Universal PI Micro

[BJBJ]

BW!
If i may ask, was the performance in your nugget test not satisfying
with the Atmel device in terms of the resolution of the 10bit AD?

Btw, hillarious nuggets!

regards
BJBJ

[bugwhiskers]

Hi BjBj,

The biggest problem with the bus prototype was noise, a whopping 80mV of it in the decay curve.
The earlier you can sample with a PI the better, especially for small nuggets. The more amplification you put on the front end the later the earliest sample time. If a low amplification is used and the overall "gain" is realised through improved resolution of the available signal voltage then it would seem to be the best all round compromise. With 24 bits to work with it is very easy to shove them right a few times (halving the value)to get rid of noise and still have a fairly chubby number to work with.

regards
bugwhiskers

[BJBJ]

Do you have a scope picture (JPG) of a typical decay curve with this 80mV noise?

What about this idea then, why not using the GPR principle?
Stroboscope the samples period then averaging out the noise?
It give far more detail to the sampled signal but it will require
fast sampling but then you could lower the number of sample
cells, i.e run it in sample frames , one frame of e.g 8 samples
then a storage period to RAM then run another frame.

[bugwhiskers]

Hi BjBj,

There is some averaging being done in software and it may well prove to be enough. It is possible to have a switchable non motion mode where the user would hover over a suspected target then some serious averaging to reduce noise could be done.

You really want to see the noise BjBj ? it isn't a pretty sight !


regards
bugwhiskers

PS What did you find so hilarious about the nuggets ?

[BJBJ]

>You really want to see the noise BjBj ? it isn't a pretty sight !

Yes please, i do!

What do you think of my frame sample idea? Bad?

>PS What did you find so hilarious about the nuggets ?

Hilariously large!


We simply dont have these here in mid Scandinavia.
The ice age have grind'ed them down to powder.

Regards
BJBJ

[BJBJ]

BW!

I don't want to discourage you but!
Studying the datasheet I see that they're quoting +/-15ppm INL which
is closer to 16-bit resolution than 24. A bit more digging suggests
that the part is really only accurate to 18 bits at very best under
premium circumstances which a detector is not!

So some of those 24 bits are quite worthless. For instance Audio
rate codecs aren't much better 24-bit ADC/DAC combinations are
pretty cheap these days, but unless you're picking the parts
carefully and surrounding them with high-quality drivers & supplies
you'll never get the kind of accuracy that simply counting bits
would suggest.

Assuming a +/- 1V signal swing, a single bit is around 0.1uV
at 24 bits. This is pretty much impossible to engineer,
especially considering the equivalent circuit SNR has to
be better than 144dB. Even if voltage swing are increased
dramatically this PI enviroment are about to trash AD bits
around.

When recording audio for instance, it's impossible to find a
microphone+preamp combination that comes anywhere close to
that. And that are in studio environment here we are out
in the field!!

So most so-called 24-bit converters are just specmanship.

Even if your signal are heavy filtered/sampled into DC i would be
supriced if you even could squese out 16bit of resolution of the Aduc.
You still have charge injection to deal with who will alter the
sampled value.

Best regards
BJBJ

[bugwhiskers]

Hi BjBj,

All R&D work is try it and see. A few things are certain, the AtoD must be inside the micro so that external transfer overhead doesn't slow things up. I am going for the maximum number of bits and then see what is left that is usable. Another certainty is the low cost and ready availability of these evaluation boards. The PGA and digital filtering is also a very big plus for the ADUC845, it will allow software programmable gain and possibly elimination of certain noise bands. Don't forget, the AtoD will be reading the voltage on a cap which is static and I can interlace the readings to allow for the best possible result.

I do appreciate all input BjBj, afterall, this is not a commercial project and I expect to make nothing out of it except a detector that will do what I want.

Knowing how the sampling works, perhaps you could provide an alternative method to achieve the same objective, I assure you my ego will be totally unaffected as I am under no illusion that I know everything.

The attached pic is of the Evaluation board running a 40X2 LCD.
Now I can start doing some sampling and see just how many of those 24 bits are realisable.

regards
bugwhiskers

[BJBJ]

[quote=bugwhiskers;54267]Hi BjBj,

>All R&D work is try it and see. A few things are certain, the AtoD must be
>inside the micro so that external transfer overhead doesn't slow things up.
>I am going for the maximum number of bits and then see what is left that
>is usable.

Well, as mentioned previously INL has already eaten up some bits
even before you actually have powered the Aduc device up!
So regarding to that specific value there is no need for test and see really.
But i do understand your approach, i would have done the same.

