Simpler PI...

[Aziz]

Improved PSRR

Hi all,

the initial amplifier circuits were prone to power supply variations or high frequency regulation ripple/noise. This improved version greatly improves the power supply rejection ratio (PSRR). Although, we have some voltage drop loss of approx. 0.8 V, there is still enough voltage swing range for the output.
Notice: the power supply filter has been "stolen" from the AN-222 publication mentioned here.

I have taken two parallel transistors without emitter load balancing. This is just a preparation for the intended BF459 usage. Note, that the 10 Ohm input resistor greatly degrates the noise performance. I hope, I can reduce it in the real usuage. Fingers crossed.

The new schematics:


Did I win the great prize? *LOL*

Cheers,
Aziz

[Aziz]

Aziz, un-bypassed emitter degeneration is proven beneficial for the 1/f noise effects in oscillators. I'd expect much faster recovery without capacitors there, and if you remain in ~10ohm ballpark you should not have noise spoiled either.

Hi Davor,

could you make a simple sketch or circuit, which describes what you mean please? I'm not sure, whether your proposal will work.
Aziz

[Davor]

Aziz, un-bypassed emitter degeneration is proven beneficial for the 1/f noise effects in oscillators. I'd expect much faster recovery without capacitors there, and if you remain in ~10ohm ballpark you should not have noise spoiled either.

[Waikiki_Sweep]

hello waikiki

did you build a prototype of your circuit already?

It is Beenthereuk schematic and I have tried to convert it to PCB in my post #29.
Not finished yet. As I understand N-mosfet have to be replaced with P-mosfet near speaker in that schematic with proper correction.
I see no more updates from Beenthereuk. I may try to finish it later. But Beenthereuk already build it and told that it sensitive to big pieces of metal only.

Right now I am trying to finish one of my ideas to fix a problem that gives me most of the pain.
I need a PI detector with iron discrimination to stop braking my arms and legs by digging all that iron trash. Easy way is to use VLF but not in salt water.
So I have added balanced "8" shape RX coil to my PI machine. As I see any other solution, even Chance PI, will gets me to dig that nasty beer caps.
I can use Fisher CZ but it reports beer caps in nickel range too. I can buy Excalibur II for $1500 but I will brake my hands moving it in water.
Another way is build hydrodynamically shaped coil for Excalibur but I am not sure in success and anyway DD coil will make more drag in water than monocoil.
Idea is to use common PI schematic with monocoil and read iron/color metal indication from separated hydrodynamically shaped RX coil with motion discrimination channel to provide VCO sound to PI circuit.

[Aziz]

Now a really bargain BJT:
BC550C (0.05 EUR)

It will work a treat and super low noise. There are more equivalent low noise BJT's out there. Almost every transistor, which has a lower noise figure than 3-4 dB will work very well. The BC550C should have a typical noise figure of 1-2 dB.

Ok, I don't have to wait. I can't wait. I'll start the soldering soon...
Cheers,
Aziz

[Aziz]

Alternate BJT Transistor

Hi all,

I have found another good BJT transistor:
MJE 340 (0.27 EUR)
As good as BF 459.

You can use NE5534 instead of OP27 of course. But the OP27 performs better.
NE5534: 0.25 EUR or 0.33 EUR A-grade
OP27: 1.70 EUR

Aziz

[Aziz]

Some Noise Figures

Hi all,

I have increased the RG1 to 4.7 Ohm (from 3.3). Depending on the input resistor Rinp, I get the following noise figures:
Rinp = 10 -> 0.64 nV/sqrt(Hz)
Rinp = 22 -> 0.8 nV/sqrt(Hz)
Rinp = 33 -> 0.9 nV/sqrt(Hz)
Rinp = 47 -> 1 nV/sqrt(Hz)

(Rinp = 1 kOhm -> 5.4 nV/sqrt(Hz) )

It seems, that I can have some input protection up to 33 Ohm input series resistor and I'm still in the ultra low noise ball park.

The power consumption is around 6 mA. Oh well, I can live with it.

BTW, one have to set the operating point accurately by adjusting the emitter resistor (R4). Set R4 to get the average output voltage in node vout (output of the op-amp) at Vcc/2, so 8.5/2 = 4.2 V. Best done with a combination of resistor and trimmer network.

Cheers,
Aziz

[bernte_one]

hello waikiki

did you build a prototype of your circuit already?

[Aziz]

Some Remarks..

Hi all,

I want to add some remarks before someone starts the soldering..

