low noise amplifier - can i build it?

hi everyone,

i'd like a bit of advice here: i need a preamplifier for measuring very small vibrations on a piece of scientific kit. the sensor (a geophone - i=2Ee. a coil and magnet) will output voltages of about 0.1 microvolts (in the range 0.1Hz to 1kHz), so i need to get hold of a really low noise preamplifier.

firstly, does anyone know of a company might sell me what i want at a reasonable price, i.e. less than $500 (=A3300)?

also, do you think i can build it myself to these specifications? i have some experience with op-amp design and i know the basic theory, but i'm not sure if i can actually get a good enough noise performance.

thanks in advance

roko mijic

Reply to
Roko
Loading thread data ...

formatting link

The LT1128 is a low-frequency version of the LT1028. I had not noticed the huge 275pF high-voltage on-chip capacitor, one of the largest I've seen. Perhaps Jim has seen (or used) bigger?

Anyway, Roko, back to your geophone amplifier... If it's low frequencies you're interested in, the LT1128's spectral-noise density starts going up below 250Hz, an effect that's called 1/f noise. Thankfully its impact is not as dramatic as the graph, page 5, implies, because when you integrate the noise density to get the RMS noise level (and hence the peak-to-peak noise), the sqrt-BW part of the calculation reduces low frequency influence.

A spectral noise of 0.9 to 2nV should be compared to the Johnson noise of associated resistors in the circuit, e_n = sqrt (4kTR) = 1.28nV for 100 ohms, and scales with sqrt R. For example your geophone coil's DC resistance may be much higher than 100 ohms. You also need to keep the value of critical gain resistors low.

There's 40 pages of detailed discussion about low-noise issues in our book, The Art of Electronics. Recommended. :>)

--
 Thanks,
    - Win
Reply to
Winfield Hill
[snip]
[snip]

275pF, if square, would be about 565u X 565u, or about 22mil X 22 mil.

I've not seen one that big before, but if you've got real estate to burn I guess it's OK ;-)

...Jim Thompson

--
|  James E.Thompson, P.E.                           |    mens     |
|  Analog Innovations, Inc.                         |     et      |
 Click to see the full signature
Reply to
Jim Thompson

Right, and hence a noise source.

And I'm quiet sure I've purchased geophones with lower resistance. But they do tend to get up there, in their attempts for sensitivity.

--
 Thanks,
    - Win
Reply to
Winfield Hill

You really need the spec sheet or to measure the resistance. I have seen some units with very much higher resistance:

formatting link

Reply to
John Popelish

colin wrote: (snip)

(snip)

I have been reviewing opamps for noise in this frequency range, but haven't yet included the AD8552. But you might look at the auto zeroing LTC1047 and LTC1051 and OP2604 or OPA604.

Reply to
John Popelish

A transformer in the front end might make things a little easier. What's the impedance of the coil?

Reply to
Ian Stirling

Just to make sure the system cannot be excited and a lock-in amplifier cannot be used. Lownoise amplifiers are bought, not really built. Since the source impedance appears to be low, have a look at the LT1128, it is specified at 0.9nV/rtHz

Rene

--=20 Ing.Buero R.Tschaggelar -

formatting link
& commercial newsgroups -
formatting link

Reply to
Rene Tschaggelar

In article , Winfield Hill wrote: [...]

IIRC, the nominal impedance of a geophone is 600 Ohms. This impedance normally appears as a resistance and not the inductance you would normally expect from a coil.

--
--
kensmith@rahul.net   forging knowledge
Reply to
Ken Smith

The name "Geoformer" comes to mind as a transformer specifically intended for this application.

--
--
kensmith@rahul.net   forging knowledge
Reply to
Ken Smith

wow! thanks for the prompt (and numerous) replies - especially to Win - i can't believe that - i was reading your book today trying to figure out how to make this thing! wow!!!!! great book btw.

ok, a few points: the sensor will be at 0.3 degrees kelvin, so thermal noise is less of a problem, and it has a resistance of 250 ohms. this means we're getting a thermal spectral noise density of 0.06 nV /root Hz.

Also, since i'll be plugging the output into a spectrum analyser, the fact that the 0.1Hz - 10Hz band doesn't contain much bandwidth (and therefore little noise power) doesn't help, because the component of my signal at those frequencies won't contain much power either.

i'm not sure about what an "auto zeroing amplifier" is but it sounds like it might just eradicate everything below a certain frequency, which isn't what i want, my signal is at 1 - 2 Hz! is this the case?

