I was going to do some tests on a couple of 6W Chinese PAR16 and PAR20 bulbs. They use 3 x 2W Cree LEDs. I hooked the thermocouple to the heat sink and got a very bogus temperature reading (80 degrees C when the bulb was at 25C) when I switched it on. Same with the thermocouple electrically insulated from the heat sink. Looks like those bulbs are throwing off a lot of electrical noise that is corrupting the temperature readings. I am going to try some more robust insulation…
No problems measuring my Sylvania/Philips/LSG bulbs.
Well, don’t buy the 16” sphere from Barnards unless you like shaping styrofoam… the inside is not spherical! There are 6” diameter flat spots at the top of each half where the wall was made thicker. A little quick measuring shows you would need to remove about 0.56” of material at the peak if you were to sand the inside to a true spherical shape.
Also, the sphere is actually 15.625” in outside diameter.
Here is a screen image of a run done on the hyper-mega-uber-wonderful-if-you-didn’t-get-one-your-life-sucks E1320 modded UF2100 running on high (not turbo) Battery was a Panasonic CGR18650EA.
Ignore the LED numbers in the upper left hand corner… those sensors are not connected.
The center line of the graphs is marked >…< Those are the REFerence values for the plots. The RMS values are the RMS ((root mean square) think of it as an average value) of the plotted curves. The values below the “cursor time” line are the recorded values where the mouse cursor was (mouse cursor not shown on the screen dump).
The LUX values shown are very close to the lumen output of the light.
Now to the interesting stuff… see those noise bits in the battery current and power plots? Me too… at first I thought that I had bobbled some of the wiring during the run. But NOOOO! I redid it and got the exact same thing. The current drawn from the battery by the driver is actually going whacky. Note that the light output curve (lux) stays constant. The current regulators are regulating the current to the LED. I wonder what the driver is doing to cause the current to do that at those points?
Gratuitous feature… sun and moon position/phase are shown on the clock face… you can work in your deep and dark modding dungeon and know when it’s safe to go out and shine your shiny things around. If the moon is greyed out or the sun is shown as a circle, it’s below the horizon…
“The current regulators are regulating the current to the LED”… Quite impossible, well, at least if thats a linear regulator like I think it is in that light… I would still suspect that’s noise from something - not sure what, mind you, but I don’t think it’s real… Especially since BATv doesn’t wiggle with ‘supposedly’ increased current draw - think about it, if you suddenly drew a bunch of extra current from the battery, voltage would surely sag at least a little - but that line is rock solid steady…
Those spikes are only a couple of hundred mA at most and those batteries can do over 10A. That might not show up in the battery voltage. Also, I’ve run it four times now and get the same spikey stuff in the same places, so it’s not random noise. It could be a wonky current sensor… I’ll try another one and see what happens.
I tried another current sensor and get the same noise burst at the beginning of the plot… I’m letting it run and will see what it looks like towards the middle.
Interesting… Still… Where’s the current going? The Regulator (It is linear, right?) can’t ‘drop’ current… Current at all points in the circuit must be equal - meaning that the LED would also be seeing those current spikes, yet your lux isn’t suggesting that. Unless, I suppose, the lux sensor is slow and doesn’t see those quick spikes…
Dunno… Weird… But if that current is real, it has to be effecting something…
I guess the other question is your ADC Reference… Are you turning something else on/off during those periods that would futz with your ADC +VRef or AVSS? That could pretty easily be the ground or adcVref bounce due to an output being turned on/off…
Any glitches on the ADC ref would affect and show up in all the other analog readings (i.e the battery voltage).
I can filter the display by any number of seconds. Those glitches still show up with a 60 second filter…
The lux sensor integrates the readings over a 250 msec interval (well probably around 150 msecs, it is read out every 250 msecs, 4 readings are averaged and a value is output every second). The lux readings are very smooth though…
The current regulator could be sending a regulated current to the LED and burning off the excess as heat. The only other explanation seems to be the current sensor is oscillating, etc at particular readings. I don’t think so since I ringed it out pretty extensively when testing its accuracy and saw no funkiness. Another possibility is something caused by the maze of clip leads it takes to get the light cobbled onto the board… maybe induced by the driver PWM.
I’m putting put one of the 20A sensors in there and see if anything changes.
Definately strange… I’m still working on the assumption that this is a linear driver, which means that the regulator can’t ‘burn off’ current - only voltage… There’s no such thing as burning off current - a linear regulator (whether current or voltage feedback) can only burn off excessive voltage - current in will still equal current out… It’s noise - it has to be. With as quickly as you’re sampling the lux sensor, you’d be seeing variations if there was any increase in LED current, and you’re right, all of the analog readings woud be doinked if it was reference bounce…
It HAS to be either noise or funkiness in the current sensors at those values… Unless the assumption i’m working on is wrong and this is using a switch mode regulator - if that’s the case, all bets are off… You’ve got a current clamp, if memory serves… toss it around the battery leads and watch with an oscilloscope… When the readings get funky - does the current really bounce around like that on the scope?
Yes, current in does equal current out… but that current out can be split into two paths (the LED and ground). After thinking about it and checking (it uses AMC7135 regs) I don’t think that is what’s happening.
I’m going to break out the heavy equipment this weekend and give it a probing it’ll never forget.
I replaced the 5A current sensor with a 20A one and saw the same thing. But it looks like the recorded amplitude of the glitches is maybe twice as big with the 20A sensor. The 5A sensor outputs 185 mV/A and the 20A one does 100 mV/A. That seems to indicate the noise or whatever is being coupled to the output of the sensor.
I did another run with the 20A sensor and looked at the data more closely. The amplitude of the glitches with the 5A and 20A sensors are pretty much the same. So maybe it is coming in on the input to the sensor…
Woo hoo! I think that I have found the source of the current glitches.
I stripped everything down to add some test points to the little Chinese current sensor board. In the process, I removed the thermocouple from the flashlight head. It was rubber-banded to the light under a styrofoam packing peanut. There should have been a piece of Kapton tape electrically insulating the thermocouple from the flashlight body… nope, not there, damn underpants gnomes will steal anything.
Anyway, did another run with the thermocouple off the light. No glitches… Now I need to see what it does with the thermocouple on the light, but electrically insulated from the body.
Knew it had to be noise from something… Those current glitches just couldn’t be real… Glad you found the source, let me know if you want to go underpants gnome hunting - it’s one of my favorite sports.