The latest ultra low noise AC amplifiers won't work in your standard PI circuits of course. They aren't DC coupled as you would require in most PI circuit cases.
They presume the induction balance (IB) coil configuration, where the high voltage from the (transmitter) coil is absent. They require very low impedance path to the RX coil to achieve the superior low noise performance.

In the traditional PI:
You have the 1 kOhm series resistor with anti-parallel diode high voltage proctection circuit usually. Unfortunately, the 1 kOhm resistor connected in series to the coil will kill all of the low noise performance. The resistor noise density of 1 kOhm is approx. 4 nV/sqrt(Hz) and you can dream of ultra low noise designs. We would require a low impedance path switch via input fet switch to overcome the 1 k series resistor.

I might increase the 3.3 Ohm resistor to 4.7 Ohm or even 10 Ohm. And I might put a 10 Ohm series input resistor just to limit the current during the high voltage spike periods (if there). And the noise performance will be affected slightly but I would have a save operation.

Cheers,
Aziz

[Aziz]

C5 will blow... Boom.... Everything else looks attractive, in simulation. Lets wait for results of soldering.

Magnetic field/geomagnetic field cancel have to be balanced well to kill that low frequency noise. With 5% tolerance capacitors it may let some part of noise to pass.

No, C5 won't blow. It has the right polarity.
There isn't any serious input high voltage protection there (except the two diodes). I hope, I don't have to increase the series input resistor.
Booooooooooooom! *LOL*
Aziz

[Waikiki_Sweep]

C5 will blow... Boom.... Everything else looks attractive, in simulation. Lets wait for results of soldering.

Magnetic field/geomagnetic field cancel have to be balanced well to kill that low frequency noise. With 5% tolerance capacitors it may let some part of noise to pass.

[Aziz]

Um didn't we want it low noise at 3 - 10 hz? That's the noisy part in most designs...

Hey Silver Dollar,

the voltage noise density gets high at lower frequencies, but the gain is low too. Voltage noise density isn't equal to total noise energy (integrated over a frequency range). And there is an inherent high pass filter.
In my application, I'm using frequencies above 1 kHz. That's the reason, why I can use the simple AC amplifier. All low frequency region isn't important to me.

The low frequency amplifier noise isn't important in a PI as well, which has the static magnetic field/geomagnetic field cancel feature.
Aziz

[Silver Dollar]

Um didn't we want it low noise at 3 - 10 hz? That's the noisy part in most designs...

[Aziz]

And this is the single transistor, single op-amp version (OP27). Still below 0.5 nV/sqrt(Hz).
Note, that OP27 has minimum load drive of 600 Ohm. In this circuit, it has 3.3 Ohm*200 = 660 Ohm. We are above 600 Ohm. That's ok.


I will use this amplifier in my next laptop detector controller project. It has been optimized for 8.5 V single supply voltage therefore. The 8.5 V needs to be stable of course (from LDO voltage regulator). And no transistor matching required.

Did I win the competition prize now?

Cheers,
Aziz

[Aziz]

Very Ultra Low Noise AC Amplifier

Hi all,

I have forgotton to mention, why the perfect matching of the transistors isn't much important here:
We don't have a differential pair amplifier topology here.
Just parallel the (input) transistors and take care, that they have enough collector current for each transistor. I have chosen a collector current of 1 - 2 mA each (actually 1.5 mA).
And we can load balance the emitters with one elco and one 100 Ohm resistor for each transistor.

Ok, I have taken the OP27 for the NJM2068 for the noise performance simulation. Both have around 3 nV/sqrt(Hz) noise density. So a dual op-amp can be used to minimize further noise and has more current drive capability in the gain resistor divider network. You can use the OP27 for the single op-amp version too. Just take care, that the resistive load does not go below 600 Ohm. You can use a series resistor at the op-amp output (just like in the dual-opamp version) if you have a very small gain resistor divider network (< 600 Ohm series).

You can use the NE5532 too. The noise density spec just increases marginally (totally neglectable IMHO). But the OP27 is a better choice for the single op-amp version (has more output voltage swing).

And you can use a single transistor as well. Then leave the emitter load balancing part (elco + 100 Ohm resistor). Noise density still way well below 1 nV/sqrt(Hz).

This is just to give you some ideas:


The "Ultimate Thermo Nuclear Melt-Down"(c)(r)(tm) is coming.... *LOL* "Rette, wer sich kann!"

Cheers,
Aziz