Tweet
Originally posted by Sean_Goddard
-
Sean is going too far into this issue of design,why asic and advanced processing, 555 timers will do the thing and they are cheaper, use dft only if u want an lcd model and use a micro and use dsp in that...dont complicate the things, u need to concentrate on the front end not on the signal fft etc and some even speak of shifting the frequency so that they will not interfere with the sampling or match with sampling or voltage doubler freq which is not required !!!! -
DSP
Originally posted by Sean_Goddard
Yea and someone who can define
just what is meant when someone
starting screaming RA RA RE DSP !!!
I said DSP first, I own it, I own it.
Someone else do it.
DSP can solve everything I know it.
You do it or you are the failure.
I said DSP first.
Companies spent millions doing DSP and
ended up with something a good analog
detector would beat on depth everytime.
What is so special about your approach.
If you guys don't KNOW what you are talking
about you have lost from the beginning.
Maybe your first priority after you figure
out your market should be to investigate
the DSP that has been done by other
companies, their cost, and their failures,
or one success barely with the GTI. Kinda.
Has not dominated the market, and yes
I KNOW you can buy fast big DSP processors
and you still will fail.
Who has the algorithms? Who has the real math?
What is special about this? Oh you are going to do
a fft. On what? a single freq on vlf or multi on PI.
So what does that tell you? probably nothing.
Coding is simple once definition has been made.
This is a big big mountain to climb.
Yea, get mad at me for poping your bubble. -
If all you want to design is a basic PI or VLF that is just like everyone else's PI & VLF, than I agree, a 555 is the way to go.Originally posted by sharky View Post
If you want to move detector design to a new level, then DSP is the next step.
- CarlComment
-
I have some ideas that are only realistic in a DSP-based architecture.Originally posted by JC1 View Post
- CarlComment
-
@ JC1
Where is the problem with DSP, math, algorithms.....
If Bruce Candy and companys is "barefoot" about DSP - the word is not....Comment
-
JC1 and Sparky you misunderstand..
I said DSP, not because i think it's clever and "hip" but becasue it IS the answer. Analogue machines as good as DSP, no sorry that's rubbish!! The reason the Garret DSP machine are rubbish is, and I'll REPEAT MYSELF FOR THE TWENTIETH TIME, THEY DO NOT have the fundamental understandings of what is possible. the GTi they made Is and as you will see if you are in the group and I am in the group, was my idea, NO I DO NOT own it, you will see the document I sent to Garett suggesting a machine using DSP and called the Garrett GTi DIGIDEC.
Right, let me say this AGAIN, I work on military communications systems, the BEST analogue Rx will give you around 10dB SINAD for a -114dBm input.
NOW consider that Planet earth noise floor is around -123dBm. THAT is the level at which the natural background noise of the planet we are on occurs, ANYTHING below that is swamped, BUT if you use correlation and many other DSP based tricks, the best Rx the company I work for makes has a SINAD of 12dB for an input of -127dBm, YES that's a signal picked up BELOW the noise floor. NO Analogue system could do that, AND be as portable as we need a detector to be, but with DSP is a breeze.
I appreciate that what you have seen on DSP based machines to date is not good, but as I have said MANY times they are TOYS....FACT Even the ML Explorer processes it's signal mainly in the analogue domain.
ML boasted that the Explorer processor executed 20Million IPS (instructions per second), the machine I have in mind uses two Blackfin DSP chips and has around 1.5 BILLION IPS, more than enough to do ANYTHING ans you will see when I make my suggestion open to the group.
I have researched my design for the last 20 years. I have approached many manufacturers, but none could understand how ot make it, or afford to (A good DSP programmer can charge around $100+ per hour).
I appreciate the comments about my doing only reaserch, a valid point.