The LLT 1128 looks good - it specifies 35 nV of noise from 0.1Hz - 10 Hz, so would i be correct in thinking that the spectral density will be approximately 35/root10 = 10 nV/rootHz? this looks good.

the SSM2019 looks nice at $3 - i may as well just buy it, but it only specifies the noise at 1kHz.

thanks guys, i'll look at the chips and maybe start ordering. am i correct in thinking that these chips will need putting on a PCB and a few resistors soldering on, or they contain all the components already?

cheers

Roko Mijic

Reply to
Roko

some good sugestions so far, Also you need to consider 1/f noise wich wil potentialy kill your noise budget at about 1hz,

1/f noise starts below a certain frequency wich may be quite high typically 1khz +

its hard to get an op amp below 0.1 uv pp noise in the range .1hz-10hz many low noise amps with 1nv/rt hz at 10khz have total peak noise of 10uv pp in the .1 10 hz range

One way to reduce 1/f noise is with an auto zeroing amplifier (also caled zero drift), this cancels out the voltage offest and also the 1/f nosie wich is equivalent to a quickly drifting offset voltage, parametric amplifiers maybe the ultimate but they dont seem to be used much anymore.

Also you need to consider thermal drift wich will easily be in your frequency range, thermoelectric effcets need to be considered, so thermal mass and windproofing needed, best to put the pre amp as close as posible to the source.

Ive been looking at similar problem of 1hz-10hz sub uv signals although I have very high src impedance I am using the ad8552 for high src impedance this has low noise (1.6uv pp) 0-10hz and extremly low curent noise 2fa/rthz at 10hz, and low offset voltage 1uv

however there are better amplifiers for lower src impedances, the ad745 has lower voltage noise of .38uv pp, but at 1hz the current noise is 100fa/rthz

ad797 has 50nv pp @.1-10hz but unspecified curent noise in this range, although it specifies src impdeance of 2k for min noise at unspecified freq.

the lt1028 as sugested is best for source impedances of

Reply to
colin

On Mon, 18 Jul 2005 15:32:46 GMT, colin wrote in Msg.

Linear's LT1051 is spec'd at 1.5uVpp 0-10Hz and .4uVpp 0-1Hz. From there on it's a pretty steady 70nV/rtHz up to 3kHz.

robert

Reply to
Robert Latest

Your requirements aren't much different to those of a high spec microphone preamplifier used in pro-audio equipment. Indeed, the bandwidth you need is less.

Check out the specs of the SSM2019.

formatting link

Graham

Reply to
Pooh Bear

Be very careful reading the low-noise specification. Many devices become noisy at lower frequencies and the specifications are often written to conceal this fact.

Check that your amplifier/amplifying device is low noise over the whole of the frequency range you are using. If it doesn't say so in the spec sheet, it probably isn't so.

--
~ Adrian Tuddenham ~
(Remove the ".invalid"s and add ".co.uk" to reply)
 Click to see the full signature
Reply to
Adrian Tuddenham

250 ohms makes things easier, low noise voltage is usualy mutualy exclusive with low noise curent.

no the auto zero doesnt zero the signal, just cancels out the errors in the op amp, otherwise the op amp behaves as ay normal op amp from dc to its gbw. for low impedance you can get just as good results with ordinary type.

the spectral density of the noise voltage rises below its corner with falling frequency @ 1/f the curent noise also rises @ 1/f below its corner frequency but the corner frequency is often higher. the 1/f noise only goes down to some lower frequency so noise density isnt very useful at this frequency range its best to look at the noise graphs on the spec sheets very carefuly, or the quoted dc to 10hz total noise. some noise spces have a very alarmingly steep rise at the bopom of their graph. The reference to pop-corn noise in AoE is amusingly apropriate I gues.

that looks good, it is probably specified at its lowest noise frequency, so below 1khz it will probably rise @ 1/f? dont forget although the current noise looks low too it will rise faster with falling frequency, even with

250 ohms it might become dominant at 0.1hz. if you want to be sure its best to use an ic that has a specified noise performance at the frequecny you are using. I think Win was interested in something like that for his programable gain amp.

just a few resistors to set the gain (1 if your lucky), coupling capacitors, bias resistors, couple of voltage regulators and some good supply decoupling capacitors, and a filter network to filter out unwanted noise, and possibly a second stage of amplification too, or combination of filter/amplification. there are some good examples in the many spec sheets/app notes. in fact i wouldnt be surprised if there is an example for a geophoneamplifier somewhere.

Colin =^.^=

Reply to
colin

On the data sheet the noise is shown down to 1Hz IIRC. It's about 3nV/rtHz at that point.

It wasn't actually designed for LF use but is still pretty decent.

They'll need a circuit designed around them. As you say, a few Rs mainly but don't forget power supply decoupling etc... Talking of which, do you have a bipolar DC supply for this amplifier ?

Graham

Reply to
Pooh Bear

There are also LN2 cooled amplifiers, at least in the area of nuclear physics. Cooling the first stage can greatly enhance your SNR.

Bye

--
Uwe Bonnes                bon@elektron.ikp.physik.tu-darmstadt.de

Institut fuer Kernphysik  Schlossgartenstrasse 9  64289 Darmstadt
 Click to see the full signature
Reply to
Uwe Bonnes

Indeed, with a switcher, everything is lost.

Rene

--
Ing.Buero R.Tschaggelar - http://www.ibrtses.com
& commercial newsgroups - http://www.talkto.net
Reply to
Rene Tschaggelar

you

Not with a SA for analyzing the output signal.

--
Thanks,
Fred.
Reply to
Fred Bartoli

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.