Kenometer Pro Review

[Mario]

That's no good for measuring time series power readings, for example to determine stability.

You can get 24-bit metrology ADCs in SOIC8s for less than 5 euro a piece. You can put them on a separate power and ground plane, far from the uC, and you can optoisolate them. They have integrated digital filters that will reduce noise to below measurable limits. They are far, far more accurate than any galvo movement at ten times the price, and they never need to be recalibrated as they never drift. (They do need temperature compensation, but you can do this easily with a couple of ten cent temperature sensors.)

The world of metrology is a digital world now. Mostly, it's better for it.

Uhm, a common single 5V 24bit A/D converter has a resolution of:

Ad_resolution = (Ad_Vin ) / AD_Bits ^2 so Ad_resolution = 5 / 24 ^2 = 0.008681

Then you have a small Ophir Head thas has a typical accuracy of 3% in a range from 60
µW to 3 W with an integration time of 1.8 S

The rest of the math is very simple...a 24 bit A/D is just an overkill.

Analog conversione and/or galvanometer / 10 € chinese cheap DMM will do the job with an accuracy within the ophir head specification.

- M

[greenalien]

I don't suppose anyone's prepared to sell just the ophir thermopile heads? I can see them being ideal candidates for a group buy, if someone would be prepared to source a few for the forum...

[tocket]

I was a little surprised to see a full arduino nano inside of the enclosure. Those are rather expensive. You could easily get away with less by putting an arduino compatible circuit on your own PCB. Here's one example: http://www.instructables.com/id/Perf...tible-circuit/

I really like the fact that this meter ships with an ophir head. Having seen the degree of engineering that goes into making a good thermal sensor on the ophir sensor I have at work (though that one measures down to 60 µW), I think it's hard to get the same accuracy from a TEC painted black and glued onto a heatsink.

The by far most important task for the actual meter is to be able to measure the voltage from the sensor accurately, which means that a lot of effort needs to be put into ensuring that it does just that. I don't know much about electronics, but I think heroic's suggestion to use a real ADC makes a lot of sense, since this is the component that does the voltage measurement. I've used the ADC on AVRs before and I would not trust it for this application.

I agree that we should not expect the same performance and accuracy from these hobbyist meters as from professional meters, but I don't think it's impossible to get close.

One thing that upsets me though is the inadequate specifications on error and accuracy. This goes for all hobbyist LPMs. At the very least you should specify the power range and power accuracy correctly. The power range is the range of powers for which the power accuracy criteria holds. This determines the useful range of the meter. Quickly looking at the laserbee graph mccarrot posted it is evident that either the power accuracy of this device is poor or the power range is incorrectly specified.

I would also be interested to know the linearity, response time, spectral range and thermal drift before I buy a meter. The device should of course also be calibrated and tested for conformity with at least the power range and accuracy criteria at two (or more) measuring points at the extremes.

[heroic]

]Stop hiding behind the curtains and tell us what you are using, if you think your meter are really that special. Otherwise, I see no proof that you aren't using an internal ADC either.

That SOT23 component in the top right of the board picture that was shown looks like an AD7476. Guaranteed linearity is important. You don't get that with uC integrated ADCs. AD7476 is a 12-bit successive approximation ADC with 1.5LSB guaranteed maximum nonlinearity at full scale. 70 dB SNR.

AD7476 costs five bucks.

Laserbee, am I right?

Even at the best settings, the resolution of the AVR ADC is only 10 bits. It has 2LSB of error, max of 3LSB. It has a nonlinearity of 0.5LSB typical or 1.5LSB max- but remember, those bits are *eight times bigger* than the bits on the AD7476. It has a gain error of 2.5 bits (1.5 typical).

The ATmega32 datasheet does not specify the SNR of the ADC; probably because it's pretty poor. Since you're using an Arduino, AVcc is coming from... where exactly? The logic plane? Dude, that's just *bad news*.

I don't care if you are trying your best. You wrote "Pro" on the box, and this is pretty fucking far from a professional quality instrument. If you're going to play at being an EE, you need to accept that when real EEs look at your hack job, they don't find it funny that your crap metrology skillz are reflecting poorly on an entire profession. A profession that some of us take seriously.

As for the two of you who have said "if you can do better, why don't you?" What do you think I get paid to do all day every day? Twiddle my thumbs and drink coffee?

[Things]

Calling it "Pro" can simply mean it's one of the better meters in his series, it doesn't mean its comparable to a $>1500 meter.

Like laserbee's meters, "laserbee deluxe", does that mean it's better than all the other meters out there?

