Advanced PI coils

[Aziz]

Oh but you can, by making significant improvements in other parts of the electronics. Been there, done that. However, that is not for this thread.

Eric.

I'm looking forward to see your improvements Eric.
(Provided that, you're going to share your ideas of course.)

Beating the best is sport & a challenge.
Aziz

[Ferric Toes]

You can't beat the best with an 1k input resistor and diode clipping input path.

Aziz

Oh but you can, by making significant improvements in other parts of the electronics. Been there, done that. However, that is not for this thread.

Eric.

[Aziz]

Hi all,

regarding the low noise & fast amplifier:

Have a look at the data sheet. The gain bandwidth product (GBP) tells you, how much max. gain you can have to the op-amp. A typical PI has a BW of 200-400 kHz.
Lets take a GBP of 5 Mhz and BW of 200 kHz.
Max gain = GBP/BW = 5 Mhz/200 kHz = 25

You see, that we need a more stage amplifier. If we don't take a multi-stage amplifier, we are just reducing our BW of the detector (=inherent low pass filter).

Fortunately, fast op-amps offer more than 5 Mhz GBP. So you can increase the gain a bit if necessary.

Regarding the impedance matching:
It's only important, if you want to transfer POWER (energy) from source to destination! We aren't transferring power here. We are measuring an induced voltage with an high input impedance amplifier (only bias currents flow to the input of the amplifier).
The bipolar input amplifiers need very low input path impedance to reduce the amplifier noise. As we know, op-amps and high flyback voltage in a low impedance path do not blend well. So we need input switching fets to take advantage of the low noise bipolar op-amps.

You can't beat the best with an 1k input resistor and diode clipping input path.

Cheers,
Aziz

[Aziz]

Hi Aziz,

Above where I have written "to cancel the combined emi from tx and tx1...."
Should read "to cancel the combined emi from tx and rx1...."

I have done the measurement of the ML DD coil and I got 316uH and about .4R for the tx and 354uH and 18R for the Rx. This is with the rx damping resistor still in parallel with the rx coil and I think the damping resistor was about 1k, though I cannot confirm this as it is inside the coil housing potted in hot glue. The damping resistor will have some effect on the inductance value of the rx coil. For an experiment I added a 680R resistor in parallel with the rx coil and meter and the measured inductance dropped to 332uH. On the tx side it had little effect only dropping the measured inductance 1uH down to 315uH, probably due to the low resistance of the windings.

Cheers Mick

Thanks Mick,

nevertheless, I'm totally confused.

Open questions:
How many coil bundles in a DD coil? 2 or 3
Is the TX coil shared as RX (RX1=TX) in the receive period? Or is the RX1 separate from the TX?
Is the left TX coil (RX1) in induction balanced position to the right RX2 coil? (Overlapped TX, RX2 coil bundles?)
I can't make the comparison at the moment until I know the exact specifications.

Aziz

[sido]

Low noise amplifiers do make a difference even though the input circuitry is not ideally matched the preamp’s characteristics. e.g. I have recently been evaluating the LME49990 against the NE5534 which I have been using for some time. Without altering anything else, the LME appears to give less audio “wobble”. I need to do a proper recorded noise comparison to be sure though. Small figure 8 TX/RX used for this test.

Eric, are you refering to the impedence transfer between low amps has no effect on amplifier gain and low noise amplification??

I have recently been evaluating the LME49990 against the NE5534 which I have been using for some time. Without altering anything else, the LME appears to give less audio “wobble”.

What could be the cause? Frequency "BW"?.........The Frequency of receive operation?

The lowest noise I have achieved in a preamp was with a fet gating circuit between the coil and amp. input. This enabled me to use a 100R input resistor rather than 1K0 as before. Obviously it was the lowering of the resistor noise that was the important factor.

What's that descrete Pre amp op that Davor was talking about......Graeme Cohen??

One thing to watch out for, particularly with the 5534 used as a single gain stage, is that the gain does not recover quite as fast as you would think by observing the dc recovery trace on the scope. This can be checked by summing a sine wave signal (anywhere between 10kHz and 100kHz) into the input when the TX and coil are running; just enough to give about 1V pk-pk on the preamp output. Full gain is not reached until after 15uS in my setup. This problem is overcome if using two amp stages in the preamp. i.e. 1^st amp x20 and 2^nd amp x25 for x500 gain. The former causes an unbalance to the 2 sample earth’s field cancellation if you have an adjustable delay control.

