Is there any opamp better than Linear Technology's LT1028 regarding low frequency noise? i.e. voltage noise about 1 nV/sqrt(Hz) and current noise about 1 pA/sqrt(Hz), and 1/f corners below 10 Hz if possible.
It doesn't matter if the GB product is as low as 1 MHz.
Regards,
Jean-Pierre Coulon (here "cacas.pam" is what others call "nospam")
AD8599 is damn close and has a GBP of 10 MHz. Might be a useful alternative candidate.
Graham
-- due to the hugely increased level of spam please make the obvious adjustment to my email address
Analog Devices do offer a bunch of low noise ampliers that compete - I've not used the AD 797 myself, but I suggested its use once and it seemed to work pretty well.
If you want to go nuts you can trying parallelling discrete transistors or transistor arrays - the Analog Devices quad MAT04 would do better than 1nV per root Hz if you paralleled a couple of them.
And there are always big JFETS
-- Bill Sloman, Nijmegen
Unfortunately the current noise 1/f corner is high.. current noise at
10Hz is 6pA/Hz^1/2I don't know why there are no monolithic JFET-input amplifiers optimized for low voltage noise (of course they don't have much current noise). Not many hybrids out there even.
Have a look at the ADA4898-1. 1.2 nV @ 10 Hz. A bit more current noise, 2.4 pA, but a very nice part. Only two gain stages, vs. 3 for the LT1028 (as I think miso pointed out), so its open loop gain isn't as high, but it's still 100 dB. The noise plots only go out to 100 kHz, so it's possible there's buried treasure out there someplace (as with the LT1028), but if so, it isn't below 1 MHz.
Cheers
Phil Hobbs
If your source impedance is low, you parallel a bunch of the 4898's or LT1028's; with the noise falling as the square root of N.
The BF862 jfet runs about 0.8 nv/rthz, with esentially no current noise. If you don't mind the capacitance, you can parallel a lot of them... they're cheap, too.
Distributed amplifiers are interesting here.
John
Hmmmm. A 32 Ohm resistor generates 1nV/root Hz at room temp. I know curiosity killed the cat but I have to ask. What are you trying to amplify that has a source impedance less than 32 Ohms and a frequency response where 10 Hz matters?
There was application note from Analog Devices about using an opamp with the external input stage made of four SSM-2210 in parallel. IIRC it had
0.5nV/1pA performance with 1/F corner at 1 Hz.Vladimir Vassilevsky DSP and Mixed Signal Design Consultant
JK17PWGBDR a écrit :
1nV/rtHz is about 60R, not 32R. Some other values good to remember are: 50R is 0.9nV/rtHz 1K is 4nV/rtHz 1M is 128nV/rtHzBut 32R isn't that far, at 0.73nV/rtHz. (obviously about a sqrt(2) factor below 1nV)
w
ise
Right. Years of knowing 50 ohms is just below
1 nv and I say 32 is. Must be getting old. Still would like to know what he is trying to measure that makes nV/rootHz important in the 10 Hz range. At that level and frequency he is going to have serious problems with thermal noise, electromechanical noise and stray rectification of EMI. Any one of those will be noisier than the op amp unless he is extremely careful with packaging and layout.
Once you reach the lower limits like this you can decrease your noise further by paralleling the best amp you can get.
Interestingly, AD have a patent on this for use at the substrate level:
Dave.
1K ohm at 4.2K only has ~0.5nV/Hz^1/2 Johnson-Nyquist noise.
Best regards, Spehro Pefhany
I am measuring current fluctuations from a photodiode from 1Hz to 10kHz.
I put a 50 Ohms in transimpedance mode, so the opamp noise voltage quadratically adds to the .9 nV/sqrt(Hz) of the 50 ohms.
Later on I will put a C-R high pass filter below 1 Hz, and a non-inverting stage with gain=10 or so. At present the AD797 seems the right choice, but I'll make noise calculation with my two stages.
Jean-Pierre Coulon
Just ranDUMBly paralleling devices will NOT give much help, if any. Theys gottsta bee matched! Oh, BTW, every DOUBLING of (matched) units gives another 3dB so it gets a bit costly after 8..
...meaning not much of a market..but OK for DIY?
Okay, but it would be nice (as in I'd be happy to pay for it) if it was available as a fairly small hybrid (fully spec'd and tested, of course, with guaranteed voltage noise and input capacitance).
Best regards, Spehro Pefhany
What is the capacitance of the photocell? Here's a low noise JFET:
g
but
Cool. I have a few suggestions for you. At the noise levels you are talking about you will need to take extraordinary precautions to prevent external noise ruining your measurements.
1) Heat sink the photodiode. Even a very small change in temp can affect the leakage current. Even somebody walking by can make enough of a draft to affect the reading. 2) Place the photodiode and first amplifier in a heavy metal box and keep it at a controlled temperature. If you are going to allow light to enter the box onto the photodiode make sure no light can get past into the rest of the box. There are several reasons for this. a) any temperature change between readings will cause the leakage currents in the photodiode to change and change your readings. It might be interesting to vary the temp and see how much the photodiode leakage is affecting your readings. b) any temp change in the box can cause the thermcouple effects. Run the box for at least an hour before taking readings to allow thermal gradients to stabilize. c) any light leaking into the box can cause photoelectric effects on any semiconductor in the box. Most semiconductor packages will leak a small amount of infrared into the package. 3) If the photodiode is not mounted on the circuit board use RTV on the connecting wires to hold them down to box and circuit board. If the can vibrate they will generate microphonic noise. 4) Take extreme measures to RF filter any wires going into or out of the box. The person two rooms down talking on his cellphone can absolutely destroy sensitive readings. 5) Use batteries for the first amplifier power. Do not rely on a regulated power supply as these can be nasty for low frequency differential noise between the + and - rails.Good luck and be prepared to see some strange things.
Also an alternative could be to run the photodiode in photovoltaic mode (zero bias). Then there will be no leakage current since there is no voltage across it.
This does increase the PD capacitance, may not be a problem at the low frequencies being mentioned here?
z.
ting
e, but
I believe that is what he is doing. The trouble is that the photodiode does have an appreciable internal impedance which will generate noise and leakage currents. How much noise depends on the diode he is using.
I should have put in something like this earlier.
Have something to add? Share your thoughts — no account required.
Ask the community — no account required