xStatiCa
0
- Joined
- Feb 7, 2010
- Messages
- 360
- Points
- 0
Make sure to place a new post in this thread with the pro version is available.
LPF Donation via Stripe | LPF Donation - Other Methods
Links below open in new window
ArcticMyst Security by Avery
FS: Kenometer USB Firmware Upgrade / 0.1mW Precision / $30
- Thread starter Trevor
- Start date
- Joined
- Sep 20, 2008
- Messages
- 17,622
- Points
- 113
Just a question Trevor....
What is the ADC resolution on the USB Kenometer...
I understood that the arduino uses a 10bit ADC...
Correct me if I'm wrong... I just may be wrong...LOL
Jerry
What is the ADC resolution on the USB Kenometer...
I understood that the arduino uses a 10bit ADC...
Correct me if I'm wrong... I just may be wrong...LOL
Jerry
Trevor
0
- Joined
- Jul 17, 2009
- Messages
- 4,386
- Points
- 113
- Joined
- Sep 20, 2008
- Messages
- 17,622
- Points
- 113
Have you added any other electronic circuitry to the Kenometer USB
IE... another higher resolution ADC or something of that nature???
Maybe I'm a noob at this... but I'm having trouble to see how
you can get 0.1mW precision over 5Volts with a 10bit ADC...
5000/1024= 4.88mV per bit...IMO
And in the case of the OPHIR head (1mV=1mA) that translates
to a resolution of 4.88mW per bit.
Like I said maybe I'm not seeing something clearly... correct me
if I'm wrong..
Jerry
Trevor
0
- Joined
- Jul 17, 2009
- Messages
- 4,386
- Points
- 113
Have you added any other electronic circuitry to the Kenometer USB
IE... another higher resolution ADC or something of that nature???
Maybe I'm a noob at this... but I'm having trouble to see how
you can get 0.1mW precision over 5Volts with a 10bit ADC...
5000/1024= 4.88mV per bit...IMO
And in the case of the OPHIR head (1mV=1mA) that translates
to a resolution of 4.88mW per bit.
Like I said maybe I'm not seeing something clearly... correct me
if I'm wrong..
Jerry
No components were added. Think outside the box. There are tons of papers authored on this subject, one of which was written by yours truly.
I'll post a video of measurements being taken on the different firmware versions to show the difference between the original version and the new one, and to demonstrate the added precision.
-Trevor
Last edited:
- Joined
- Sep 20, 2008
- Messages
- 17,622
- Points
- 113
Last edited:
Trevor
0
- Joined
- Jul 17, 2009
- Messages
- 4,386
- Points
- 113
This is a research university, so a lot of the work done by faculty and students ends up as the university's property. I mentioned my work to show you that someone with as little training as a college sophomore in the field can learn to write code that will do this. This does, of course, assume that the fact that I can produce graphs that show this code in action doesn't convince you.
Seriously, Google it.
Specifically, the 10-bit ADC in conjunction with my code has more precision than a 15-bit ADC.
-Trevor
Last edited:
- Joined
- May 4, 2009
- Messages
- 5,443
- Points
- 113
Trevor
0
- Joined
- Jul 17, 2009
- Messages
- 4,386
- Points
- 113
Flaminpyro: Thanks, will do.
As promised, here's my argon at low power being metered with the old and new firmware.
Before
After
Any questions?
-Trevor
As promised, here's my argon at low power being metered with the old and new firmware.
Before
After
Any questions?
-Trevor
Bluefan
0
- Joined
- Aug 15, 2009
- Messages
- 1,443
- Points
- 48
I still have doubts. You said you have written a paper, please provide a link. A least your name will be on top. I can acces nearly anything that's published.
Second, you can't create precision where it isn't. Your 10 bit ADC can't reach 15 bit ADC precision without knowing the noise distribution, do some complicated averaging and know EVERY error from every bit in your ADC, and most ADC have a differential nonlinearity of at least 1/2 LSB.
A 10 bit ADC for a 5V signal gives a 4.88mV/bit. Some averaging if there is a noise present with the right statistics can bring this down a bit, but I doubt to 0.1mW. Also because the ophir thermopiles have a noise power of 0.5mW and probably a silimar drift. You can filter in time to get more precision at the expensive of a fast response, but drift can't be compensated.
