Is DSP useful?

[JC1]

Easy Way

This way that you produce is really very hard....

ADSP-21369 from Analog Device is wrong for IB/PI MD - you need ADC > 1 Mega samples
(ADSP-21369, 24-bit codec is good but sampling is very very slow only 96kS, very big price)

QAM and digital TV technology is for Analog IB but for Digital, the error is too big...
FFT(like Garret) is too slow and need very fast DSP for small errors...
"delays and the phase errors" is constants and you can correct digitally...

Chirped Waveform, Compression Radar, SARs... is for radars and not for MD, but you can get some stuff from Radars

PC softwares for Lock in Amp, good, but methods is wrong and slow and is for analog metal detectors....

"The SR850 is a digital lock-in amplifier based on an innovative DSP" - is precise device for accurate measuring,
for MD you need DETECTOR not meter....

PS: I say NOT DSP is best of the best - but is good way....
and sorry for my English......

Hi Wizard,

This is the easy way, board is already laid out

and built and tested. 16 bit ADCs 100 kHz is fast

enough for VLF detector and that is all I am talk

about with this. It also has 16 D/A converters which

can make the transmit sign wave at say 7 kHz.

The demodulation, filtering, lock in amp, everything

is done in the DSP board.

So just need software mainly and could make it go.

though I haven't examined all this to be sure..

And there may be better chips or boards out there.

[TheWizard]

Hi JC1

ok but your ADC is too slow 100kSamples / 7kHz = 14,2857 degrees range/error and if you make DSP demodulation the sense will bad...
I test this before 4 years in PC - hardware and software simulations...

[Qiaozhi]

DSP Development Kit

Has anyone got one of these?
http://www.mikroe.com/en/tools/dspicpro3/
If so, what do you think of it?

[JC1]

You are right

Hi JC1

ok but your ADC is too slow 100kSamples / 7kHz = 14,2857 degrees range/error and if you make DSP demodulation the sense will bad...
I test this before 4 years in PC - hardware and software simulations...

Hi Wizard,

You are right, I post the wrong evaluation board for

the ADSP-21061 processor which the paper says

16 bit A/D at 48 kHz. I can't find a eval board for

this exact part. It may be too old. At any rate I would

also prefer faster conversion, but not for the same reason.

In DSP alot of interpolation of data goes on and I don't

need a sample/degree to get accurate phase information.

With DSP methods.

Now as far as directly comparing phase in microprocessor

you would need what you said or 360 degrees times 7 kHz

and that is getting a little too fast to easy realize.

But even with DSP math still would be nice to have

a min sample rate of 10 times that and that would still

be 70 kHz.

Guess I have more looking to do.

But you get the general idea.

[JC1]

Nyquist

The SR850 is a digital lock-in amplifier based on an innovative DSP (Digital Signal Processing) architecture. The SR850 boasts a number of significant performance advantages over traditional lock-in amplifiers—higher dynamic reserve, lower drift, lower distortion, and dramatically higher phase resolution. In addition, the CRT display and 65,536 point memory make it possible to display and process data in a variety of formats unavailable with conventional lock-ins.

Digital Precision

At the input of the SR850 is a precision 18-bit A/D converter which digitizes the input signal at 256 kHz.

Hi Wizard,

Note that the instrument I posted is a 100 kHz

lock in amp, with sampling to 256 kHz.

this 2.56 times the input freq which barely meets

the Nyquist Sampling Criterion, and like I said I

would prefer 10 times but this will make it.

http://www.thinksrs.com/products/SR850.htm

[JC1]

Don't have one

Has anyone got one of these?
http://www.mikroe.com/en/tools/dspicpro3/
If so, what do you think of it?

Hi Qiaozhi,

Hmmm,,,,, don't know certainly cheaper.

Like Wizard is saying how fast and accurate

is the A/D converter?

[TheWizard]

I see but I dont know how this device work really, We read only SR850 advertisement...
Too many methods for phase measurement...

[JC1]

...

I see but I dont know how this device work really, We read only SR850 advertisement...
Too many methods for phase measurement...

Hi Wizard,

If you read the lock in PDF you will

see that they say it will work to 20 kHz.

This is with 48 kHz sampling which is

2.4 times the 20 kHz.



The theorem states that
“ Exact reconstruction of a continuous-time baseband signal from its samples is possible if the signal is bandlimited and the sampling frequency is greater than twice the signal bandwidth.

In practice, neither of the two statements of the sampling theorem described above can be completely satisfied, and neither can the reconstruction formula be precisely implemented. The reconstruction process that involves scaled and delayed sinc functions can be described as ideal. It cannot be realized in practice since it implies that each sample contributes to the reconstructed signal at almost all time points, requiring summing an infinite number of terms. Instead, some type of approximation of the sinc functions, finite in length, has to be used. The error that corresponds to the sinc-function approximation is referred to as interpolation error.

http://en.wikipedia.org/wiki/Nyquist...mpling_theorem

http://zone.ni.com/devzone/cda/tut/p/id/3016

[JC1]

DSP Class

The conventional phase measurement techniques introduce error in the phase when the input signals are distorted by harmonics. A novel technique, known as adaptive sampling, for high-precision phase measurement is introduced. A digital signal-processing approach is used in this technique. The maximum sampling rate required for this technique is h+2 samples/cycle of the input signals, i.e. (h+2) f sampless, where h, is the highest harmonic present in the signals and f is the fundamental frequency of the signals. This sampling rate is way below the Nyquist sampling rate (more than 2hf samples/s) when h is a large number. In the adaptive sampling technique the sampling rate is started from three samples/cycle and then is gradually increased until the phase is correctly measured. This phase measurement technique has been verified using synthesized signals.

http://ieeexplore.ieee.org/xpl/freea...arnumber=39036

[JC1]

Phase

Many methods some better than others.

