FS: Kenometer USB Firmware Upgrade / 0.1mW Precision / $30

[lasersbee]

That's not noise. That's the ADC flipping between states, despite the voltage being the same. All ADC's do this. The variation in those readings has absolutely nothing to do with noise.

The only difference between those two datasets is the emulation of ADC bits for higher precision. There is zero line noise in that signal.

Emulation is a widely accepted technique to squeeze more precision out of an ADC, while sacrificing some sample speed. So in short, yes, I did increase precision.

Hope that clears it up.

-Trevor

Sorry for the double post...
Yep... once you stated that....that cleared it up.. Thanks

Jerry

 

[MarioMaster]

Please do forgive my poor eyesight. Maybe you could be so kind as to point it out for me?

And you are correct, you haven't sent me any of your meters but that doesn't mean I haven't seen them.

 

[lasersbee]

Things aren't going to well MM... now your memory is affected...:evil:

Kenometer Pro Review

Trevor even posted it above....
Heroic found it on just a photo that you posted....

Jerry

 

[MarioMaster]

Heroic took a guess. So Heroic is correct in that the SOT23 component is the ADC? That's all you had to say.

 

[Trevor]

Things aren't going to well MM... now your memory is affected...:evil:

Kenometer Pro Review

Trevor even posted it above....
Heroic found it on a just a photo that you posted....

Jerry

Okay, for pure curiosity's sake... just want to clear this up.

LB I
12-bit ADC, measured up to 1.05W in 1mW increments (from your site - correct if wrong)
-The ADC technically offered 0.26mW steps.
-I would've cut it off at 1mW too. There's no pretty way to use more of that resolution.

LB II
16-bit ADC, measures to 1W in 0.1mW increments, to 3200mW in 1mW increments.
-The ADC offers steps below 1W in ~0.015mW increments.
-The ADC offers steps across the whole range (up to 3.2W) in ~0.049mW increments.

That's where I'm slightly confused. I'd have maybe made resolution below 1W to 0.05mW - but I do understand why you chose 0.1 (0.05 isn't super pretty). But with no line noise, wouldn't it be quite doable to get 0.1mW resolution out of it up to 3.2Wr?

-----

Actually, I just went and looked at one in action in one of DTR's threads... did you cut it off at 1mW resolution due to running out of room on the LCD screen?

-----

Dunno, just curious I guess. As a programmer, the Laserbee II just screams "potential" at me.

-Trevor

 

[lasersbee]

You are correct Trevor...
and yes we ran out of room on the LCD...
Good catch...

@MM... it's no fun if you are handed all the answers...
Heroic was correct...

Jerry

 

[willpine]

Hey It's good to see you back LaserBeeee! I was wondering what you would have to say about this, i read this the other day. It seems you all are very knowledgeable about this subject; I like to hear different point of views and nothing like good argument to stimulate the ole' brain.

 

[Trevor]

In other news, I've got a copy out to someone to test and analyze.

-Trevor

 

[willpine]

Dunno, just curious I guess. As a programmer, the Laserbee II just screams "potential" at me.

-Trevor

So Trevor do you think you can make tweaks and program laserbees's, and do you think you can program Kenoms or laserbees builds to use other sensors?

 

[Trevor]

So Trevor do you think you can make tweaks and program laserbees's

Given access (this has been discussed; won't happen) and the ability to flash the microcontrollers (which I don't have), then yes I could.

and do you think you can program Kenoms or laserbees builds to use other sensors?

For the Kenometers, yes. For Laserbees, see above.

-Trevor

 

[Bluefan]

...Even if the linearity of the ADC changes by a relatively large amount (+2 LSB, yielding a variation in readings of +/-19.5mV), a precision of +/-0.3-0.4mV will be retained. That's a worst case scenario. ...

My question is on this point: if the differential nonlinearity is 1/2 LSB, then the step size of the 4096 steps of the ADC won't be 1/4096 of the full range, but can be as low as half of that. Interpolating between counts will not lower this error, you'll have an interpolation with a proportional smaller or larger stepsize.
So if the nonlinearity is 19.5mV, how can you claim 0.3~0.4mV precision, if your ADC can be as much as 19.5mV off without indication? It's a systematic error.
I don't have to see your source code, but I'm definately interested in the theoretical explanation on how to solve this problem.

 

[Trevor]

My question is on this point: if the differential nonlinearity is 1/2 LSB, then the step size of the 4096 steps of the ADC won't be 1/4096 of the full range, but can be as low as half of that. Interpolating between counts will not lower this error, you'll have an interpolation with a proportional smaller or larger stepsize.
So if the nonlinearity is 19.5mV, how can you claim 0.3~0.4mV precision, if your ADC can be as much as 19.5mV off without indication? It's a systematic error.
I don't have to see your source code, but I'm definately interested in the theoretical explanation on how to solve this problem.