I was once told that you can't overcome the laws of physics just because you are using a DSP, BUT you can characterise and compensate for them. THAT'S where Garrett went wrong, all they did was implement a direct "hard coded" version of a not very good analogue system, if they had approached it as an entirely different beast, then it would have worked...I'm SURE of it!Comment
-
SAME here Carl. What I have in mind is impossible using analogue (well portable analog anyhow)Originally posted by Carl-NC View PostComment
-
CORRECT, the initial signal prep can be done using analogue, but then DC level remove and decimate so that a PROPERLY IMPLEMENTED DSP based detector can do it's magic, it's NOT about whether of not DSP can do better than analog, it's about IMPLEMENTATION.Originally posted by sharky View Post
If th implementation is rubbish thne so will be the result. remember, Garbage in...Garbage out (an old programmers idiom). BTW I spoke to a guy at Qualcomm I used ot work with who has over 20 patents to his name, and he suggested EXACTLY what to use , even drew me flowcharts and block diagrams, and an FFT AIN'T IT!!
Of course the above, AND my suggestion to use DSP is ONLY my suggestion, I am NOT trying to force anyone to say that my deas are best, far from it, Carl is a LOT cleverer than me as is Qaiozhi, and a good deal many others on here. I ws merely pointing out that it is an OPTION at this stage. Maybe there will be two or MORE design groups each working on different ideas.
We need a project manager
.
Comment
-
@Sean
You not need Blackfin DSP for metal detector - The MD math is very slow for Blackfin.... you want to kill Ant with Elephant
but the idea is good
Comment
-
No, for what I have in mind you WILL need one. A MD/GPR hybrid?? I'd love to see someone try that in analogue, BUT hey I'll eat my laptop if it can be done on a reasonable platform
.
Comment
-
I won't hold my breath
Yea I have heard the DSP spouted by
technicians and junior engineers that
have absolutely no idea what they are
talking about.
As the solution to any and everything.
It is not true.
I am NOT saying IT cannot
be done believe me I KNOW there are
missile systems which can do amazing
things. I also KNOW the level of effort
these things require to work properly.
And the cost of these man hours.
Now what I am saying is that a good
analog beeper WILL beat any digitized
version period, no matter what math
or whatever you do to the digitized
signal.
As far as raw sensitivity is concerned.
When you digitize it you
quantize it and limit it right there.
And then the math errors just grow
from that, on any further processing.
Now I am not saying you can't do some
interesting things, and taking something
like the ArcTangent is practical in analog,
but you WILL NOT
improve the basic sensitivity of the detector
by digitizing the signal and doing whatever.
In digital.
Image what you will.
Keeping it real.Comment
-
not easy
Taking the ArcTangent in analog
is NOT easy. Mistype.
Anti-alias filters do a number
on Group Delay.
Oh and WHITEs spent millions
chasing the DSP monster and never
came up with anything to market
that couldn't be bested by any of the
analogs.
Whoops.
Big DSP sales job.
But hey there are hundreds of these
stories.Comment
-
I want my money back
Poor Garrett,
They wanted to get their money back.
And now they sale something nobody
is bragging much about.
Real DSP processor and all.
But hey they want to get their money
back. Going to have to sell a bunch of
these.Comment
-
DSP can easilly pick out signals below the noise floor so there is merit in that approach. Maybe the best option is to make use of the best traits of both analogue and digital techniques. The thing that lets most detectors down is in the way the ground balance circuitry works, subtraction is subtraction as like throwing the baby out with the bath water. I have carried out many experiments with GB on GB off and the detection range and sensitivity increase in GB off are substantial. To come up with a better detector one has to think outside the square. Carl should select a core group of engineers to look at various submissions from the site members, if any of these look prospective they could be passed on to members of a sub group to investigate the concept. What you will need to do Carl is set up a data base of members expertise and get the ball rolling.Comment
-
JC,
The point of opening these discussions is to explore the possibility of forming a collaborative design group, not to debate detector architectures. Let's please stay focused, and first see if this is worth doing.
- CarlComment
Comment