zombie seconds

[goryunov]

Hi Tim,

thank you very much for the code. I am using it with Teensy 3.1 and crystal powered from external 5v source. Having it running for a while I noticed that periodically "zombie" seconds being left on (led left with light on with the same color as second) until full circle for second completes, then it erases the "zombie" and continues. I have not found any pattern in appearance of the additional second light, so just wondering if you experienced the same and if yes might be you have corrected code for it.

Regards,
Andrey

[timpear]

Hi Andrey,

It's great that you've built the clock!

I actually haven't run the clock for an extended time myself, since I gave it to my dad and haven't built a new one yet. But I did notice what you're talking about one time when running version 1 of the code. In fact, the zombie pixels were so bad that it became some sort of hypnotic glitch art -- which is OK in some cases, but not ideal if you just want a clock. It turned out that it was a power issue -- I tried a different power supply, and it seemed better.

BUT -- looking into it further just now, it seems that Adafruit has updated their NeoPixel Überguide since the first time I read it, and they've added a few recommendations about power regulation for larger numbers of pixels, like the strips. They recommend adding a 1000 uF µF, 6.3V or higher capacitor across the + and - terminals, and a 300 to 500 Ohm resistor between the data output and the NeoPixel connector. This will prevent damaging the LEDs -- but I think the bigger issue is that the Teensy 3.1 uses 3.3V data, and the NeoPixel strip is 5V. It seems that the manufacturer recommends that the data is at least 70% of the power voltage, and what we're supplying is 4% short of that. It just means that the pixels might not hear the data signal over the power. I bet it's close enough to work most of the time, but occasionally it misses. That must be it.

https://learn.adafruit.com/adafruit-neopixel-uberguide/power

The solution is easy and inexpensive ($3.95), but unfortunately does require building a new circuit. Adafruit sells a bi-directional logic level converter: http://www.adafruit.com/products/757 which will bump the data voltage up. Here's a wiring guide in their example, going from a 5V board to a 3.3V part: https://learn.adafruit.com/adafruit-1-wire-thermocouple-amplifier-max31850k/wiring-and-test. Here, they add a 4.7K resistor from the data out line to 3.3V -- I'm going to write to a forum to see if that's the same if up-converting, or if it goes on the 5V side in that case. I've never used one of these myself.

I'm also going to ask if there's any risk of damaging the parts if you don't bother with the converter. My guess is that since it's a lower voltage, and especially if you add the other capacitor and resistor to prevent power spikes, that it won't harm the NeoPIxels -- it'll just fail to send the message every so often. If that's the case, then there is a code solution: write each pixel two or three times to increase your chance of success. The only question is if the power resistor will make it even harder for the data to get through. The other question would be if there's a risk of feedback damage to the Teensy, but one of the features of the 3.1 is that all the digital pins are 5 volt tolerant, so I believe it is OK.

All that said, I don't guarantee that the code is perfect. I'm still learning. But this seems more likely given the unpredictability of it.

I'm sorry I didn't know all this before.

I'm working on the code now -- I'll let you know when it's updated.

Best,
Tim

On May 23, 2014, at 3:46 AM, Andrey Goryunov notifications@github.com wrote:

Hi Tim,

thank you very much for the code. I am using it with Teensy 3.1 and crystal powered from external 5v source. Having it running for a while I noticed that periodically "zombie" seconds being left on (led left with light on with the same color as second) until full circle for second completes, then it erases the "zombie" and continues. I have not found any pattern in appearance of the additional second light, so just wondering if you experienced the same and if yes might be you have corrected code for it.

Regards,
Andrey

—
Reply to this email directly or view it on GitHub.

[timpear]

Andrey,

I've added redundancy to turning off the pixels, so hopefully that helps (if it is a power issue). But I did not get to test it on a real clock, so PLEASE let me know if it breaks everything, and use the previous version.

I also made a V3, which erases the entire strip and then writes the pixels out every cycle (10 times per second), so that simplifies the code and should also make any glitches last for 1/10th of a second. On the other hand, it might introduce 1/10th of a second glitches more often…

I'll let you know when I hear back from the forums about the power issue -- mainly if it's cool to keep using the Teensy 3.1 without the logic level converter, if we can fix the problem in code.

Tim

On May 24, 2014, at 5:00 PM, Tim Bartlett timbartlettpear@gmail.com wrote:

Hi Andrey,

It's great that you've built the clock!

I actually haven't run the clock for an extended time myself, since I gave it to my dad and haven't built a new one yet. But I did notice what you're talking about one time when running version 1 of the code. In fact, the zombie pixels were so bad that it became some sort of hypnotic glitch art -- which is OK in some cases, but not ideal if you just want a clock. It turned out that it was a power issue -- I tried a different power supply, and it seemed better.

