time in a continuous circuit, this looks like three seperate hits of 4, 5, and 4 ms each, each of which would have registered under the old system but which won't register under a strict interpretation of the new system. However, from this sample, it's pretty clear to see that the tip was actually depressed for 15ms since there's no way that a human could remise in 1ms three times. If this is to be considered a good hit then we need some sort of algorithm to determine the difference between that and something like this:

100010001000101

or this:

100000000000001

Hope this helps.

James.

[keith]

Originally Posted by jBirch

If we digitise the circuit with a 1 ms sampling rate we may get a data set like this:

111101111101111

well really you have two circuits

Tip/blade 111000000000000000 111
tip/lame 000111100111100111 000

.. or a near infinite combination thereof.

So was the rule change designed to demand that a hit must be delivered with >500gms of contact through the tip for >=15ms. God knows, and until the Pope gets his voice back we are just going to have to wait.

... so how do you separate between a slapped on hit, and a technically perfect hit with microbreaks/shorts? Should one only filter the lame/tip signal, or only the tip/blade signal, or both?

[C.J.]

Simple. You'll need both, and tip to lame has priority over tip to blade.

[garyhayenga]

Originally Posted by jBirch

Tip touches lame. Connection one. Base of the tip touches contact inside the barrell that is kept apart by a spring that is fully compressed when a force of 500g is applied to it. Connection two. Both connections need to be complete for 15ms (debounce time).

A slight correction. In foil the base of the tip does *not* touch the contact inside the barrel to complete a connection.

Instead when the tip is in it's normal undepressed state the current flows down the wire to the contact and through the spring to the tip and through the collar of the tip to the barrel and back down the blade.

When the tip is depressed the it no longer makes contact with the barrel and the circuit/connection is broken and an off or on-target light to go off.

The spring then pushes the tip back up and into contact with the barrel again.

gary hayenga

[jBirch]

Originally Posted by garyhayenga

A slight correction. In foil the base of the tip does *not* touch the contact inside the barrel to complete a connection.

Instead when the tip is in it's normal undepressed state the current flows down the wire to the contact and through the spring to the tip and through the collar of the tip to the barrel and back down the blade.

When the tip is depressed the it no longer makes contact with the barrel and the circuit/connection is broken and an off or on-target light to go off.

The spring then pushes the tip back up and into contact with the barrel again.

gary hayenga

Yeah, I know, but I was trying to be simple.

James.