Replies to Minipulse Plus

[green]

Been getting a brain cramp trying to figure frequency response and gain for the integrator. Tried building it in spice. Changing sample time changed frequency response and gain. Don't know if I did it right. Could someone give the formula for gain and frequency response for the integrator.

[mickstv]

Been getting a brain cramp trying to figure frequency response and gain for the integrator. Tried building it in spice. Changing sample time changed frequency response and gain. Don't know if I did it right. Could someone give the formula for gain and frequency response for the integrator.

Hi Green, check the link below.


http://www.geotech1.com/forums/showt...ntegrator-gain

[josewashere]

If you are still interested in experimenting, try the R11/C11 configuration that results in 15 usec (6.8k, 2n2) and see if the response is a little better for a normal detector swing rate. The combination that give 10 usec may require a little slower swing rate to achieve adequate integration time.

As far as coils... my favorite 2 coils, both of which I made 7 yrs ago and keep using on all my PI's, are a 10.5" DD and a 9" × 4" eliptical shaped figure 8. In both, the tx is ~275 uH and rx is ~380 (I was shooting for 300 and 400 respectively but missed by about 1 turn in each). I love the dd... it gives me the added advantage of ferrous metal discrimination. The figure 8 I use in noisy envirionments and for bench testing as it is extremely immune to environment and ground (salty damp soil) noise. It's only drawback is the small detection profile (the front 4.5" of the figure requireing at least double the scans to cover the same area as the dd.

I think its ok for me were it is at the moment but yes if you swing too fast you don't get a response. Its definitely worth some experimentation to increase the delay if the sensitivity stays the same. I tried to twist various coils I made into a figure eight and the noise went down but the sensitivity went down with it by a lot. A double D design is interesting to me as it should be quieter on bad ground.
Only one of my home made coils is as good as the commercial coiltek coil. Its smaller 9 inch vs 11 inch but when you take the size into account its as good as the coiltek. All others are worse in some way.
I think I am going to buy some proper litz wire one day and give that a good try. I just want to eliminate eddy currents within the coil. Also want to have a go at a basket weave coil one day.

[josewashere]

Today I tried a MAX412 (as suggested by a member) in the preamp in place of the Ne5532 . Its better than the ne5532 and about the same as the LM4562 that works best in my MPP. I think that's it for me trying op amps. LM4562 stays.

[green]

Hi Green, check the link below.


http://www.geotech1.com/forums/showt...ntegrator-gain

Thanks, but I think this integrator is different. It acts different in spice. Maybe I'm missing something. What is the gain and frequency response formula?

[mickstv]

Thanks, but I think this integrator is different. It acts different in spice. Maybe I'm missing something. What is the gain and frequency response formula?

Here's another link. Eric posted the thread, one of the circuits is the same as the MPP, you might find some useful info.


http://www.geotech1.com/forums/showt...ircuits-for-PI

[Tinkerer]

KingJL

Probably the next best thing after the Litz wire, is fine strand tinned insulated wire. It seems that the eddy currents are much less and the insulation reduces the inter-wire capacitance. Twisted, fine stranded, tinned insulated tween leads in parallel give even better results.

[KingJL]

I tried to twist various coils I made into a figure eight and the noise went down but the sensitivity went down with it by a lot.

In the figure 8, only the rx coil is formed into the 8
( or a facimile thereof ), the tx coil circumscribes the entire rx figure 8.

[KingJL]

Thanks, but I think this integrator is different. It acts different in spice. Maybe I'm missing something. What is the gain and frequency response formula?

Gain is R23/R25, LP cutoff is 1/( 2×3.1416×R20×C13 ), band stop is 1/( 2×3.1416×R24×C13 ).

[josewashere]

In the figure 8, only the rx coil is formed into the 8
( or a facimile thereof ), the tx coil circumscribes the entire rx figure 8.

Ok I tried it with a mono coil. It does work and with much reduced noise but very low sensitivity.

[josewashere]

KingJL

Probably the next best thing after the Litz wire, is fine strand tinned insulated wire. It seems that the eddy currents are much less and the insulation reduces the inter-wire capacitance. Twisted, fine stranded, tinned insulated tween leads in parallel give even better results.

I normally use stranded tinned wire with teflon coating to make coils.
So let me see if I understand this right. I take a twisted pair of stranded wires and wind a coil and I then connect the two wires together at both ends so effectively they are in parallel ? I can get some cat 5 cable that is exactly that. 4x twisted pairs of stranded wire and a 30m length will be enough for 4 twisted pairs (4 coils). This is why I want to use litz wire. Its made up of 50 thin enamelled wires that are woven (twisted) an the 50 strands are connected at each end to create a coil with a higher q factor. I can get litz wire as used in high end commercial coils locally for $1 per meter.

[KingJL]

Gain is R23/R25, LP cutoff is 1/( 2×3.1416×R20×C13 ), band stop is 1/( 2×3.1416×R24×C13 ).

This of course is modified by the sample width and the filtering of the input and output circuits. My Minipulse Plus implementation with 15 usec samples and additional filtering of the output (a 10k in series ahead of C15 and 100n in parallel with R29 ) results in an effective gain of ~1.75 and a 6 dB cutoff of 11.5 Hz. I determined the effective bandwidth by using the FFT function of the simulated output.

[KingJL]

Ok I tried it with a mono coil. It does work and with much reduced noise but very low sensitivity.

If carefully made ( carefully balancing the 2 haves of the figure 8 and carefull layout ) the figure 8 can be very sensitive, but only in the center of 1 of the figure 8 halves.

[Tinkerer]

I normally use stranded tinned wire with teflon coating to make coils.
So let me see if I understand this right. I take a twisted pair of stranded wires and wind a coil and I then connect the two wires together at both ends so effectively they are in parallel ? I can get some cat 5 cable that is exactly that. 4x twisted pairs of stranded wire and a 30m length will be enough for 4 twisted pairs (4 coils). This is why I want to use litz wire. Its made up of 50 thin enamelled wires that are woven (twisted) an the 50 strands are connected at each end to create a coil with a higher q factor. I can get litz wire as used in high end commercial coils locally for $1 per meter.

1) It is not just stranded wire, but tinned stranded wire. The Tin helps confining most of the eddy currents within the single strands.
2) the twisting of the 2 wires makes that there are only short pieces of wire lie parallel to each other and therefore generate little inter-wire capacitance.
3) The insulation and the twisting put more distance between the wire turns and therefore reduce the inter-wire capacitance even more.

This combination of effects comes near the effect of basket weaving, but can also be used when basket weaving to enhance the effect.

[Davor]

... I determined the effective bandwidth by using the FFT function of the simulated output.

There is a much more effective way of establishing the bandwidth and its nature in time domain. What you need is a step stimulus that you already have in a form of Ncycles in a sinus voltage source, and you observe the output. It most certainly lags against the stimulus, but the slope and its curvature say a lot. You'll notice that there is a slew rate limited slope, and not a classic LPF slope. It means this integrator will follow input stimulus very accurately, and will not flutter unless screwed up completely.
You can estimate inverse corner frequency (1/f for -3dB point) as 3 x rise time to 85% or 3 x fall time to 15%. You don't have to be extremely accurate on this.
More importantly, you'll notice that this integrator provides more of a constant slope, rather than an exponential one, which indicates a slew rate constraint.
In practice it means that you'll be able to miss some targets if you swing your coil as a lumberjack, but it will still be much faster to respond than a normal LPF response detector, and even more so than the one with overshooting problem.