BUT -- looking into it further just now, it seems that Adafruit has updated their NeoPixel Überguide since the first time I read it, and they've added a few recommendations about power regulation for larger numbers of pixels, like the strips. They recommend adding a 1000 uF µF, 6.3V or higher capacitor across the + and - terminals, and a 300 to 500 Ohm resistor between the data output and the NeoPixel connector. This will prevent damaging the LEDs -- but I think the bigger issue is that the Teensy 3.1 uses 3.3V data, and the NeoPixel strip is 5V. It seems that the manufacturer recommends that the data is at least 70% of the power voltage, and what we're supplying is 4% short of that. It just means that the pixels might not hear the data signal over the power. I bet it's close enough to work most of the time, but occasionally it misses. That must be it.

https://learn.adafruit.com/adafruit-neopixel-uberguide/power

The solution is easy and inexpensive ($3.95), but unfortunately does require building a new circuit. Adafruit sells a bi-directional logic level converter: http://www.adafruit.com/products/757 which will bump the data voltage up. Here's a wiring guide in their example, going from a 5V board to a 3.3V part: https://learn.adafruit.com/adafruit-1-wire-thermocouple-amplifier-max31850k/wiring-and-test. Here, they add a 4.7K resistor from the data out line to 3.3V -- I'm going to write to a forum to see if that's the same if up-converting, or if it goes on the 5V side in that case. I've never used one of these myself.

I'm also going to ask if there's any risk of damaging the parts if you don't bother with the converter. My guess is that since it's a lower voltage, and especially if you add the other capacitor and resistor to prevent power spikes, that it won't harm the NeoPIxels -- it'll just fail to send the message every so often. If that's the case, then there is a code solution: write each pixel two or three times to increase your chance of success. The only question is if the power resistor will make it even harder for the data to get through. The other question would be if there's a risk of feedback damage to the Teensy, but one of the features of the 3.1 is that all the digital pins are 5 volt tolerant, so I believe it is OK.

All that said, I don't guarantee that the code is perfect. I'm still learning. But this seems more likely given the unpredictability of it.

I'm sorry I didn't know all this before.

I'm working on the code now -- I'll let you know when it's updated.

Best,
Tim

On May 23, 2014, at 3:46 AM, Andrey Goryunov notifications@github.com wrote:

Hi Tim,

thank you very much for the code. I am using it with Teensy 3.1 and crystal powered from external 5v source. Having it running for a while I noticed that periodically "zombie" seconds being left on (led left with light on with the same color as second) until full circle for second completes, then it erases the "zombie" and continues. I have not found any pattern in appearance of the additional second light, so just wondering if you experienced the same and if yes might be you have corrected code for it.

Regards,
Andrey

—
Reply to this email directly or view it on GitHub.

[goryunov]

Hi Tim,

I am using capacitor for power and resistor for data pin, so might be as you said it is the issue with Teensy using 3.3V data. I tested new v.2 and got the same zombie seconds and now LEDs that are on with some color flashing when there is second color just left it. And when I tested v.3 - all LEDs which are on are flashing for ~4-5 times per second constantly.
I checked the code and modified eraseSecond function and made it setting all pixels to off. Since all further functions in loop are setting all needed colors I think it should not be a problem in clock functionality. I am having that modified code running for now and have not noticed any of zombie LEDs yet.

Thanks,
Andrey

[timpear]

Andrey,

Thanks for letting me know about it! I hate to post code without running a test, so I'm glad you got it to work with erasing all the pixels. I'll get rid of that v.3 since it clearly doesn't do what it should.

Did you include strip.show() in the new code to erase all the pixels in eraseSecond? Or just set them all to 0,0,0 there, and wait until after all the other pixels are set to show?

It sounds like as long as it's working, the Teensy won't damage the NeoPixels. Here's what I heard back from Adafruit:

The newer "B" version of the WS2812 is a little more fussy about voltage levels than the "A" version. They do not like 3.3v logic when powered by 5v. You won't damage them, but operation will be unreliable.
The 74AHCT125 has been tested to work well for driving Neopixels from 3.3v devices such as the Teensy.

Thanks,
Tim

On May 25, 2014, at 12:42 AM, Andrey Goryunov notifications@github.com wrote:

Hi Tim,

I am using capacitor for power and resistor for data pin, so might be as you said it is the issue with Teensy using 3.3V data. I tested new v.2 and got the same zombie seconds and now LEDs that are on with some color flashing when there is second color just left it. And when I tested v.3 - all LEDs which are on are flashing for ~4-5 times per second constantly.
I checked the code and modified eraseSecond function and made it setting all pixels to off. Since all further functions in loop are setting all needed colors I think it should not be a problem in clock functionality. I am having that modified code running for now and have not noticed any of zombie LEDs yet.

Thanks,
Andrey

—
Reply to this email directly or view it on GitHub.

[goryunov]

Tim,

it is simply just setting 0,0,0 to all of them

void eraseSecond(int s) {
for (int i = 0; i < 60; i++) {
strip.setPixelColor(i, 0, 0, 0);
}
}

I will current circuit as it is but I got Trinket 5V as well, so might be will try transistor with it for next clock if there are any "zombies"

Thanks,
Andrey