Switch Torture Test (preliminary)

The small white Convoy switch is an Omten 1288

If RMM’s black Omten 1288 is the same as the ‘small white Omten’ found in Convoys, then I have plenty of those and already have them listed in the ‘switches to be tested’.

I think the real value of this test is not in trying to find any ‘absolute’ reliability numbers, but in simply comparing switches A, B, C, D, etc. to each other. As long as they are all going to get the same treatment, the comparison should be valid. As for the reason, pflexpro is a flashlight modder/builder/seller. He has an interest in making the best possible product by using the best components within his reach. So, the comparative testing is going to help him know what to use, and will help all our BLF modders as well when he shares the info.

Thanks, pflexpro, for taking the time to not only do this test, but to share it with us!

Ahh, okay, I guess I didn’t realize that… thanks PD. and, never mind pflexpro! :wink:

My intention for the test is to determine the durability of several ‘common’ switches that I (we) use on a daily basis…both electrically and mechanically.
I tried to choose a current, click rate and total click life that would represent a good balance between electrical and mechanical reliability.
The number of clicks* doesn’t really matter because, unless some number of the test switches fail, then the test is meaningless. I chose 20,000 clicks as a pass on the original test, but if no switches fail within that number of clicks, then I would perform a second round of testing with the original switches…it’s just, I would be happy if a switch makes 20,000 clicks running at twice it’s rating.

Since I will have the investment (both time and money) in the test rig, then it would be no problem to repurpose it for an additional ‘hot-rod’ test.
What I would need is some input on the basic test parameters…power, click rate and switch selection. It seems 10A may be good for the hot-rod test, but I’m sure, some my want the current even higher.

*I’m no longer using the term ‘cycle’ because it causes some confusion about what a cycle is…is it one click…is it on/off (two clicks)…so now I will just refer to it as ‘clicks’ and I think everyone will know what I mean.

Tail cap switches are sold separately and often require no soldering to remove/install so they are definitely easy to change. Just a retaining ring to unscrew.

I like this idea- we’ve had only sporadic empirical evidence of durability so far while other switch aspects have been covered much better.

I’m not ‘picky’ about most things but I am very picky about things functioning as expected to. I seriously doubt that I’ll ever wear out a flashlight switch but if there’s a ‘lesser’ switch in my lights I’ll soon know which one(s) could stand an upgrade and what needs to go in their place.

Subscribed!
Phil

Yes, this should be a great test.

20,000 clicks may sound like a lot, but I probably average at least 50 clicks a day across all my lights. Especially on electronic switches, there’s lots of mode switching, etc. If I used one light all the time, and did 50 clicks a day, it would only take a little over a year before I did 20,000 clicks on it.

I think electronic switches are built for well over 100,000 clicks. Mechanical ones probably a lot less.

It looks like the Omten switches are typically rated for 50,000 cycles…some are 100,000 cycles, but these numbers are based on ‘rated current’. On a good switch, I really don’t expect to see a mechanical failure at 20,000 cycles, but you never know. I’m not sure what effect increasing the current anywhere from 2.5 to 5 times the rated current will have, but it will definitely shorten electrical life and will probably shorten the mechanical life because of heat.

It is BUT it isn’t, the small white switches are chinese knockoffs of the omten 1288’s. There is a world of difference between the two as far as the quality of construction materials used. If we could talk them into making them out if delrin…

I’ve been throwing around the term ‘rated life’ (from the manufacturer) as if it means something to us…well, it does and it doesn’t. The rated life is based on 250v AC. AC current is much less destructive on a switch. The ‘contact break’ is the most electrically destructive portion of the cycle. With an AC current, the break usually doesn’t happen while the AC current is peaked…it often happens when the wave is close to zero and the load is very light. I seem to remember a de-rating of a 250VAC switch to 14VDC. Also, there is a de-rating for the type of load…resistive, inductive, capacitive. I’m not really sure what type of load we’re working with…I know, some drivers actually have an inductor mounted on them, so they probably create an inductive load. The 7135 drivers, I would guess are probably a resistive load…or an inductive load, or both…or neither…I’m so confused!
There is also the possibility that the manufacturers ratings aren’t accurate in the first place…if ratings are even available.
One switch I will be testing has a manufacturer’s rating of 6A -it’s printed right on the switch, so it must be true…yeah right!

