Induction Balance Stuff - Single/Multi Frequency Response, GB, Disc, Measurements, Ideas, Fun, etc.

[Aziz]

Hi all,

I am almost finished with the pre-amp design. AI chat helps to optimize it further. But Google AI is talking at least 50 - 60 % bull$hit.
I am operating the ZTX851 on its sweet spot now (at linear operation point). And I am getting excellent THD figures.
A rock solid gain stability at various temperatures and battery voltage levels.

With the NE5534, the 10 Mhz GBWP (Gain Bandwidth Product) allows a gain of approx. x 200 for signals up to 50 kHz.
If we really need more gain, we can upgrade the NE5534 with OPA1611.

Aziz

[Aziz]

Hi all,

this is it. The ultra low noise AC pre-amp with compromises.

With Rs=2 Ohm (RX coil resistance) and even with relative low gain feedback loop resistor R8 of 470 Ohm, it is achieving at my operation frequency of 45 kHz 0.35 nV/rtHz input referred noise density. This is acceptable and realistic. The op-amp should be able to drive the R8 with ease without any significant THD degradation.
The capacitor C4 (Cf) is absolutely mandatory for stable operation of the NE5534 (uncompensated).
The capacitor C2 (100 nF) at the non-inverting input of the op-amp reduces further noise.
Some capacitors must have low ESR or must be C0G/NP0 type.
The Diode D1 is to protect the transistor Q1 during start up and down conditions (or electrostatic stress cause). It prevents reverse biasing of Q1.
Input impedance Zin is 5 kOhm (Zin = R1 || R2 )
I have taken C1 as high as possible to lower the noise (lowering source impedance) and to cut-off the unused low frequency part.

BTW, lots of op-amp spice models do not model the dynamic power supply consumption. So the total power consumption is 10 - 12 mA depending on the source signal level at 9 V battery operation.
AI tought me a lot. But I have corrected him a lot more.

This is the schematics:

The zipped LTspice model file will be added for your convenience.

Cheers,
Aziz

[Aziz]

Hi all,

upgrading the NE5534 to a better op-amp and higher GBWP isn't really required. The 10 Mhz GBWP of the NE5534 is by far fast enough to handle even much higher gain.
This is owed to the fact of fast transistor ZTX851 (100 Mhz) and is doing the most part of the work.

I'm keeping it at the very cheap NE5534 solution. As the NE5534 isn't internally frequency compensated (for fast operation), we have to do it with C4 (Cf) in the feedback loop. Without C4, the amplifier will turn into an oscillator at high frequencies.

We can even rise C4 to 220 pF or more to reduce the bandwidth of the amplifier and making it more stable.

The amplifier is capable to go down to en of 0.28 nV/rtHz, if we set Rs=1 Ohm, R9 = 1 Ohm, R8 >= 470 Ohm (high gain).
This is the minimum low noise figure I can get out of the circuit.
Parallelling more ZTX851 does not give any significant benefits (the amplifier stability degrades heavily).

Cheers
Aziz

[Aziz]

Hi all,

I think, we have everything to finish the High-Q dual frequency mono/IB-detector now.
The TX-coil current is big enough (200 - 300 mA peak) to deliver enough signals in the IB-RX-coil. If we use an anti-interference (figure - 8 ) IB-RX-coil, we can push up the gain as much as we need.

The search coil will be a combo coil: acting as a mono coil and IB-coil (both signals will be processed). The most sensitivity comes from the RX coil of course.
We get good and strong ground signals at the TX coil, which will even help to improve the GB on the RX-channel.

We have to be careful, not to overload the RX-channel (amplifier) by heavy ground signals. So the final gain depends on it.
Aziz

[Aziz]

Hi all,

I have managed Claude AI for giving a short review of the latest AC amplifier schematics.
Claude AI is damn good but I am often running short of time limit.

Anyway. Interesting report (PDF).

Aziz

[Aziz]

Hi all,

look, what I have found: A matched pair of ZTX851 for true differential amplifier stages:

https://ka-electronics.com/shop/index.php?route=product/product&product_id=118

25 bucks, Vbe match < 1 mV. Thermally sealed. Nice.
2 x ZTX851 cost about 1.5 EUR.

You can buy 33 ZTX851 and look for a good match for the price.

Claude Code has suggested a DC-servo control for absolute stable mid-point (Vm) stabilisation (DC operating point stabilisation).
But this is not really required. Any low frequency change (drift) won't be seen at such high operating frequencies.
An offset voltage drift of Vm by approx. 200 mV caused by temperature changes is acceptable.
I'll keep the circuit as simple as possible. I will even leave R5 and the trimmer UP1 with appropriate emitter resistor R4.