[ElektroFreak]

I don't suppose anyone's prepared to sell just the ophir thermopile heads? I can see them being ideal candidates for a group buy, if someone would be prepared to source a few for the forum...

The ones used for this meter and nospin's 5W meter were very cheaply bought as surplus, but that source has dried up I believe. Presently these are the cheapest ones I can find.. all you need is a DMM to read them.

[heroic]

Calling it "Pro" can simply mean it's one of the better meters in his series, it doesn't mean its comparable to a $>1500 meter.

Like laserbee's meters, "laserbee deluxe", does that mean it's better than all the other meters out there?

You and I both know you're stretching there.

If you built the Kenometer Pro as a class project in a high school electronics class you'd fail dismally.

[Things]

And ...?

If you don't like it, it's as simple as not buying it.

Kenom has 3 versions, the regular Kenometer, Kenometer USB, and the Kenometer Pro.

I think the naming is fine, unless you really want it to be called "The Kenometer with a graphic LCD and computer control functionality"

The Pro simply means it's one of the ones with more features than the previous meters, does it not?

[RedlumX]

Guys,

I think one shouldn't overestimate the possible accuracy of power meter. Even when factory new, the good ones from eg Newport etc are not more accurate than a few percent. And to keep this, one needs to have the meter recalibrated every year or so. I have seen meters that were off by more than 20% whose last cal date was just a few years back. I guess to hope that with any sort of DYI sensor one could better do than 10% is unrealistc (unless very lucky). Just vary the beam size and location over the sensor and you typically see already a variation by a few %, and this even for commercial sensors (I happened to have one of those Ophirs in the past as well). And thermal sensors are anyway very sensitive to background IR, just walking next to the table of the meter can make the reading change by a few mW.

So all-in-all, there is no need to involve 16bit ADCs etc, this just creates a false impression of accuracy. If you want to be realistic, aim at 10% unless you really have a NIST calibrated transfer standard and regularly recalibrate (even transferring a calibration from sensor A to sensor B is tricky and easily buys you a few percent of inaccuracy unless extremely careful - this involves a very controlled environment, and at least a parallel monitoring of laser power - recall that lasers often fluctuate in their power due to mode cycling etc).

There is a very good reason why commercial meters are so expensive - and its not in the electronics!

[lasersbee]

Guys,

I think one shouldn't overestimate the possible accuracy of power meter. Even when factory new, the good ones from eg Newport etc are not more accurate than a few percent. And to keep this, one needs to have the meter recalibrated every year or so. I have seen meters that were off by more than 20% whose last cal date was just a few years back. I guess to hope that with any sort of DYI sensor one could better do than 10% is unrealistc (unless very lucky). Just vary the beam size and location over the sensor and you typically see already a variation by a few %, and this even for commercial sensors (I happened to have one of those Ophirs in the past as well). And thermal sensors are anyway very sensitive to background IR, just walking next to the table of the meter can make the reading change by a few mW.

So all-in-all, there is no need to involve 16bit ADCs etc, this just creates a false impression of accuracy. If you want to be realistic, aim at 10% unless you really have a NIST calibrated transfer standard and regularly recalibrate (even transferring a calibration from sensor A to sensor B is tricky and easily buys you a few percent of inaccuracy unless extremely careful - this involves a very controlled environment, and at least a parallel monitoring of laser power - recall that lasers often fluctuate in their power due to mode cycling etc).

There is a very good reason why commercial meters are so expensive - and its not in the electronics!

I agree with this statement....
There are basically 2 types of Thermoile sensors the Radial Thermopile type
such as the OHIR 20C-A and the Wafer Thermopile type similar to what is
used by the Laserbee products.

We have done some in-house tests on those OHIR 20C-A heads... We noticed
that if you shine your (2mm dia. for testing) beam exactly in the center of the
12mm dia. active Sensor area the readings a quite accurate... but if you stray
off center... the readings vary quite a bit...

Here is the results we obtained testing the four OHIR 20C-A heads... and they
all exhibited similar variations...



We performed the same tests over the entire active surface area of the LaserBee II
Thermopile Sensor at the same Laser powers and the reading did not change with
different Active area locations.

It should not be a big problem if the user of the Radial type is informed by the seller
of this reading issue...


BTW... For the LaserBee II to be able to reliably read and display our Thermopile
Sensor's output to within 0.1mW.... we required a 16bit ADC to achieve this...

(We test in a stable temperature controlled environment with test equipment
and Lasers warmed up and stable for at least an hour before performing any
tests... We try to keep the air currents around the Test Area to an absolute
minimum...)


Jerry