So please, elaborate before you claim more than the ADC and thermopile is spec'd for. I can believe you obtain 0.1mW resolution by a slow filter (the drift isn't very big if you let the thermopile stabilize, I'd say 0.2mW~0.3mW from what I've seen), but that ADC stuff requires more explanation.
Edit: you've provided some graphs, but none with a resolution of 0.1mW. Would be nice to have one of a HeNe laser and compare it with a good photodiode based power meter.
Second, you can't create precision where it isn't. Your 10 bit ADC can't reach 15 bit ADC precision without knowing the noise distribution, do some complicated averaging and know EVERY error from every bit in your ADC, and most ADC have a differential nonlinearity of at least 1/2 LSB.
A 10 bit ADC for a 5V signal gives a 4.88mV/bit. Some averaging if there is a noise present with the right statistics can bring this down a bit, but I doubt to 0.1mW. Also because the ophir thermopiles have a noise power of 0.5mW and probably a silimar drift. You can filter in time to get more precision at the expensive of a fast response, but drift can't be compensated.
So please, elaborate before you claim more than the ADC and thermopile is spec'd for. I can believe you obtain 0.1mW resolution by a slow filter (the drift isn't very big if you let the thermopile stabilize, I'd say 0.2mW~0.3mW from what I've seen), but that ADC stuff requires more explanation.
Edit: you've provided some graphs, but none with a resolution of 0.1mW. Would be nice to have one of a HeNe laser and compare it with a good photodiode based power meter.
- Joined
- Sep 20, 2008
- Messages
- 17,622
- Points
- 113
Flaminpyro: Thanks, will do.
As promised, here's my argon at low power being metered with the old and new firmware.
Before
After
Any questions?
-Trevor
I have no doubt that you are an academic as you stated...
But from the graphs that you posted...
All I see is that you reduced the inherent noise that the
Pro and USB LPMs produce by averaging the total detected
value + noise...
I did a similar thing with my Kenometer PRO with a Resistor
and Capacitor...
Like I said you probably have more knowledge about this
than I do... but 1+1 still adds up to 2 for me...
Jerry
Last edited:
Trevor
0
- Joined
- Jul 17, 2009
- Messages
- 4,386
- Points
- 113
...okay, pretend I didn't mention that. I only threw that in there to communicate to lasersbee that there's plenty of information out there on this topic.
Bits can be emulated with a large sample size. This has to do with the way an ADC works. I'm not posting a LMGTFY link, sorry.
I'll address most of this in my next segment, but please note that below 1W the reference voltage being used is 1.1V.
Simple explanation:
1. Take large sample and preprocess as it streams in.
2. Process dataset using logic (thermopile response time, baseline level of electrical noise, etc).
3. Output average of processed dataset.
And yes, to get full advantage of the code, the thermopile's got to be stabilized.
Got a Coherent Lasercheck for comparative purposes?
lasersbee said:
I assure you I didn't add any components, and did not simply average the values. With this code in place, your modified Pro would achieve an absurd level of precision.
-Trevor
- Joined
- Sep 20, 2008
- Messages
- 17,622
- Points
- 113
...okay, pretend I didn't mention that. I only threw that in there to communicate to lasersbee that there's plenty of information out there on this topic.
Bits can be emulated with a large sample size. This has to do with the way an ADC works. I'm not posting a LMGTFY link, sorry.
I'll address most of this in my next segment, but please note that below 1W the reference voltage being used is 1.1V.
Simple explanation:
1. Take large sample and preprocess as it streams in.
2. Process dataset using logic (thermopile response time, baseline level of electrical noise, etc).
3. Output average of processed dataset.
And yes, to get full advantage of the code, the thermopile's got to be stabilized.
Got a Coherent Lasercheck for comparative purposes?
I assure you I didn't add any components, and did not simply average the values. With this code in place, your modified Pro would achieve an absurd level of precision.
-Trevor
You posted that YOU wrote a Paper on this subject....
Why can't we see it... Did you even write one or was that
just bragging...
I would be proud to show off a "technical paper" had I written
one...
To get at least 0.1 mV true (unmanipulated) precision with
the OPHIR Head and 10bit ADC... you would need to limit the
input Voltage to the ADC to 102.4mV. (102.4mV/1024= 0.1mV per Bit)
That would mean the LPM could only mearasure to 102.4 mV
max with a 0.1mW precision... (1mV=1mW)...
