field test unit no 001 "model T"

[UrbanFox]

Back again.

While I ignore Ufox (he can't even write my name)
Aziz

OK Aziz, a legitimate question which I am sure forum members would be interested in hearing an answer to.

If we are talking PI, where does the X component come from for Ground Balance???

[Aziz]

Hey, using the semi analog-digital demodulator/modulator concept, we can modulate four samples into a stereo input channel of the sound card. That should be enough.. It will even work with two samples only (high frequency/low frequency components).

One of the highlights of the Tablet PC processing is, that you can make everything in the software!!!
Cheers,
Aziz

[Aziz]

OK Aziz, a legitimate question which I am sure forum members would be interested in hearing an answer to.

If we are talking PI, where does the X component come from for Ground Balance???

Why should I talk about GB with you? You won't get my WBGB.... but the Chinese.... or the QED2 dev team...
MUHAHAHAHAHAHAHA

[UrbanFox]

Why should I talk about GB with you? You won't get my WBGB.... but the Chinese.... or the QED2 dev team...
MUHAHAHAHAHAHAHA

Don't talk about GB with me then, Aziz. Talk to the members of this forum and explain where, with a PI, the X component is coming from for Ground Balance.

[chudster]

Back again.

... Where is the significant upper frequency limit?

Consider the following practical example:
We have a damping period of 6µs - 10 µs until we can get some samples. While looking at the coil current slope (dI(t)/dt), we can expect of having not much significant high frequency components after the flyback period. The later sampling times eating a lot of high frequency response and we effectively end up of having maybe a bandwidth of 200-400 kHz.

So 1 Mega samples/s is just an overkill and should be enough for experiments.

Aziz

I think Carl explained it well in his post. It depends on the approach you take and also whether you integrate or demod the signal in some way and then sample that.

A low sample rate with a few samples at strategic times is possible and approximates sample and hold techniques of some analog circuits. Another way is to ask what is the target signal, determine the highest freq components in that signal (including "features"), sample it to the digital domain and with an accurate representation of the signal "oscilloscope style" it can be processed as wanted digitally. This seems to be the approach Moodz is taking and is the approach I would have expected for the UMD (that's the Ultimate Metal Detector). A direct sampling approach seems to indicate 1MSPS to 2MSPS.

Chudster

[Davor]

We have a damping period of 6µs - 10 µs until we can get some samples.

You can observe this as a successive process. Let me get there... There is a pulse, a dead period, and a sample. For every subsequent sample you can also observe it as a very same process: pulse, somewhat longer period, sample. For each such sample you have successively lower high frequency content information.
Now, considering first sample as a combination of sample time (~1us) + waiting time, and frequency content as ~1/T, there is a significantly lower frequency content in each of such samples than the sample time (~1us) may suggest.
It also appears that fourier may not be the best transform for such samples.

[Davor]

Don't talk about GB with me then, Aziz. Talk to the members of this forum and explain where, with a PI, the X component is coming from for Ground Balance.

as if you have any idea
Why does every troll have to pose as a stuffy professor, or as a nagging woman :duh:

[Carl-NC]

Bandwidth limitting makes sense.

Consider, you have a TX transmitting a wide band pulse, let's say 10 Mhz bandwidth (ultra fast coil).
And you can detect only a bandwidth let's say 1 Mhz bandwidth.
Does it makes sense to transmit the other 9 Mhz bandwidth too?

No, not at all. You could move the "wasted" transmitted energy into the frequency region you take the measurements. And you have more bang to the targets.

Fortunately, the coil with it's parasitic capacitances and damping resistor forms a low-pass filter, which is limitting the reasonable frequency response into the 500kHz to approx. 1 MHz region. If you are not using (sampling) this frequency region, you are just wasting TX energy.

I'll have to disagree with this. On most PI detectors the "detection bandwidth" through the sampling integrator(s) is only 100Hz or so. The same can easily be true of direct sampling with an ADC. That is, if you synchronous sample a single point 10us after the flyback then the required ADC BW is very low. So do you still need more than 100Hz BW on the transmit?