Agree, Carl had this mentioned, but not implemented in his proposed 2 stage pre-amp consideration for the Hammerhead Pi.
Reduced lower input gain in first stage amp provides lower ideal coupling of unwanted signals entering the second stage.
Is this discrepancy "frequency" dependant?

I have found that Philips 5534's (with white lettering) have a noticeably wider bandwidth than Texas , or any other brand I've tried.

In respect to this, even the same manufacturer has discrepancies within then device specs, in return compromising an ideal uniform performance.


"But we are going to beat the best". This needs a bit of clarification. No detector is going to be the best at everything, as there are always tradeoffs.

Eric.


Totally agree..

Regards Sid

[Ferric Toes]

Davor,

Ultra low noise amplifier noise density := 1nV/sqrt(Hz)
(This is enough for us and we need some noise for the Gaussian distribution.)

The ultra low noise amplifier is important for AI configurations, where you have to compensate for the response losses. You can't beat the best without considering the amplifier noise. But we are going to beat the best.

Aziz

Low noise amplifiers do make a difference even though the input circuitry is not ideally matched the preamp’s characteristics. e.g. I have recently been evaluating the LME49990 against the NE5534 which I have been using for some time. Without altering anything else, the LME appears to give less audio “wobble”. I need to do a proper recorded noise comparison to be sure though. Small figure 8 TX/RX used for this test.

The lowest noise I have achieved in a preamp was with a fet gating circuit between the coil and amp. input. This enabled me to use a 100R input resistor rather than 1K0 as before. Obviously it was the lowering of the resistor noise that was the important factor.

One thing to watch out for, particularly with the 5534 used as a single gain stage, is that the gain does not recover quite as fast as you would think by observing the dc recovery trace on the scope. This can be checked by summing a sine wave signal (anywhere between 10kHz and 100kHz) into the input when the TX and coil are running; just enough to give about 1V pk-pk on the preamp output. Full gain is not reached until after 15uS in my setup. This problem is overcome if using two amp stages in the preamp. i.e. 1^st amp x20 and 2^nd amp x25 for x500 gain. The former causes an unbalance to the 2 sample earth’s field cancellation if you have an adjustable delay control.

I have found that Philips 5534's (with white lettering) have a noticeably wider bandwidth than Texas , or any other brand I've tried.


"But we are going to beat the best". This needs a bit of clarification. No detector is going to be the best at everything, as there are always tradeoffs.

Eric.

[Mechanic]

Hi Aziz,

Above where I have written "to cancel the combined emi from tx and tx1...."
Should read "to cancel the combined emi from tx and rx1...."

I have done the measurement of the ML DD coil and I got 316uH and about .4R for the tx and 354uH and 18R for the Rx. This is with the rx damping resistor still in parallel with the rx coil and I think the damping resistor was about 1k, though I cannot confirm this as it is inside the coil housing potted in hot glue. The damping resistor will have some effect on the inductance value of the rx coil. For an experiment I added a 680R resistor in parallel with the rx coil and meter and the measured inductance dropped to 332uH. On the tx side it had little effect only dropping the measured inductance 1uH down to 315uH, probably due to the low resistance of the windings.

Cheers Mick

[Mechanic]

HI Aziz,

From memory the tx on a DD is 310uH and the rx is about 450uH. The RX side is done with a fine litz wire and I think the resistance was about 40R including the parallel rx damping resistor. A bit later on I will measure these again to confirm....

Also with the magic tophat coil try tx(receive) +rx1+(-tophat rx2) with the windings/gain of rx2 to cancel the combined emi from tx and tx1....

Cheers Mick

[Aziz]

Hi all,

I think I should make a DD-PI coil vs. mono coil comparison for all three Rx modes in a ML-PI detector:
(According to ML detector manual.)
- Double D (TX left, RX+ right),
- Monoloop (TX left = RX+ left, RX+ right),
- Cancel (TX left = RX+ left, RX- right)

So ML fans know, what they can expect from the DD coil in their ML-PI's.
Unfortunately, the cancel mode requires, that the right coil part needs to be equivalent to the left TX part. So both coil halves will be matched to 300 µH. When doing this, one half RX does indeed induce 1/0.62=1.6 times less EMI noise (compared to the full round mono coil).

That's going to be interesting. Even to PJ.... *LOL*

Coil simulation is already running. Stay tuned..
Aziz

[Aziz]

Hi Mick,

yup, oscillations (interaction between TX and RX coil) can easily happen with different TC TX and RX coils (and different coil capacitances). When the RX coil induces some voltage the parasitic coil capacitance gets charged up and then the RX coil draws some current, which has to be critically damped. It would require its own damping resistor.