>Another certainty is the low cost and ready availability of these evaluation
>boards. The PGA and digital filtering is also a very big plus for the
>ADUC845, it will allow software programmable gain and possibly elimination
>of certain noise bands. Don't forget, the AtoD will be reading the voltage
>on a cap which is static and I can interlace the readings to allow for the >best possible result.

>I do appreciate all input BjBj, afterall, this is not a commercial project and >I expect to make nothing out of it except a detector that will do what I >want.

I really hope you succeed, because if you do you have a buyer
of the kit over here!

>Knowing how the sampling works, perhaps you could provide an alternative
>method to achieve the same objective, I assure you my ego will be totally
>unaffected as I am under no illusion that I know everything.

Well, i hardly know anything i just try to help out!

At the moment your multiple sample cell idea is a very good solution i suppose.
There is always the direct sampling with video ADC plus a FIFO or fast ram but
that will eat battery power and you need a micro with a fast access bus to external
ram, all a bad idea.

>The attached pic is of the Evaluation board running a 40X2 LCD.
>Now I can start doing some sampling and see just how many of those 24
>bits are realisable.

Splendid!

I think your doing great BW, i just try to see all sides of the subject in
a way to help out even if it might sound harsh sometimes but reading
data sheets that promises gold and glory raises my suspiciousness a bit.

There are plenty of bonuses with the Aduc to test it to see where
it ends and i don't think you need the 24 bits even 16 bits would do
more then fine if its achievable.

How much did you pay for the starter kit, and what tools do you use?

Regards
BJBJ

[bugwhiskers]

Hi BjBj,

The Evaluation board is available from any worldwide Analog Devices distributor for around $75US. It comes with a full IDE for assembler and also a Simulator that works. There is a "C" kit available but it is a lot more expensive.
There are code examples for the 8051/2 everywhere on the web and the instruction set is easy to learn and very efficient.

Today I hope to finish the routine that takes the 3 byte output from the AtoD, converts it to 8 digit ASCII (with leading zero blanking) and displays it on the LCD screen.

regards
bugwhiskers

[bugwhiskers]

The routine to get an AtoD reading onto the screen is working as can be seen in the photo.
Leading zero blanking is working and if the input number is zero the last zero is left intact.
Notice some jitter on the second last digit. The input is being fed from a 10 turn pot connected between AVDD and GND. After watching it for a while I concluded that there is a jitter of ~ 1000 without any digital filtering or S/W averaging.
The next step is to hook up the multiplexers, pre-amp and coil drive circuitry.

regards
bugwhiskers

[bugwhiskers]

Changing the input to the pot from AVDD to REF+ has reduced the noise to ~200. The test was done with the fastest conversion speed which also has the worst noise.

regards
bugwhiskers

[bugwhiskers]

For the code hungry.

Attached is the source file containing the LCD screen driver with leading zero blanking. The extension has been changed from .asm to .txt to allow me to post it here.

The program takes 4*24 bit samples from AIN1 to AIN4.

At the moment it is set up for the fastest possible sample time and I am staggered at just how stable it is. Mind you, it is being fed from a string of resistors hooked up between +REF and AGND.

regards
bugwhiskers

[Ungoplopp]

Attached is the source file containing the LCD screen driver with leading zero blanking. The extension has been changed from .asm to .txt to allow me to post it here.

The program takes 4*24 bit samples from AIN1 to AIN4.

At the moment it is set up for the fastest possible sample time and I am staggered at just how stable it is. Mind you, it is being fed from a string of resistors hooked up between +REF and AGND.

regards
bugwhiskers

Breath on the string and watch the numbers!
BJBJ

[bugwhiskers]

Hi BjBj,

I did even worse than that, I put a wet finger on the adjustment pot leads and the 4 digit noise was evident. Remove the finger and the numbers return to exactly what they were, about 5 times in a row this was performed.

I can email you the video if you want.

I must admit I had trouble believing it also but a picture is worth a thousand words and a video even more.

regards
bugwhiskers

[BJBJ]

In what way should i intrepetet that?

Good or bad?

Bad if its 24 bits, good if its 16bits?

What pot are you using a +100ppm ,50ppm, 1ppm?

Your test are at home in somwhat controlled enviroment step
out into the your Aussie desert at 50deg c for aminute os so
and expose the ADC/Mux/Caps/resistors ,etc to the heat and
it will change.

Better yet put in a LM 35 or similar and continuely read it and calibrate
away the temp change to the sampled coil value!

Btw if you have still pictures from scope of the sample noise it would be nice!