FIELD OF THE INVENTION

The present invention relates generally to measuring techniques of the magnitude and phase of a sinusoidal signal. In particular the present invention is directed, in one aspect, to a digital measurement technique of the magnitude and phase using a modified least square error method, and in a further aspect the technique is applied to the detection of load coils in a telephone subscriber's loop circuit.

BACKGROUND OF THE INVENTION

The amplitude (magnitude) A (A>0) and phase φ of the sinusoidal signal X(t) of the form

X(t)=A sin(ωt+φ)

where ω is a known angular frequency in radian per second, are important parameters to measure in many practical measurement and control applications in the power system or the like. Many methods have been proposed for the purpose and include the zero-crossing detection (ZCD) method, the discrete Fourier transform (DFT) method, and the least square error method (LSM). The zero-crossing detection method transforms the signals into two square waves, and then measures the time difference between the zero-crossings of the two square waves which is proportional to the phase. This method suffers from phase errors clue to the hysteresis and different signal amplitudes. An article entitled "A Novel Digital Phase Meter" by Ibrahim et al in IEEE Transactions on Instrumentation and Measurement, Vol. IM-36, pp 711-716, September 1987, describes such a measurement.

The discrete Fourier transform method performs the DFT on N digitized samples of the signal X(t) which are sampled uniformly at a higher rate than the Nyquist sampling rate. The signal X(t) is converted into a causal discrete-time sinusoidal signal x(n)

x(n)=A sin(ωnT+φ) for n=0, 1, 2, . . . (1)

where T is a sampling period in seconds. Subsequently the DFT coefficient is obtained and using this coefficient the amplitude and phase of the signal are calculated. The method however suffers the effect of spectral leakage from the edge discontinuities arising from the beginning and ending of the samples.

In an article entitled "Phase Angle Measurement Between Two Sinusoidal Signals" by R. Micheletti in IEEE Transactions on Instrumentation and Measurement, Vol. 40, No. 1, pp 40-42, February 1991, a new algorithm for measurement of phase angle is described to be based on the least square error method (LSM). The algorithm processes digitized samples of two input signals to calculate the phase angle between them. The algorithm is a general form of measuring technique and uses any arbitrary discrete sampling frequencies and sample sizes. Computation is quite complicated.

http://www.patentstorm.us/patents/54...scription.html

http://ieeexplore.ieee.org/iel5/19/2...rnumber=948289

[Qiaozhi]

Hi Qiaozhi,

Hmmm,,,,, don't know certainly cheaper.

Like Wizard is saying how fast and accurate

is the A/D converter?

If no-one here has used the MikroElektronica stuff, then what are they using?
How about you Wizard? I see you're using UEStudio http://www.ultraedit.com/index.php?n...owpage&pid=150
but what about the programmer/debugger?

I'm thinking about buying the MikroElektronica EasyStart PIC kit, and would like to get some feedback from any users. Good or bad?

[TheWizard]

I dont know MikroElektronica stuffs...
I work with Microchip MPLAB-IDE, ASM, C... ICD2..
UEstudio like text/hex editor...
About Demo boards - I can not give you right answer
If I need - I make demo board....

[Delbert grady]

DickySPy

So is this correct

You have a receive wave that is say 5 khz and you want to do an AD sample as many times as possible each cycle.

You end up with a lot of digital numbers that you want to process some way.

If this is correct. How many times each cycle do you want to sample.

[JC1]

Correct

So is this correct

You have a receive wave that is say 5 khz and you want to do an AD sample as many times as possible each cycle.

You end up with a lot of digital numbers that you want to process some way.

If this is correct. How many times each cycle do you want to sample.

Hi Delbert,

You want to sample as much as is possible.

You have to sample greater than twice the

highest incoming frequency. Nyquist.

Otherwise aliasing will occur such as you see

on a digital scope when the time base is set too low.

Now the theory is correct but assumes you have

digitized the data with infinite resolution.

Which of course you cannot do in the real world.

Audio CDs were first digitized at 44 kHz to allow for

20 kHz bandwidth. Now I believe the higher fidelity

ones are digitized at higher rates and more A/D bits.

Now this is all fine but as you stated there is more

processing that we would wish to do for a metal detector.

Or whatever. In which case all this processing, in fact

just about every computation on a computer will

generate errors. So the more computing you do the

worst things can get. So now how fast do we need to

sample? well now that question just became more complex.

My rule of thumb is to sample at least 10 times the

highest frequency coming in. 50 kHz in your case.

20 times probably even better, but at some point

higher sample rates may not yield any more accuracy.

The Nyquist sampling theory, or so it is called by many,

is not very intuitive to understand.

But then this is true of alot of DSP.

So I wonder who on the board is qualified for this type

of work. And alot of work it will be, for sure, before

all the little details are worked out.

[JC1]

...

If no-one here has used the MikroElektronica stuff, then what are they using?
How about you Wizard? I see you're using UEStudio http://www.ultraedit.com/index.php?n...owpage&pid=150
but what about the programmer/debugger?

I'm thinking about buying the MikroElektronica EasyStart PIC kit, and would like to get some feedback from any users. Good or bad?

Hi Qiaozhi,

That is probably fine for a nice microprocessor

project.

Not so good for DSP.

In my opinion.

Now for DSP there are lots of high speed processors

with higher speed high precision A/D converters etc.

But then guess what?


FPGAs can outperform DSPs, says study

http://www.dspdesignline.com/news/19...RSKHSCJUNN2JVN

So the story continues.


Now in a microprosser you can still do some

maybe even alot of processing of the signal.

Like near the end of the signal path you can

add up the samples and divide by the number

of samples (averaging) which is technically

signal processing. Just not very much or very

exciting.