Well, I'm using a 5 strand ribbon cable for the RX coils. Although the insulation isn't ideal like teflon insulation. You don't need much solder for the wire joints.

Adapting the whole principle into an existing detector isn't trivial to meet all the critical timings, which are typically hard-wired in a detector. I think, the AI IB variants should be easier to go.

The VLF/LF detectors however aren't that critical. The coil capacitance doesn't matter much.

Regarding the planar DD AI: Yes, these achieve a good EMI noise rejection but also a good target response rejection (heavy loss). I should demonstrate this next time.

Cheers,
Aziz

[Mechanic]

Hi Aziz,

Of course I cant substitute 16 turns of 10 strand for 160 turns of 1 strand!!!! A bit of brain fade was going on there I think!!!!! LOL

Now using extra strands in parallel should help reduce the inductance and increase the resonant frequency of the coil which we want to match the tx coil or make the rx coils even faster.
When I was doing my AI mono coil trial setup, perhaps there was a mismatch in resonant frequency(AI Rx coil too slow) between the 2 coils, thus the frequency's being picked up were not matched. With my preamps I can select whether they add or subtract, so the phase orientation was not so important. The other problem I was having was that the tx coil was seeing the RX coil after turn off and in turn causing oscillations as if the coils were under damped and when the damping was adjusted so as not to oscillate, there was a long decay curve noted on the tx coil. So the coil setup was no longer as fast at it originally was with just the mono coil. I did have a damping resistor(variable) on the rx coil. I think from memory when I made 2 identical receive coils and tested them without a TX there was a significant reduction in noise, however it was not as quiet as a DD coil set in cancel mode, but I don't remember what the turns count of the rx coils were or the inductance, but I think it was somewhere around 500~600uH. Will have to make some new RX coils and try it all again!

Ribbon cable would be good to make coils from as all of the strands are tin coated for 1 and the wires are all evenly spaced. What about the spot where you join all of the strands with solder? The solder beads would have to be very minimal otherwise they will support eddy currents in to the receive period. How many strands wide cable do you use?

Merry Christmas to you all!

Cheers Mick

[Davor]

I have no problem with that. My point is that even with a real low noise pre you may beef up the Rx coil a little bit so that S/N becomes better than monocoil, and you actually can do that without much effort. 10dB is 10dB. It is not free, but it is well worth the effort and the extra wire.

[Aziz]

You often mention "ultra low noise preamplifier" lately, but to what end? We may assume an amplifier to be very low noise at 1nV/sqrt(Hz) and that's equivalent to ~60ohm source resistance. Going lower than that is not so practical because all designs with less noise draw enormous currents. Going along with a Muhammad and the hill proverb, it is a better compromise to reach the noise limit from the coil side, rather than the preamp side. A practical PI with first sample being later than 10us will be bandwidth limited to below ~100kHz, which would not be a significant limiting factor for coils up to maybe 10mH, and resistances ~30ohm, which is close to a noise limiting resistance. I'd say this separate Rx approach is much closer to the ideal S/N in terms of a possible turns ratio gain.

Davor,

Ultra low noise amplifier noise density := 1nV/sqrt(Hz)
(This is enough for us and we need some noise for the Gaussian distribution.)

The ultra low noise amplifier is important for AI configurations, where you have to compensate for the response losses. You can't beat the best without considering the amplifier noise. But we are going to beat the best.

Aziz

[Davor]

You often mention "ultra low noise preamplifier" lately, but to what end? We may assume an amplifier to be very low noise at 1nV/sqrt(Hz) and that's equivalent to ~60ohm source resistance. Going lower than that is not so practical because all designs with less noise draw enormous currents. Going along with a Muhammad and the hill proverb, it is a better compromise to reach the noise limit from the coil side, rather than the preamp side. A practical PI with first sample being later than 10us will be bandwidth limited to below ~100kHz, which would not be a significant limiting factor for coils up to maybe 10mH, and resistances ~30ohm, which is close to a noise limiting resistance. I'd say this separate Rx approach is much closer to the ideal S/N in terms of a possible turns ratio gain.

[Aziz]

A better magnetic flux linkage relation RX/TX of the graphics above:



The RX factors between 0.2 .. 0.3 are hypothetical and aren't convenient for us as you would require more RX loop turns (=more coil capacitance). I would even go for 0.4 and up.

Cheers,
Aziz