I've shown precision as high as 0.1mV in a range that "should" resolve to 4.88mV. My DMM agrees with the readings acquired.

What I did with the battery on the original firmware demonstrates a 1-bit error. Now look at the battery dataset on the new firmware. That's a 1 LSB error that is effortlessly attenuated. 2 LSB would be only one step above trivial.

As for linearity, the simple answer is that I can't account for it - but neither can the original firmware. But I'd still happily pit my Kenometer USB with this firmware against a calibrated LPM.

If you address nothing else in my post, address this:

Why do you even care? My target market is Kenometer USB (and eventually Pro) owners. This firmware is designed to add a level of precision that is not present when using the stock firmware. It does that. With flying colors.

Now consider this: if I wasn't limited by hardware, and I had a platform with a stock 12-bit (or more) ADC without line noise, imagine what some good programming could do. Stay tuned.

-Trevor

 

[Bluefan]

You've shown a resolution on 0.1mV, not a precision of 0.1mV. Or I missed the DMM reading, you state "The battery is quite obviously outputting 1.6088V" a few posts back, What DMM did you use?
"As for linearity, the simple answer is that I can't account for it - but neither can the original firmware."
I think that means the 0.1mV doesn't have a 0.1mV precision. Unless I'm missing something here.

The noise reduction alone is worth the upgrade. I'm just interested in how stuff works, that's why I ask. Does that "stay tuned" promise something?

 

[Trevor]

You've shown a resolution on 0.1mV, not a precision of 0.1mV. Or I missed the DMM reading, you state "The battery is quite obviously outputting 1.6088V" a few posts back, What DMM did you use?

My Craftsman one. Ran me something like $40 or $50. I think. :thinking:

Photos later if you actually care...?

I think that means the 0.1mV doesn't have a 0.1mV precision. Unless I'm missing something here.

It's demonstrated 0.1mV precision on low-power HeNe's that have been verified on calibrated LPM's. I've been unable to test other ranges, but I'm optimimistic.

Like I said, I'd be perfectly willing to put my development unit head-to-head with a calibrated LPM to find out.

The noise reduction alone is worth the upgrade. I'm just interested in how stuff works, that's why I ask.

I'm sure that's not your only reason for being here.

Does that "stay tuned" promise something?

Yes.

-Trevor

 

[Bionic-Badger]

Why do you even care? My target market is Kenometer USB (and eventually Pro) owners. This firmware is designed to add a level of precision that is not present when using the stock firmware. It does that. With flying colors.

We care because when somebody makes seemingly outrageous claims, we'd like to see some evidence to back it up. So far, the "evidence" you've provided comprises data-points that have higher numerical precision, and graphs that are quite a bit smoother than the original firmware. Neither of those imply that your techniques improve actual precision of the meter. The graph itself could easily be imitated using a simple averaging technique and just outputting higher precision numbers for the data.

I'm not saying your claims are snake oil. There are good techniques for improving dynamic range (sigma-delta-like techniques, dithering, etc.) that you may be employing. Even ignoring precision, simply using an averaging algorithm would improve the output for of the firmware as a worthy upgrade.

I realize the techniques you're employing may seem like trade secrets, but disclosing a general description of your techniques (or at least the theory behind it) has its benefits too. It gives your product credibility, far more than just your word. Plus, it would settle doubts from skeptics who may influence potential buyers.

 

[Trevor]

We care because when somebody makes seemingly outrageous claims, we'd like to see some evidence to back it up. So far, the "evidence" you've provided comprises data-points that have higher numerical precision, and graphs that are quite a bit smoother than the original firmware. Neither of those imply that your techniques improve actual precision of the meter. The graph itself could easily be imitated using a simple averaging technique and just outputting higher precision numbers for the data.

I think this is the third time I've mentioned this. The outputs of my newer gas lasers (this is in the low, low, low range of course) perfectly match what the seller measured on their calibrated LPM. This is currently the best I can do.

What this shows is that the Ophir sensor is well calibrated, and I'm just cutting through the noise and acquiring precision.

I realize the techniques you're employing may seem like trade secrets, but disclosing a general description of your techniques (or at least the theory behind it) has its benefits too. It gives your product credibility, far more than just your word. Plus, it would settle doubts from skeptics who may influence potential buyers.

My general description of how it works is just my word too. Once again I'll reiterate that I'm now waiting on field testing results (with a customer - not an early tester). That will be the most telling.

I'm working on proving the benefits of the upgrade by securing a test from a trusted, neutral member who has a calibrated professional LPM.

I might even settle for a head-to-head test with a Laserbee on principle.

Explaining the method proves nothing.

-Trevor