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

[Aziz]

Oh man,

LTspice can calculate totally wrong THD (total harmonic distortion) if you don't setup some internal LTspice parameters convenient. The AI solved the issue finally.
The op-amp version has very very low TDH of 0.000948% now. And I was wondering, why the pre-amp is so much bad (THD of 2.2%).
Nevertheless, the next version must be optimized further. I will take the AI suggestions into the consideration.

Cheers

[Aziz]

Hi all,

the dual op-amp version gets even much more quiet. The load drive resistance of the NE5532 can even go below the specified 600 Ohm. And we can go with the input referred noise density (en) below 0.3 nV/sqrt(Hz) at 45 kHz. This is a new record.

With a single ZTX851 (NPN) only. Paralleling of more ZTX851 isn't required at all. Would not give any benefit.

More later...
Aziz

[Aziz]

Hi again,

Ok, I have to set some critical node voltages containing capacitors with the .ic command with some other .op commands. This solves the issue and I am getting correct .ac and .noise analysis now.

One NPN input transistor (ZTX851) and one op-amp is as simple as the discrete solution. But the op-amp version is more stable.
Well, I have to start a bunch of circuit simulations. If I am satisfied with the results, you will get the circuit of course.

How about two output op-amps? The NE5532 (has two)? Parallel mode with lower noise and higher drive load capability? I will check this.

Cheers

[Aziz]

Hi all,

I will look at the transistor (NPN) + op-amp version back again. But I have to solve the buggy .ac and .noise analysis output (DC operating point issue).
The op-amp version delivers a rock solid gain stability of course.

For some reason, LTspice doesn't find the optimal operating point in the circuit simulation. I have to ask the AI and look for a solution.
Aziz

[Aziz]

Hi all,

the power consumption was too much of the latest discrete amplifier. By reducing the collector currents and changing some parts value, it is still possible to get good low noise performance.
I have reduced the power consumption to 9-10 mA now.


This simple circuit has two major disadvantages however:
- Gain is dependent on temperature changes (can change up to 5 %)
- Gain is also dependent on battery voltage level (battery voltage diminishing)

It may cause heavy drift issues on ground balance part. I'm sure, this can be solved in the software part. This is the price of being simple circuit.

To overcome the issues, we have to use an op-amp for the second stage. I have to look for an old design yet.
I have prepared the LTspice commands for the simulation. Just activate them and look at the circuit simulation.

Cheers,
Aziz

[Aziz]

LTspice file ..

[Aziz]

Hi all,

after simulating over a dozen discrete transistor AC amplifiers, I have finally found a good solution.
Features:
- two stage discrete transistor amplifier (2x NPN transistors)
- enough gain (> 200)
- auto biasing, simple design, with gain loop back compensation and high frequency filter
- ultra fkn low noise design (0.314 nV/sqrt(Hz) @ 45 kHz operating frequency)
- less distortion (THD ~ 0.08 %)
- good voltage swing (between GND and source voltage rail)
- single supply (9 V)

But it has it's price: ~ 20 mA supply current. The 9 V block battery will get quickly empty.
On the other hand, it is an ultra low noise design and still a simple circuit. No op-amps required.
Single NPN ZTX851 (the low noise transistor), which is cheaper than the PNP ZTX951 variant. The second NPN (output stage) is a very cheap NPN transistor (what you have in your box).

This is the current schematics:

It might work well for my application. But I have to try it out.

Cheers,
Aziz

[Aziz]

Hi all,

due to the IB nulling mixer, I need more gain in the amplifier.
A single stage transistor amplifier won't get enough gain. A two-stage will do it with negative feedback loop gain control. This has less distortion and gain stability benefits.

It will still have good enough en of 0.4 nV/sqrt(Hz) with NPN ZTX 851 transistor.

I'll continue in 12 days. No free time at the moment.
Cheers,
Aziz

[Aziz]

Hi all,

back to the roots.

This basic single stage PNP transistor AC amplifier outperforms any op-amp designs. Single supply (9 V).
At my operating frequency of 45 kHz, en is 0.3 nV/sqrt(Hz).



Oh fck!, the PNP ZTX951 isn't available and is much more expensive than NPN ZTX851 now. What happened?
No problem, we can do this amplifier with NPN ZTX851 type too. Would get approx. same low noise performance.

I have to check it out yet, whether temperature drift and gain stability would cause much problems.
Aziz

[Aziz]

Hi all,

I'm so sorry due to the lack of time. I won't have for the next 11-12 days much free time as well.

Dual frequency VLF project changes:
- TX mono coil channel:
Active nulling on mono coil TX does not make any sense. A TX coil inductance change of 0.5 - 1 % due to heavy ground mineralization would definitely overload any planned amplifier stage. The reactive response on the mono coil TX is 2-3 orders high.
I will leave the mono coil channel as is (no active nulling, no amplifier).