This seems like a valid point.

  • As far as the proposed high-speed test being a level playing field, I think that’s a bit misleading. Changing one variable (frequency of clicks in this case) doesn’t always yield results which apply equally to real world usage scenarios. Imagine a basketball game where gravity is reduced to 0.5x Earth gravity. Would the results of this game have much bearing on the results of a normal gravity game? Probably not! Some players might gain a large advantage while others might gain no advantage at all.
  • If the light meter samples with a frequency of 1s and doesn’t rely on a “trigger” input it seems that this will yield poor results when used to monitor a light being switched on and off at at a higher frequency. It seems to me that either synchronizing the light meter or running it at a much higher sample frequency is required.
  • Attempting to test two variables at once is often a bad idea. Maybe high-speed mechanical switch testing would be best done with no current or minimal current. Medium/high currents could be tested separately at a lower frequency.

EDIT: Here’s a picture demonstrating the issue with a too-low sample rate: File:CPT-sound-nyquist-thereom-1.5percycle.svg - Wikipedia

Having opened up a few of them I’ve found that the main cause of failure in the smallest, cheapest switches is from having current flow through the spring. Next is the size and thickness of the contacts and how smooth they are. And the final factor is contact material and plating which is heat resistant and doesn’t melt. Switch life can be prolonged by cleaning the contacts of carbon deposits and resurfacing but unfortunately copper while lower in initial resistance is more prone to damage from arcing than nickel plated contacts and flashlight switches aren’t made to be serviced.

BS or not? Just my impressions from experience with flashlights and power tools.

A cam or rotary push system will induce an addition factor…lateral force. While the lateral force should not affect the electrical durability, it will alter the mechanical functions…and the switches are designed and typically used with a simple straight force. The solenoid will will have a silicone push plate which will soften and somewhat slow down the quick cycle. The speed of the contact closure will not change with a fast press vs a slow press.

I don’t plan on changing the click rate after the testing begins. Right now, I’m leaning toward 2 clicks per second so all switches will be tested at this rate. Testing 8 hours per day, it will be running for over 4 days …this doesn’t include each switch setup and intermediate resistant readings for each switch at certain intervals. If I slowed the click rate to 1 per 4 seconds which may still be faster than typical, the duration of the test would be over a month if I test 8 hours every day. Although, I won’t need to watch the test continuously, it will not be running completely without intervention.

As far as synchronizing the light meter…if I use a PLC to control both the solenoid and the polling of the meter, this would be possible, but I can’t control the meter’s polling. If I tried to manually sync the relay speed with the meter, I may get it to match initially, but unless it’s perfect, it would be off after some amount of time. The relay will have two timing functions…the duration of the stroke (which will be short and dependent on the requirements of the solenoid) and the delay from the end of the solenoid stroke to the beginning of the next solenoid stroke. Either adjustment will change the full stroke cycle duration and will cause a mis-match with meter polling sync.

My solution for the output is actually simple. What I’ve done is create several macros in excel. These macros will look at the varying output levels…split them into groups of 30 seconds (or any amount of time I specify), look for the peak readings from the 30 second sample and replace all of the readings for that 30 seconds with the high reading. I’ve use an extreme method to verify it. I recorded the output from a light set in strobe mode for 60 minutes…of course, graphing the results looked a mess. After running the macros and re-plotting the results, the graph looked perfect…you could actually see the decline of the battery from the recorded output. A strobe mode is much more challenging than a 2 or 4 cycles per second. Using the 4 cycles per seconds, I calculated there would be a failure in the macro method at 1 out of 10.8 billion.

