Simpler PI...

[mickstv]

Someone built audio power amplifier this way, to drive speaker directly, by paralleling some 40 op-amps, not OPA 37, but some LMC type, will post schematic of this contraption if i find it.

elektor had a kit using NE5532's


http://www.elektor.com/magazines/201....1529745.lynkx

[Waikiki_Sweep]

Interesting, how low noise can be achieved if several LME49990 (0.9nV/√Hz) will be paralleled?
Is there resource to get more depth for metal detector?

Signal-to-Noise Ratio ∞ √N since amplifier noise is uncorrelated.

Looks like it may help that super sensetive QED detector to be even more sensitive.

[Davor]

In case your input impedance is much lower than 50ohm you can theoretically reach lower noise than 0.9nV/sqrt(Hz), but your purse will get immensely lighter in a process. A discrete preamp with BJT transistors seem a more straightforward approach. Also a microphone transformer can get you there. But in all cases it is limited with source impedance.

[Waikiki_Sweep]

In case your input impedance is much lower than 50ohm you can theoretically reach lower noise than 0.9nV/sqrt(Hz).

It is great because coil itself has 1.3 Ohm resistance.
So just proper schematic idea is required to connect input straight to coil.
May be some new commutation or voltage limiting solutions will help.

[Davor]

Beside the coil resistance there is an uncompensated inductance, and it will ruin your low input resistance a bit. I don't feel like modelling it right now, but yeah, it works in that direction.
Have in mind that low input noise is also related with input device's quiescent current, and in all cases except the microphone transformer it will get high.

[Waikiki_Sweep]

Beside the coil resistance there is an uncompensated inductance, and it will ruin your low input resistance a bit. I don't feel like modelling it right now, but yeah, it works in that direction.
Have in mind that low input noise is also related with input device's quiescent current, and in all cases except the microphone transformer it will get high.

Looks like microphone transformer is a keyword here! Probably it is good solution to solve whole set of noise problems including parasitic capacitance to and galvanic connection with environment (ground, water, wet grass) or even radio frequency noises.
Finally we are working in audio frequency range and fighting against that noises problems which audio engineers already solved for that high quality and high sensitivity microphones and their preamps like SSM2019 that already mentioned on forum several times.

[Tinkerer]

Looks like microphone transformer is a keyword here! Probably it is good solution to solve whole set of noise problems including parasitic capacitance to and galvanic connection with environment (ground, water, wet grass) or even radio frequency noises.
Finally we are working in audio frequency range and fighting against that noises problems which audio engineers already solved for that high quality and high sensitivity microphones and their preamps like SSM2019 that already mentioned on forum several times.

The problem with the SSM2019, is that the noise figure is not good at all with low gain. only at gain 1000 it is good. However, we like to use a 2 stage preamp with low gain each.

The question arises: how important is the noise figure really at 10 Hz? for the input stage, since we chop the signal with sampling anyway?

Making a 2 or 3 amp differential instrumentation opamp with 0pa37's has good band-with, but problems with matching the resistors. Without precision matched resistors the common mode noise figure deteriorates quickly.

[Aziz]

Hi all,

I personally would take the NJM2068 instead of NE5532. It's an excellent and low noise dual-opamp and works even with low supply voltages. It's cheap too (0.57 EUR at Reichelt.de) .
Aziz

[Waikiki_Sweep]

The question arises: how important is the noise figure really at 10 Hz? for the input stage, since we chop the signal with sampling anyway?

Impulses with frequency less than 10Hz in your threshold perfectly imitating targets. Once you rise amplification to get that deep targets you will get falsing instead.
I don't think chopping will give any difference. All that 10Hz garbage will pass all the way from preamp to audio because it fits our target signal requirements.

[Davor]

On the contrary, 1/f noise acts as a wandering offset, and subtraction of the "earth field" - which is also a wandering offset - in effect turns your preamp into an autozero chopper, and that's precisely what gives PI edge over VLF.