- IB-RX channel:
An high-gain amplifier will be used for the IB-RX channel (either figure-8 or true IB-coil configuration). Due to IB configuration 2-3 orders of reactive response is avoided. But the hot ground has still up to 3-4 orders more reactive response compared to PI technology.
In this configuration, active nulling makes sense. So the IB-coil setup should be easy and need not to be perfect induction balanced. Gain of 100 (40 dB) should be enough. An ultra-low-noise amplifier design makes sense in this case.
I tend to use a simple discrete transistor based bipolar amplifier technology instead of op-amps. A ZTX951 bipolar transistor (pnp) with 5-9 mA collector current should deliver excellent ultra low noise figures.
If we use a figure-8 RX coil, we have to amplify more of course. An active nulling on the IB-RX channel would avoid overloading at high gain.

I am planning to use a simple thin shielded stereo coil cable for the coil (3-wire).
- Ground (cable shielding)
- TX (in)
- RX (IB-channel out)

TX-out or TX-reference will be taken out of the controller box/circuit board.

I don't have any idea, how to realise this configuration yet. I will start with the simple amplifier first. Well, we will see..

Aziz

[Aziz]

Hi all,

so far, the j-fet input op-amp OPA2134 does not get any significant improvement. It is more noisy.
I can use even the NE5532 in direct mode as the non-inverting (+) input does have enough high input impedance. En (input referred noise voltage density) is approx. 9.3 nV/sqrt(Hz) in direct mode.This is even slightly better than OPA2134. On the other hand, the OPA2134 is a rail-to-rail op-amp providing more dynamic range.

I will look at the ultra-low-noise front-end with the BJT ZTX951 (PNP) + NE5532 as a feedback loop driver. This will reduce the noise density 10 times.
And yes, it makes really sense to use the ultra-low-noise front-end in this case.

I will be in Berlin for the next 5 days. The analysis will be continued, when I am back.
Cheers,
Aziz

[Aziz]

Hi all,

I am testing different solutions via LTspice simulations.

Mode of transmitter
- Direct injection
- Transformer coupled injection

Active nulling via:
- Simple mixer ( a*(Vsig+Vcarrier) + b*Vcarrier = 0)
- Difference amplifier topoplogy

Amplifier topology regarding noise generation and source impedance issue:
- bipolar amplifier (NE5532)
- j-fet input amplifier (OPA2134)

There are a lot of trade-offs, design considerations, pros/cons, etc.. depending of the mode.

So far:
The NE5532 can be operated at approx. 7 nV/sqrt(Hz) input referred noise at operating frequency.
The OPA2134 at approx. 9-10 nV/sqrt(Hz).
Do we really need input referred noise of <1 nV/sqrt(Hz)? How much would cost us using some more parts to the NE5532?

It takes time to travel and analyse all the possible ways.
Aziz

[Aziz]

Hi all,

the LTspice simulation confirms, that the complex nulling signal reference can be found easily by varying amplitude and phase of the two independent frequencies f1 and f2.
Four parameters for nulling:
A1 (amplitude f1),
Ph1 (phase f1),
A2 (amplitude f2),
Ph2 (phase f2)
for exact nulling reference signal. So we can amplify the difference (target signal) easily.
The nulling will be done with two small capacitors instead of resistors.

The amplifier will add some noise but it isn't really critical. We can design it to produce less noise.
Aziz

[Aziz]

Hi all,

the next step is getting most out of the mono coil VLF system with active nulling and amplifying the difference. For active nulling, I have to use one of the line-out channels for the nulling signal. And I have to drive the LC-tank with single line-out channel this time (single ended drive). Note: dual frequency drive.

Realising the search coil as a transformator (primary + secondary coil) has still huge benefits:
Impedance and voltage transformation. The secondary coil (the real resonant LC-tank) won't be loaded much and is keeping a high Q. The high voltage in the LC-tank is down transformed and I don't have to touch the resonant LC-tank. Minimizing additional cable capacitance effects, etc..

In this case, I don't need any j-fet buffer or j-fet input op-amp. The NE5532 would be sufficient enough and has enough high input impedance.

Two possible configurations are available:
1. Pure mono coil VLF only, single op-amp as amplifier used, the other op-amp as 0 V buffer (no floating inputs).
2. Mono coil VLF + a third figure-8 coil inside the search coil as a secondary signal source, both op-amps used.
The LC-tank has enough coil current for detecting target signals in the figure-8 coil. The second op-amp will amplify the signals of the figure-8 coil.

Let's start with the option 1 first.
Aziz

[Aziz]

Hi all,

I am looking for low noise jfets. I have finally found a relative cheap dual n-jfets (SMD).
NSVJ5908DSG5 (n-ch, 0.83 EUR)
It's noise figure is 1 dB.
It can be used in a fully differential stage or single ended and paralelled jfets.

I have not decided yet, which configuration I am going to use it.
Aziz