As far as testing 2 variables at once, part of it is a time issue…I will already be dedicating a week to the actual testing at 2 cycles per second (unless there are some early failures). If few or no failures occur, then the testing will continue in 2 to 3 day increments. Another reason I’m testing 2 variables simultaneously is…‘that’s the way we use the switches’…in a flashlight, everytime we click it on, we’re depending on both electrical and mechanical functions to work. In reality, I don’t care if it fails electrically or mechanically…a fail is a fail.

I’m not trying to be difficult, but I have to look at how many days or weeks I’m willing to dedicate to evaluating 6 switches.

Great project. This should be very interesting.

You open switches…I throw them away.
So, one question I have is: Do typical flashlight switches (clicky) like the ones being tested…do they all use the internal spring as a conductor?

Clearly time is a big factor here (that applies for most of us!). So is the availability of tools.

I recognized in my suggestion that you’d be unable to sync the light meter readings… clearly if the meter had a trigger input or similar a PLC could be used as you mentioned. Filtering the results in the way you describe will get you “something”, but it may not get you what you want. What precisely are you hoping to gather from the light meter readings? [For example, the light meter won’t give us switch resistance. Such heavily filtered results also certainly will not tell us whether the switch fails to actuate 1 time out of 10.]

It sounds like you’ve already planned to minimize some of the problems with the solenoid, so that’s good. If a motor and cam assembly was used it could certainly move a piston to remove the lateral forces. Setting that up doesn’t have to be a big deal - but I recognize that it’s one more thing and it sounds like you may have the solenoid idea under control. IMO as long as it’s not bottoming out the switch that’s probably the main thing. (The other issue, as Lazy-R-us wrote, is acceleration. I’m not that familiar with operating solenoids, but it seems like a motorized cam would be much easier to manipulate in that respect.)

Saying “that’s how we use them” about the current/clicks question is definitely hand waving IMO. We’ve already established this test is not about “how we use them” since it’s being done at a constant rate of multiple clicks per second. IMO a key focus in this conversation is to (hopefully!) figure out ways that the test conditions could sway results. Nobody wants the results to be swayed. For example, if high-frequency high-current clicking causes a higher operating temperature than normal in some switches but not others this could be very significant. It could allow metal and plastic to deform in the affected switches, throwing off the results significantly.

This is deceptive, but really the big question here is: what variables do you hope to quantify in this experiment? Certainly the answer can’t be “everything about a switch”. We know that we won’t can’t walk away from the results with a “best switch” crowned.

I’d rather wait to bring this up, but I’d better jot it down while it’s on my mind. Of course how relevant this is depends quite a bit on the answer to the question at the bottom of my previous post.

Anyway here goes! To minimize time I think it would be reasonable to omit the light meter readings. As I alluded to up above, I think that as-described those readings provide no useful information. Omitting them allows you to do the clicking in parallel. Servos are cheap and a setup like scaru’s can be quickly and easily adapted to use a pushrod, eliminating enough lateral forces to fit the bill. If really desired a piston could be used to entirely constrain lateral motion, but I think that’s a bit much: a pushrod will be closer to real world usage. Once the clicking is going on in parallel you’re much more free to run the test at your own pace. After every 1 thousand clicks (or X number of clicks) tests can be done on actual resistance (using voltage drop under heavy load). Potentially other relevant tests can be done at that time.

Furthermore if we wish to measure successful on/off clicks a PLC or microcontroller can be used along with a simple logging system to just give Boolean (true/false) outputs for whether the switch was in the expected state after being clicked. Frankly there’s little need for synchronization if the logging frequency for that Boolean data is increased to something like 100hz. At that rate there are less than 500k samples per hour per switch. No big deal at all storage space wise.

I open them to find out why I needed to throw them away or change the way the contacts exit the case. Tofty, Judco, Omten 1288, Kan and the like do not use the spring to carry current and so last much better. 501b/502b switches have very poorly made cases of weak plastic.

After trashing the test switches, I will open the to check the condition.
The Solarforce is an important switch to me and I wonder if it’s using the spring.
Also, concerning the McClicky switch, I’ve read that ‘the spring’ will heat up over 3A…I wonder now if they’re referring to the contact spring or internal spring.