preamp types

[golfnut]

Looking around

most preamps discussed are opamp types. Looking at this in detail the noise figure of these things are not good.

With £10 op amps its tough to get a NF of less than 5dB (according to this link)

http://www.youtube.com/watch?v=BTAQ07gG2CE

Reading through some of the Bipolar specs - you could out perform a £10 opamp with a £0.50 bipolar.

Further you could get a <2dB NF amp from a fet like a 2SK3557 for £0.35


good reads

http://users.tpg.com.au/users/ldbutler/AmpNoise.htm

http://www.vishay.com/docs/70599/70599.pdf

Steve

[golfnut]

This is a fairly new good opamp - check out the NF though

S

[Carl-NC]

Noise figure isn't usually applied to metal detector design. Besides, I can easily get an outstanding NF with a lowly NE5534 opamp.

[golfnut]

Sure, Im aware NF isnt often raised - which is odd to me.

I cant explain why the guy in the nat semi vid sees ~6dB NF on his amp.

I looked at the 5534 spec - your right 1-2dB should be reachable with 5k source res..

Sub 2 should be fine. I get jittery when I convince myself of 6 plus

Im missing a few basics somewhere - gotta sharpen up - its been a while since I did my cellular stuff.

S

[Aziz]

The Poor Man's Ultra-Low-Noise Amp Tip:

BC327-40 & BC337-40 in a differential input stage with a low noise op-amp as feed-back driver (NE5534, OP27 or similar). Both n and p transistors need to be matched.
The higher beta of the transistor pair, the better.

Comes as close as possible to the 1nV/sqrt(Hz) amplifier.

Aziz

[golfnut]

Reading more over lunch

1) The worst noise performer is the full differential (like IGSL)

best is non inverting

2) The coil is your Rs and its about a few hundred ohms Reactive so generates little Voltage noise

useful to us as coil has only tens of real ohms - so good.


3) As Rs is small the opamp input Voltage noise is dominant. - rather than the opamp input current noise.

good, pick an amp with low input volts noise.


4) Best place for amp is in coil head.

Jury is out - is the cable loss a significant contribution at few kHz ?

S

[golfnut]

Saying the above this Ti paper leave me cold a bit. These NF are high,

[Tinkerer]

Maybe you would like to apply your knowledge and expertise on a real front-end?
At the post below, we are redesigning the RX front-end.
I am thinking of using the AD621.

Your input would be most appreciated.

http://www.geotech1.com/forums/showpost.php?p=146595&postcount=5

Tinkerer

[Skippy]

Aziz:You can obtain 1nV/root-hertz noise levels from cheap opamps nowadays. The Fisher F75/ Teknetics T2 machines use an opamp with that spec, and they perform quite well. To take advantage of the discrete front-end, you should be looking at parallel matched pairs/triples of npn transistors. Figures of 0.3 nV/root-hertz should be achievable that way, especially if you don't use the long-tailed differential pair.I plan on building one this year, just for curiosity. I've matched up a load of BC337's ready.

[Carl-NC]

GN, save noise figure for radio systems, don't worry about it for metal detectors.

Here's some examples to show why; calculate the following noise figures:

1. A typical PI front end; a 300uH coil has a 680-ohm damping R and a series 1k clamping R with clamp diodes; applied to a NE5534 with 4nV/rtHz.

2. Same PI front end, but the series R is magically reduced to 100 ohms.

3. VLF front end, where the RX coil (10 mH/10 ohms) is applied to a NE5534 preamp.

You'll quickly see why we don't give a squat about NF.

[Tepco]

High noise amps

What about "high noise" video amps (old and partially obsolete) like NE592 (LM733), trading noise for speed? Specified 12uV over 10MHz BW is not that bad, heavily overdriven recovery time is in tens of nS range, eliminating speed and recovery issues completely. I never used this in detector, but in test rig made to measure target response times etc, did anyone tried something similar in actual detector design?

[Aziz]

Resistor Noise

Hi all,

to give you an impression of the resistor noise (thermal noise), I have prepared a table for you:
(at temperature = 300 K (27 °C) )
1 Ohm: 0.129 nV/sqrt(Hz)
10 Ohm: 0.407 nV/sqrt(Hz)
50 Ohm: 0.910 nV/sqrt(Hz)
100 Ohm: 1.287 nV/sqrt(Hz)
500 Ohm: 2.878 nV/sqrt(Hz)
1 kOhm: 4.070 nV/sqrt(Hz)

Pretty fine if you can go for a 10-50 Ohm series resistor and 1nV/sqrt(Hz) amplifier. But have a look at the induced EMI!!! The EMI noise is much, much bigger.
Aziz

[Davor]

True. That's why Carl is so resolute about noise being a non-issue here.

There is, however, one other facet of the issue: optimum source resistance. It is calculated from the specs voltage and current noise performance (Vn and In -> Ropt=Vn/In). With NE5534 you get Ropt at ~9kohm, OPA134pa would have Ropt ~ 1800ohm, while with LT1115 you get Ropt at ~750ohm. Hence, LT1115 is better for low Z operation. For LMH6629 Ropt reads 350ohm but with this price...

It only means that for source resistance above the optimum, noise will bi dominated by resistor noise only.

Check "Voltage Noise vs Source Resistance" for further ideas.

It can be improved with liquid helium though

[Carl-NC]

1. A typical PI front end; a 300uH coil has a 680-ohm damping R and a series 1k clamping R with clamp diodes; applied to a NE5534 with 4nV/rtHz.

The damping resistor and diodes make no difference, only the 1k series. It has a NSD of 4nV/rtHz, same as the input-referred noise of the opamp. Ignoring the noise of the opamp's feedback resistors (which I can easily make insignificant) the noise figure of this preamp is 3dB.

2. Same PI front end, but the series R is magically reduced to 100 ohms.

Despite the fact that we now have less noise than with #1, the noise figure of the preamp has increased to 10.3dB!

3. VLF front end, where the RX coil (10 mH/10 ohms) is applied to a NE5534 preamp.

The noise figure for this case is 19.9dB. Holy Smokes, this sucks! Or, it just doesn't matter.

- Carl

[detectormods]

Most engineers do not take the actual ground noise into consideration, even at 17 c (290K) the noise figure from the ground is 3db so using a preamp with a lower noise figure just amplifies the ground noise. The important aspects of the input stage is dynamic range, slew rate, settling time and low distortion.

http://cp.literature.agilent.com/litweb/pdf/5952-3706E.pdf

The Link provides some useful information on noise measurement.