More contrast at the sensor might be helpful so I charted some contrasts using a similar sensor, OPTEK K-2362. Vcc 5.26V, diode 40mA, transistor as emitter follower, 1k load.
Bright Al.---- 2.5V, Al. tape --- 2.6 V, white paper --- 190 mV, black tape --- 12 mV,
Shiney white paper, --- 200 mV, open ---1.5 mV
22mA diode, R 1k, Al tape --- 1.18 V.
Diode-40 mA, R= 2k7
AL. tape --- 4.9V, black marker pen on paper, --- 123mV.
I would put a comparator within an inch or 2 of the sensor if I had a noise problem.

 

Based on this data it seems that the best choice is bright aluminum tape, with the rest of the surface a dull flat black. Raising the initial signal is a great way to improve the signal to noise ratio.

 

XPP, have you been able to improve the signal to noise ratio enough to minimise the 5k trash ?

 

Actually not. I haven't had contact with the motor since last friday because it's at work. Tomorrow I will grab it again and retake the topic. The 5k signal still leaks sometimes affecting my measurements.

 

Update: I twisted the motor wires and attached a clamp on ferrite bead. The noise seems to be gone. I'm measuring the speed correctly because I also used a digital tachometer, so the micro and the tach agree. However, I have to do a software adjustment since the controller is not "always" giving me the right speed. It's like I have to do something to reject non-expected values. But surely the 5kHz trash is gone.

 

How often are you reading the sensor FB?
What processor are you using?
Any kind of PI or PID or just simple correction.
Max.

 

How often are you reading the sensor FB?
What processor are you using?
Any kind of PI or PID or just simple correction.
Max.

I'm using a L432KC nucleo from STM32.
I still don't have the PID controller. I'm just reading speed at open loop without load, so the speed should remain constant. For example, I always shoud measure 341 RPM. But the micro readings jump between 341 and 22k. The tach remains always a 341 RPM.

 

Do you know how long or how many samples show the 22k reading?
If that much error for a very short inconsequential period you could use a sample filter to reject 'Noise' or what ever is causing it .
The 'scope may give you a better picture of what is happening.
Max.

 

Do you know how long or how many samples show the 22k reading?
If that much error for a very short inconsequential period you could use a sample filter to reject 'Noise' or what ever is causing it .
The 'scope may give you a better picture of what is happening.
Max.

I really don't know. I will see what the scope says. Although, a simple solution would be if I create a condition that ignores all the readings above the max speed I plan to measure.

 

a simple solution would be if I create a condition that ignores all the readings above the max speed I plan to measure.

Or any reading above what would be a credible reading.
Max.

 

Or any reading above what would be a credible reading.
Max.

Yeah, for instance, I said
sample_speed();
if RPM>2000 {
RPM=0;
}
else{
RPM_True = RPM

It's been working well like that.

 

I'm using a L432KC nucleo from STM32.
I still don't have the PID controller. I'm just reading speed at open loop without load, so the speed should remain constant. For example, I always shoud measure 341 RPM. But the micro readings jump between 341 and 22k. The tach remains always a 341 RPM.

It is undoubtedly a software issue, depending on how the speed is calculated, it may be some variable is occasionally used for some other purpose and gets a wrong number loaded in. So my suggestion is to fully understand exactly how the speed is determined. An easy way out might be to average several readings before making any corrections in the speed control output. It may be a random interrupt happening occasionally.

 

Nov 4, it has been a while. I have one more thought that possibly the occasional jump in displayed speed might come from an irregularity in the speed sensor signal. If one of the pulses is narrower and the software is using pulse width or time between leading edges that could explain the problem. The simple fix is to use averaged pulse frequency to calculate speed. Of course it may not be that simple a software change, but it would require no hardware fixes. I have been stuck with chesing noise when the problem is in the software and it can be a pain indeed.

 

Nov 4, it has been a while. I have one more thought

Almost a year later (last post 11-7-18) and your still thinking on this problem??? Quick on your feet aren't you?

 

Almost a year later (last post 11-7-18) and your still thinking on this problem??? Quick on your feet aren't you?

Well, Excuse Me! This post just popped up on that short list of current posts, so it looked like something current.

 

Hey...better late than never!

 

Hey...better late than never!

It looks like I have been fooled one more time. But like I stated, this thread was in the 5 boxes posted as the most recent ones. And that "old thread" notification never popped up like it used to do. So once again I will say it: "At best only HALF of all changes are improvements."

 

Just get a stroboscope and adjust the frequency until the the spinning motor appears to be standing still. Painting a single white dot or stripe would make the measurement less ambiguous on a symmetric surface.

View attachment 161965

Only problem with this method is knowing you have the right RPM. For instance, if you strobe your strobe at 500 strobes per minute (SPM) and the motor is spinning at 1K RPM the motor will spin two revolutions between strobes. You may think you have 500 RPM when in fact you have a higher speed. IF you strobe at 2K SPM and the motor is spinning at 500 RPM you will see multiple points simultaneously (or what would appear to be simultaneously). Should show four points, each 90˚ from the previous (assuming you have exact speeds and strobe rates). (any imbalance between the two rates and the dots or stripes will appear to move in one direction or the other.

[edit] YOU MEAN THERE ARE SIX PAGES? Guess I should do a little more reading before I post a comment. Sorry for backtracking so drastically. [end edit]

 

Yes, that happens.

I used stroboscope in one job for a couple of years and don't remember this as being a real problem, just like aliasing on my digital scope doesn't present a real problem.

 

[edit] YOU MEAN THERE ARE SIX PAGES? Guess I should do a little more reading before I post a comment. Sorry for backtracking so drastically. [end edit]

And not only that it's a necropost that MrBill woke up.