Aziz

BTW, Claude Code is also talking a lot of bull$hit.

[Aziz]

Hi all,

the RX coil impedance (not DC resistance Rs) would cause heavy noise in the pre-amp of course (current noise) and will defeat any effort in ultra-low-noise design.

We have to go near the resonance frequency (operating frequency) by adapting the coupling capacitor C1 (in conjunction with the RX coil inductance).
The series reactance must be minimized down to 0 Ohm.
If we don't match the input impedance, we will go up to 0.7 - 1nV/rtHz. I'll check this in the LTspice simulation next time.

[Aziz]

Hi all,

regarding my last post (#427):
This bull$hit advice was taken out of Claude Code AI. And I have trusted in it this time until I made the spice simulation right now.

Yet another big mistake of AI.
Oh man, we are measuring only (tiny) voltages. There is almost no current flowing through the RX-coil (µA only). We don't have to match any impedance.
Leave C1 as specified.

AI is absolutely not good in electronics engineering.
Aziz

[Aziz]

Hi all,

regarding the IB nulling:

We specify a maximum residual IB RX-coil voltage of 50 mV peak. The RX-coil must be induction balanced to meet this specification. This is a manual process during the making of the coil.
The residual RX-coil voltage alone would totally overload our amplifier of course. So we have to electronically null the residual RX-coil voltage.
And we specify an IB nulling correction voltage source of 3 V peak (standard output voltage of the sound card).
Line output of the sound card:
Channel 1 will feed the LC-tank (TX) and
Channel 2 will deliver the IB nulling correction voltage

So we need a nice mixer to electronically induction balance nulling the residual RX coil voltage to avoid the overloading the amplifier.
Output impedance of the mixer at operating frequency must be very very low for ultra low noise operation however.

Fortunately, the amplifier has a built-in mixer in the input stage. Let's add an additional input line to the amplifier.
Input 1: RX-signal out of the IB RX coil
Input 2: IB nulling correction voltage.

If we look at the amplifier, the input signal will be fed via C1 (330 nF) capacitor to the base of the transistor Q1.
Let's take it and make a capacitive mixer and realise the second input line to the amplifier for the IB nulling correction.

We have to translate the high 3 V correction voltage to 50 mV range.
Factor: 3 V / 50 mV = 60
If we have for C1 330 nF, the second line capacitor coupling input must be:
C = 330 nF / 60 = 5.5 nF -> nearest is 4.7 nF

There we have our good mixer and saved some additional parts.

The following schematic shows the modified amplifier with two input lines:


The ultra low noise operation is not affected by the mixer as the RX coil impedance line is still very very low. The impedance of the IB nulling input is high. The voltage Vib (+noise in the channel) will be divided by factor 60. The signal (Vsig2) will not be attenuated much (1/60, approx 2%).

In the schematics above, L2 is my RX coil. This is not showing the real situation of course. It was required to confirm/review the bull$hit the AI was telling me (post #427, #428 ) .
So we have an amplifier with two inputs now.

Cheers,
Aziz

[Aziz]

Hi all,

for all ultra low noise freaks and "audiophile" metal detector frontend designers:

Must see link:

https://www.hifisonix.com/articles/mc-head-amp-circuit-compendium/

Includes a lot of info and circuit examples (pdf + spice simulation files in a zipped file).
Great finding.
Aziz

[Aziz]

Hi all,

I have used the original Zetex ZTX851/ZTX951 spice models. For exact noise analysis, RB parameters are set wrong or set for (high current) switching characteristics in those models.

The typical measured RB parameters are:
ZTX851: RB=1.6 Ohm
ZTX951: RB=1.2 Ohm

So all the nice noise analysis get slightly worse.
No worries, we are still below en=0.4 nV/rt Hz.
I have to change all the spice model files now.
Aziz

[pito]

e are still below en=0.4 nV/rt Hz. = in Upp nV for 10 kHz signal ?

[Aziz]

Hi pito,

e are still below en=0.4 nV/rt Hz. = in Upp nV for 10 kHz signal ?

if you rise C1 a bit, then yes: (Gain=210, Rf2=470, Rf1=2.2, Rs=2 Ohm)
C1=330n, en=0.42 nV/ rt Hz @ 10 kHz
C1=470n, en=0.40 nV/rt Hz @ 10 kHz
C1=1µ, en=0.39 nV/rt Hz @ 10 kHz

But I want to use frequencies above 30 kHz up to 48 kHz. This is the reason, why I kept C1 as low as possible.
Aziz

[pito]

Vpp ?

[Aziz]

Vpp ?

I don't know, what do you mean by that.