I'm using a Cateye wireless, the cheaper unit which doesn't do average speed. It works nicely, records mileage per trip and total, which is what I wanted it for. It's nice to see my instantaneous speed too. However the max speed reading ends up with ridiculously high numbers almost every ride. I don't normally check max speed while I'm riding, but the last few miles of my typical ride include a good downhill with some serious bumps. When I check it after that section it is often ridiculous.

I figure the high speed readings could be the result of the magnetic switch bouncing, especially while the magnet itself is going by. This could result in multiple readings from the detector that really don't correspond to complete rotations of the wheel.

On the other hand, maybe I really did hit 87mph on that downhill today. But I doubt I ever hit 203mph.

Has anyone else ever seen this kind of behavior?

No, problems with the magnet and the pickup would only cause a loss of "clicks". That is, problems here would only underestimate the speed.

The problem you are having is extra clicks being "detected". This is due to interference from some other radio wave source (some other "transmitter"). This is a fairly common problem with non-coded wireless computers.

The only way to fix this is to either get a wired computer, a more-expensive computer (one that uses a coded transmitter), or avoiding the interference!

Problems with the magnet CAN cause extra clicks - if you mount the magnet wrong so that the north/south poles are sideways, each one can trigger the sensor. However, this would result in double speed readings at all times, particularly at slower speeds.

I agree though, it's probably interference. One of the reasons I wouldn't even consider wireless.

Thanks, both of you. Maybe the fact that this section is near a radio station's antenna array has something to do with it. The bogus readings do always seem to come in this section of road. BTW, I can get extra clicks by using two magnets to go twice as fast! Of course, a well-designed receiver wouldn't be susceptible to double-clicking if the input was de-bounced correctly in software. In any case, if this is a known problem with wireless units then I won't worry about it. (An alternative explanation could have been that a bit or trace or solder joint is bad in the unit, and once it hits some speed which uses that bit the result becomes bogus.)

But please, don't disillusion me so matter-of-factly. I really did hit 87 today. I did, I tell you, I did! I hit 87.

Note that this would be a faulty installation (if the device is installed incorrectly, all bets are off!).

I'll have to see if I can get this to "work" with my computer.

"De-bounced"?

The newer wireless computers send a code to the head unit. This avoids interference from other bicyclists using the same computer and avoids extra count issues.

Ideally, the computers would ignore ridiculously high speeds or look at a moving average to detect unreasonable high speed values.

By "de-bouncing" I meant a software scheme to rule out clicks that came too close together. The term comes from the fact that mechanical switches have significant bounce when closed. If software is polling the switch or servicing interrupts fast enough it can read the bouncing as more than one switch action. In a worst case the switch is a momentary contact which is supposed to be a toggle between on and off. In this case the final state of the switch depends on whether the system reads an even or odd number of closures. If it is supposed to merely read the final state of a two (or more) position switch it will eventually read that state correctly. But if it is supposed to take certain actions on each transition it might take them unnecessarily for the bogus closure events.

In either case a de-bouncing scheme is often implemented where the switch is read several times and the same value must be read so many times in a row. In the case of simple events it can read one event then exclude all others for some fixed amount of time. The challenge for the designer then is to know how long that time should be so that real events are detected but bogus ones are not. This could be implemented on the microprocessor in the detector unit or in the display unit. Of course all this assumes that the wireless transmission itself is not confused by other nearby systems.

Ah, jargon!

There certainly could be some scheme for sanity-checking. The basic computers don't appear to do that.

Implementing it in the transmitter unit would be more expensive.

Note that the "extra clicks" issue isn't a problem with the magnet and the pickup (since wired computers don't have this problem).

The whole problem with wireless computers is "other nearby systems" (that is, the problem is interference from other things).

The fancier computers use a coded transmission, which eliminates the confusion.

Well, yeah. The computer&software biz tends to be that way. Not at all like the bike biz where they use ordinary words like brifter and freehub and drops.

Not necessarily. The transmitter must have some sort of microprocessor to control the radio and to shut it off when idle. I wouldn't be surprised to find an inexpensive low power PIC or similar device inside. The code to de-bounce the magnet detection events would be just a few lines, easily within the code space of even the cheapest devices.

That is indeed the key observation. So it's a uniquely wireless problem. Some other transmisioxxxyx-brrrrpt-zz

It would be more expensive, somehow. (Maybe, not a lot more expensive.) Those extra "few lines" has to be written and tested too. I suspect that the transmitters are designed to be as minimal as possible. It's much easier to do that work in the head unit anyway (since it is already calculating a bunch of other stuff).

And the code could be shared between wireless and wired head units (another reason it would be less expensive to do it in the head unit).

And, for the wireless units, that code could be used to handle interference errors too! (It doesn't make engineering sense to do the correction in the transmitter.)

That's right, and consider my experience as well.

For years I've used rollers as trainers and had wired computers for speed/odo functions. Over the years, I discovered I could "wear out" the magnetic reed switch in the pickups!

That's right. After several years of use they would become "sticky" and start sending erroneous info to the computer. They would work perfectly at 20-30mph, but during sprint efforts and 35mph+ they would "lose it" and drop out to zero intermittently.

Radio interference is always a concern with low power consumer gadgets. You've been "gotcha-ed"