HELP: very low noise pre-amp, 0.01mV, 100KHz

Yes, and what's the expected signal-to-noise ratio? Assuming a gain of

300k and an optimistic noise bandwidth of 150 kHz (1-pole rolloff with fc=100 kHz), a 1 nV/sqrt(Hz) noise source becomes 1e-9*300000*sqrt(150000)=0.116 V rms.

Cheers,

Phil Hobbs

Nope, if you amplify 300000 times with 3 transistors, there will hardly be any feedback, thus very low bandwidth unlinearity and distortion. The noise will be high,(0.14Vrms, 0.85Vpeak) with that much of gain in a wideband amp.

To decide which type of transistor/opamp to choose, it is important to have some data about the sensor. Is it a piezotransducer or a dynamic mic with a voicecoil? I think the latter, because 1kHz sounds pretty low.

You probably need also a filter, which can improve the S/N ratio. A transformer can help if the sensor impedance is very low. Then I would use at least 3 stages of wideband opamps like OP37, better would be a balanced amplifier (THS4130) 110dB is quite demanding layout wise as well.

ciao Ban

You're asking for a gain of 300k at a bandwidth of 100kHz. This can only be achieved with multiple stages. To start with, what is the source impedance ?

Rene

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

Noise of an amplifier with a given gain and bandwidth depends on the source resistance, which you did not specify. Roughly speaking, a bipolar input stage will have lower noise, especially at "low" input source impedances. A MOS FET input stage in general will always give a higher noise at any source impedance. A JFET input stage in general is the best choice for lowest noise for "high" input source impedances. Use of a low noise op-amp is a good choice because the design tradeoffs have already been done - but the design should be "matched" to the source impedance you have. That usually means a bipolar front end for low impedances, and a JFET front end for high impedances. If your source has an unusually low impedance or an unusually high impedance, a transformer between the source and the amplifier may be useful to reduce the overall noise.

any chance that a logamp would be more suitable?

AD8307 ISTR, near DC to almost light bandwidth

martin

Many thanks for the replies.

The input is piezo transducer. Not 100% sure about their input impedence. My impression is they are in mega ohms range (please correct me if I am wrong). Should I use discrete JFET or op-amp with JFET input stage? Any recommended part numbers?

The transducer is used to measure mechanical parts impact sound. I am now using an 'ordinary (not low noise) op-amp amplifier' in front of a PC sound card. The impact sound signal spread from about 1 to 18 kHz.

Have checked

and apparently, the lowest noise op-amp is about 1nv/sqrt(Hz). So, if I scale down the freq to 20kHz, noise is

1e-9*300000*sqrt(20000) = 0.042 V, giving 20log(0.042/3)= 37 dB signal to noise ratio. Is that right?

Is there other company making op-amp with noise lower than

1nv/sqrt(Hz)?

Some text book suggest that since the first stage determine the overall noise perfromance. One can use a discret 'lowest-possible noise' JFET as front end and then use op-amp to get the needed gain.

Any recommendation and (JFET, MOS or bipolar) transistor part number for such an approach? What are the ball park figure of signal to noise ratio that one can expect (this is a medium cost instrument, so, can afford upper-medium price/quality chips)?

Presume 20kHz bandwidth and gain at 300,000 what type of shielding and layout may be needed? Assume a three to four stages design, would it need to make four metal partition, like those used in VHF/UHF receiver layout?

Many thanks in advance.

Impacts of hard materials on hard materials go *much* higher in frequency than that. You could probably get a huge SNR improvement by going to a faster transducer and increasing the BW. This sounds paradoxical, but the peak power in a pulse goes as the BW**2, whereas noise power goes as BW, so you win by widening the BW until the pulse stops getting narrower. Small-object collisions can easily get up into the megahertz range--and impacts of micron-size particles go up to ~ 1 GHz.

Cheers,

Phil Hobbs

Basically yes, but... Also note that those ultralownoise amps having that little noise are mainly wideband amplifiers. Artificially limiting the amplifier with an RC in the feedback for the noise doesn't work either. You have to fully use the gain bandwidth. Sad enough, the ultralow noise amps rely on low source impedance eg the LT1028 is having these

0.9nV/rtHz with a 100 Ohms. Have a look at their datasheets.

Rene

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

GHz.

I first was stuned reading that GHz story, but further thinking (speed in the 1000s m/s and micron size) finally get us in the GHz region. I wouldn't have thought.

--
Thanks,
Fred.

Yes, the typical width of the pulse is the particle diameter divided by the speed of sound in the material. It gets up into big numbers for small particles.

Cheers,

Phil Hobbs

What's the source impedance from your transducer ?

Can't advise on best method without that knowledge.

Is the input required to be differential too ?

Graham