Using tablet PCs for processing data and displaying results

[Tec]

to integrate in a linear way

HI Aziz,

I'm using the mobile phone, but only to display conductivity and signal.
Using it as a convenient LCD replacement basically.

I have found the processing can all be done at the lowest level and the real important stuff is the front end..eg the way you are measuring the pulse widths.

My system uses only 2 pulse widths, called V1 and V2. (your proposal uses 4).
They are being added up (integrated) through a constant current (see attached).
I happen to be using 35 pulses, and then taking a measurement on the capacitor.
After the measurement, the capacitor is discharged and the pulses start again.

Note, the results (attached).. you will find the ratio of the pulse width changes are constant for a particular conductivity..i.e, When v1 (pulse width1) and v2(pulse width2) are plotted on a graph, they make a straight line..and the slope indicates conductivity. This tells me that the constant current method is good, and the system works as expected.

Anyway, the signal is very small at distances greater than 20cm, so all the processing power in the world can not help determine the target in this case...it is just a small beep, which could be almost anything.

I have really been concentrating on getting greater front-end performance..deeper, and less noise.
One way is to increase the pulse width, or take more samples..


Regarding the Bluetooth I'm using RN-42 from sparkfun...

[Aziz]

Disclosure of real patent troll and ideas thieves.

If you don't have anything valuable to this idea or discussion, then please shut up!

Aziz

[Aziz]

Hi Tec,

two samples wouldn't be enough. More than four samples would make the processing algorithm and the PI controller more complex.

We need the simplest PI controller hardware with the jitter free timings (a small challenge).

And we don't need the "linearised" ADC conversion using the constant current method (integrating ADC). All the non-linearity handling can be done in the processing element via software. We just have to optimize the dynamic range of the four samples.

Once we have the four samples in a PC environment, we can find a good algorithm for their processing.

Aziz

[WM6]

If you don't have anything valuable to this idea or discussion, then please shut up!

Aziz, I am glad to follow your extremly valuable contribution, keep going in peace.

[mario]

Any news guys?

I still like the bluetooth interface idea.

I am currently investigating a convenient µC.
PIC vs AVR?

Well, I personally tend to use the AVR platform as I have everything I need at home.
~20 MHz clock frequency should be high enough.

Aziz

Have you considered using the new USB xmegas, they have full speed USB, can be easily overclocked to 48MHZ (fully functional and without problems!), and have advanced features like cascading timers, an overhauled and fixed ADC, DAC (but limited to 65khz pwm), and the event system (which is unlike anything found on other systems).

With a bootloader, this is one serious piece of hardware.

The only drawback I see is the need for a PDI programmer. Once with a bootloader it won't matter, but if no bootloader, you need JTAG ICE MKII, or JTAG ICE 3, AVR Dragon or other PDI programmer. They're not AVRISP MKII compatible.

I saw the errata found on the DSPics and was put off by the bugs. Like the previous generation xmegas, I think those DSPics should have been recalled by the manufacturer.

[Aziz]

Have you considered using the new USB xmegas, they have full speed USB, can be easily overclocked to 48MHZ (fully functional and without problems!), and have advanced features like cascading timers, an overhauled and fixed ADC, DAC (but limited to 65khz pwm), and the event system (which is unlike anything found on other systems).

With a bootloader, this is one serious piece of hardware.

The only drawback I see is the need for a PDI programmer. Once with a bootloader it won't matter, but if no bootloader, you need JTAG ICE MKII, or JTAG ICE 3, AVR Dragon or other PDI programmer. They're not AVRISP MKII compatible.

I saw the errata found on the DSPics and was put off by the bugs. Like the previous generation xmegas, I think those DSPics should have been recalled by the manufacturer.

Hi Mario,

xmega is an ice cream.
But the development platform is expensive and not convenient for amatuer usage.

I tend to use the simplest µC, which can also be soldered & programmed easy.
(K.I.S.S.-principle)

Well, the task list for the µC is merely:
- to control the PI pulse (jitter-free)
- to measure only 4 time periods (jitter-free, 16-bit)
- to multiplex (control) a reference voltage
- to control a gated integrator
- to transmit the 4 samples to a host system via UART (bluetooth)

If this simple job can't be done by one µC, no problem. We could use two of them.

Aziz

[Tec]

2 samples is enough,.. to measure conductivity.

I actually set the trigger voltage using a PWM, so I can scan up and down to any trigger voltage..
(but this is slow)

If you don't have linear current to your capacitor, then pulse 1 is weighted more than the last pulse (when the capacitor is almost full).
The resolution lost as the capacitor charges can not be re-gained in software.


Optimizing the dynamic range:
The pulses can be similar length if each pulse is measured from the end of the last (not from the start)... using an XOR gate on the outputs of the comparators..
eg pulse2 and pulse3 are long, but only need to measure when pulse2 XOR pulse1.
pulse3 XOR pulse2.

In software

pulse2 + pulse1
pulse3 + pulse2 +pulse1

etc

[Aziz]

2 samples is enough,.. to measure conductivity.

I actually set the trigger voltage using a PWM, so I can scan up and down to any trigger voltage..
(but this is slow)

If you don't have linear current to your capacitor, then pulse 1 is weighted more than the last pulse (when the capacitor is almost full).
The resolution lost as the capacitor charges can not be re-gained in software.


Optimizing the dynamic range:
The pulses can be similar length if each pulse is measured from the end of the last (not from the start)... using an XOR gate on the outputs of the comparators..
eg pulse2 and pulse3 are long, but only need to measure when pulse2 XOR pulse1.
pulse3 XOR pulse2.

In software

pulse2 + pulse1
pulse3 + pulse2 +pulse1

etc

Hi Tec and interested members,

two samples aren't enough to do a proper ground balancing (GB). A good GB should be targeted in the first place. But four samples would do it.

I personally would make a switchable (programable) voltage divider to make the reference voltage. This has a fast settling time so all the successive time measurements can be made in the same pulse period frame.

For precision time measurements:
- crystal clock source is a must have condition
- no PLL!!! (causing phase noise)
- input capture unit for the timer (jitter-free time stamp)
- a fast and external comparator (decoupled from the noisy µC)
- a stable (low noise) reference voltage (preferably external)

The optimization of the dynamic range of the samples can be made easily with a gated integrator with programable different time constants. The integrator has the benefit of reducing (high frequency) noise and increasing the signal-to-noise ratio. So we aren't going to sample just voltages on the decay curve. We are going to sample the integrated voltages on the decay curve. We can control the integrator to start from a defined condition (i.e. 0 voltage).

We only have in the PI controller:
- an integrator (programable),
- a comparator (we could use the internal one of the µC but not preferred),
- a reference voltage (programable voltage divider),
- a TX driver,
- a pre-amplifier,
- a µC (a simple one, ADC isn't necessary)
- and an UART->bluetooth module.
That's it.

On the host side (the processing element) we can use any convenient platform and operating system. I personally would use my PC and Netbook to analyse the samples (to find a good algorithm to process the samples). In the finaly version, one can use a smaller and nice processing platform (Android Pads/Tabs/Phones). That's the reason, why I see a big advantage here. The complete platform will be very cheap and very powerful and can compete with the big ones!

Cheers,
Aziz