[Aziz]

I'll have to disagree with this. On most PI detectors the "detection bandwidth" through the sampling integrator(s) is only 100Hz or so. The same can easily be true of direct sampling with an ADC. That is, if you synchronous sample a single point 10us after the flyback then the required ADC BW is very low. So do you still need more than 100Hz BW on the transmit?

No, I meant the bandwidth before the synchronous demodulation occurs.
Does it make sense to transmit at frequencies, you can't detect at all?
Clearly not. Just save the TX power energy or use it additionally in the frequency range, you can detect.

Aziz

[Aziz]

as if you have any idea
Why does every troll have to pose as a stuffy professor, or as a nagging woman :duh:

This guy/girl is looking for the "Made in Germany" WBGB technology.
*LOL*

PS: Due to bad behaviour of patent trolls, the technology isn't published to date.
If I would do this, someone would steal and patent it with no shame.

[Aziz]

BTW,

the integrator in the front-end is also limitting the detectable frequency response. Most aren't even well designed to get the high frequency components. I dare to say, that most PI detectors can't detect bandwidth of greater than 100 kHz.
*LOL*
Aziz

[Tinkerer]

Hey Tinkerer,

Thanks for your response. Its helpful for the advancement of stuff with some real data to chew on.

In summary with a 2uS signal we need 1MSPS or higher to have a decent chance of detection or smaller targets. Moodz working at 2MSPS is 'overkill' but not unreasonable as its good to oversample (where is that antialiasing filter though Moodz). At 48KSPS, Aziz's soundcard seems OK for cannonballs with very long time constants (160uS) but not a tiny nugget.

Note we are talking about direct sampled solutions here. We can do all sorts of analog processing to have a half analog half digital solution (including integration) to mimic an analog detector with lower sample rates.

Chudster

Direct sampling:

The way I see it, direct sampling would be directly off the RX coil.
If we add a preamp, we might just as well add an integrator/anti alias filter.

However, the signal response of a 1us TC target is of really short duration, so to get the peak response, the sample should also be short, otherwise we dilute the peak amplitude with a long sample time.

One of the secrets of FE discrimination, with PI, is that the X response may have a TC of 2us, while the R response of the same target, has a TC several times as much.

Unless we are capable of sampling the X response separately from the R response, we will not be able to differentiate this FE target, as the sum of the X+R responses will mask the weaker response.

In other words, we need not worry about 2us TC targets with a traditional PI, since we do not have the discrimination capability.

With the TINKERERS TEM PI, however, that is capable of sampling a 1usTC target at it's peak, the capability of taking several samples of 1us in quick succession, can make a difference.

So, again, we need to specify which technology platform we are talking about.

Moodz has his own technology platform. Only he knows what the capabilities are and how to make best use of these capabilities.

We must not assume that all PI platforms are equal.

Tinkerer

[Carl-NC]

Does it make sense to transmit at frequencies, you can't detect at all?
Clearly not. Just save the TX power energy or use it additionally in the frequency range, you can detect.
Aziz

In PI, it does make sense. The BW of the TX pulse determines how hard the target gets kicked. The harder you kick it, the stronger the response. Once the target has been kicked, it is true that you don't need to maintain all that BW through the receiver. In fact, you can direct sample with only 100Hz of BW.

[Tinkerer]

yep I have seen the kickstarter.com site ...very interesting concept ...who is up for it ?

I understand you need to be a legal US resident to participate in Kickstarter.

Tinkerer

[Carl-NC]

Direct sampling:

The way I see it, direct sampling would be directly off the RX coil.
If we add a preamp, we might just as well add an integrator/anti alias filter.

Not true. The preamp can provide a better SNR than you can get with an ADC alone, unless the ADC has a built-in LNA. In direct sampled radios there is always a LNA in front of the ADC. Also, a lot of these high-res ADCs are rather picky about how they are driven, and often need a preamp with a particular matching network to minimize sampling kickback. You sure don't want the sampling kickback to see a coil.

Also, an integrator establishes a "channel" that limits your processing, i.e., you need a separate integrator for every unique portion of the response you want to process, but a preamp alone does not limit processing.