[Waikiki_Sweep]

On the contrary, 1/f noise acts as a wandering offset, and subtraction of the "earth field" - which is also a wandering offset - in effect turns your preamp into an autozero chopper, and that's precisely what gives PI edge over VLF.

It works for frequencies less 1 Hz. Such very low frequencies (up to DC) signals sure have to be (auto) zeroed.
But imagine you moving coil over ground with speed 3 ft. per second. Set of targets will give you a signal in range 1-10Hz as result of move of your coil over detecting objects.
If you will try to zero signals in that range you will zero your targets. And input noise with frequency 1-10 Hz will act just like your set of targets.

[Tinkerer]

It works for frequencies less 1 Hz. Such very low frequencies (up to DC) signals sure have to be (auto) zeroed.
But imagine you moving coil over ground with speed 3 ft. per second. Set of targets will give you a signal in range 1-10Hz as result of move of your coil over detecting objects.
If you will try to zero signals in that range you will zero your targets. And input noise with frequency 1-10 Hz will act just like your set of targets.

The 1/f noise does not have a frequency of 1 to 10 Hz, it is random noise that is noticeable at these frequencies. Better opamps often show a scope picture in their datasheets, showing this noise. I think that if we take many samples and average them in the integrator, this noise is much reduced. However, when we amplify the 1 to 10 Hz output of the integrator, we again get new 1/f noise and amplify it. It seems that re-sampling and re-integrating would help.

[Waikiki_Sweep]

The 1/f noise does not have a frequency of 1 to 10 Hz, it is random noise that is noticeable at these frequencies. Better opamps often show a scope picture in their datasheets, showing this noise. I think that if we take many samples and average them in the integrator, this noise is much reduced. However, when we amplify the 1 to 10 Hz output of the integrator, we again get new 1/f noise and amplify it. It seems that re-sampling and re-integrating would help.

With differential integrator it may work. For example if noise has shifted output of opamp for 0.2 sec. and TX frequency 600 pps. than we will have 120 shifted samples to positive input of integrator and 120 shifted samples to negative input.
So if samples have same length and integrator build precisely low frequency noise have to be compensated. And high frequency noise will be just averaged. OK!!! I will go to buy noisy cheap opamp NJM2068!!!

Probably there is even a practical way to adjusting integrator for precise work:
1. Disconnecting TX and shorting coil.
2. Increasing gain of preamp to hear noise in audio.
3. Adjusting balancing pot (like you see in Hammerhead) between + and - inputs of integrator to position when noise in audio is lowest.
4. Return preamp gain to norm.

[Waikiki_Sweep]

OK, here is the circuit with a choice of clocks for either a pMOS or nMOS FET coil drive stage. You only need to make one clock using the second TL082. Both circuits produce pulses that are 50 uS wide with a frequency of 333 Hz using the values shown. Make R15 a 200 k pot to tweak frequency.

[ATTACH]28226[/ATTACH]
[ATTACH]28227[/ATTACH]
[ATTACH]28228[/ATTACH]
[ATTACH]28229[/ATTACH]

And the LTSPICE file for the whole circuit:

Because this schematic is very interesting and probably may be really small and cool device I have designed PCB by using authors sketch for P-channel MOSFET.
Not yet tried is it works at all. Hope it works. What replacement for 2N7000 I can use? Is J201 OK?









SPEAKER is PIEZOELECTRIC. Do not connect 8 or 32 Ohm or it will burn.

Board size 1.35" X 1.675" inches or 34.3 X 42.5 mm. Really small.

Any corrections required?

[Aziz]

NJM2068

Hi all,

have a look at the Japanese op-amp comparison table:
http://www.oocities.org/de/c_schirp/...comparison.pdf

The NJM2068 is around 3 nV/sqrt(Hz) input referred voltage noise density.
You can forget your NE5532/NE5534!

I also have found the SPICE models!
See here: http://semicon.njr.co.jp/eng/macro/index.html

Cheers,
Aziz