Have you not verified your "processed" info against a known
calibrated LPM to see if your data-set is indeed valid..
I have looked at the chart in your 1st post and it does not
seem look right to me... it looks like data was somehow invented...
http://www.safelasers.org/Luminosity/Images/Luminosity 2.0 2010.10.20.01.png
I wasn't trying to say that you had added components but that
I had... and those components got rid of the noise...
Instead of reading up to 6mW of noise with no laser on the PRO
Sensor it now holds a steady 0mW with no Laser Input.
That I will believe when I see it and compare it with a Calibrated
Lab LPM... and/or have read and agree with your "paper"....
BTW I only colored my responses to keep them in order...
Jerry
Last edited:
Trevor
0
- Joined
- Jul 17, 2009
- Messages
- 4,386
- Points
- 113
It's a document that I got research credits for within my major. Because of this, I do not own the rights to it and cannot distribute it. I shouldn't have mentioned it. Seriously, drop it.
As flattered as I am that you think I'm the only reliable source on this topic, why don't you try reading some of the other source material that's online? I can tell from your replies that you have not.
I know of at least one document put out by a semiconductor company that describes every gory detail the process of emulating ADC bits. I would think they know better than anyone on this forum.
And no, I won't post a link to it. You have to do your own product research.
Those are two separate datasets that were taken with a firmware update in between. It is not "invented."
That's insulting.
Ah.
That I will believe when I see it and compare it with a Calibrated
Lab LPM... and/or have read and agree with your "paper"....
You're free to purchase the Pro firmware upgrade when it comes out. When you post an analysis, I won't even say that you tried to invent data. :tired:
I'll let my product speak for itself.
-Trevor
- Joined
- Sep 20, 2008
- Messages
- 17,622
- Points
- 113
That's fine... that clears up the "There are tons of papers authored
on this subject, one of which was written by yours truly."
Yes... you are right... I have not searched the entire web... Yet...
My thought was to go straight to the horse's mouth... after all you did
write the firmware so I thought you should know how it does what it
does. You are claiming 0.1mW precision...
I'm not looking to do research on a product... I am just curious to know
where you read that you can get 15Bit -16bit resolution with a 10bit ADC....
No... you don't need to post a link... I'll look it up... Would you know the
Search subject I should use ???
Sorry... I really didn't mean to insult you... my apologies.... It was not
directed at you personally...:beer:
I was trying to say the software "invented/manipulated" the original
data to something that looks odd on the graph in your fist post.
If you look at the original data RED line at ~00:00:02 it dips to
~80mW... The BLUE line at the same time slot goes to ~104mW...
That's one he!! of a difference... It happens again at 00:00:11 and
00:00:21...
Just looking at that chart... it looks like the firmware finds the noise
factor value... then seems to add that to the overall average of the
non modified actual data with noise...
That is what I meant by invented data... Somehow it just doesn't look
right to me...:undecided:
That's why the only way to see if the output is indeed valid is with a
known calibrated LPM as a base.
Yeah... I will probably buy it for my PRO... since I can't get any updates
for it anyway... nor do I expect any...:crackup:
Jerry
Bluefan
0
- Joined
- Aug 15, 2009
- Messages
- 1,443
- Points
- 48
Bits cannot be emulated with a large sample size, this has to do with the way an ADC work. I'm posing a LMGTFY:Bits can be emulated with a large sample size. This has to do with the way an ADC works. I'm not posting a LMGTFY link, sorry.
Let me google that for you
sorry
That would give the 1mV resolution
A difficult one would probably be better.
How do you process it? A simple low pass filter?
What logic do you use? Noise isn't something you can just "remove".
If you average here, what can you possibly process more?
Still won't allow you to get better then the ADC's nonlinearity. And those are usually around 1/2 LSB
I don't have a kenometer (I do have thermopiles) or I would do it for you, but I can make you a decent photodiode based LPM with nanowatt resolution, more than good enough to test the accuracy of that 0.1mW resolution.
Then what did you do? Information can't be created, you don't just have the quantization noise but also it's systematic error. Considering each ADC is different, you can't do better then it's specs. That you can put out 0.1mW resolution doesn't mean it has 0.1mW accuracy.I assure you I didn't add any components, and did not simply average the values. With this code in place, your modified Pro would achieve an absurd level of precision.
-Trevor
