Yannick said:
I have done some noise calculations for my transimpedance amplifier to
detect signals up to 50Mhz , i have still some questions , this is
what i have found already :
for example i want to detect 2na current through my photodiode (C =
5pf at 20v reverse bias). Soo the noise in the feedback resistance
should be lower then 2nA. In² = 4KTdf/R = > df = 50Mhz, In = 5nA ,K =
1.38*10^-23 and T = 290°K
soo i get R >= 200Kohm to have less then 2na current feedback noise.
But then you also have equivalent voltage noise at the input of the
amplifier wich is 6nv/sqrt Hz for the OPA655 => this noise and the
Total input capacitance will give noise integration : vn² = KT/Ctot
with Ctot = Cd+Cf+Cin = 5pf+0.2pf+1pf = 6.2pf.
veq² = 645pV => veq = 25.4µv then this noise is increased due opamp
voltage gain R2/R1 with R2 = 250K and R1 = opamp input impedance
parrallelel with photodiode impedance wich will be about 500ohm for
50Mhz and 6.2pf total capacitance wich will give R2/R1 = 250K/500 =
500 voltage gain, soo Vnoise = 12.7mv at the output for this noise
contributor . is this correct of am i doing it all wrong?
i have also found something to reject daylight with succes, i put a
coil with a resistor from 20Volt power supply to my photodiode to
reverse bias it, then i use a 1nf capacitance between the photodiode
and inverting input of my OPA655. is this a good way of rejecting
daylight? i understand this resistor has to be high because of current
noise wich will be added to the total noise like the feedback resistor
does.
Further i was thinking of just using discrete components like a simple
HF transistor in place of the OPA655, then using a current mirror to
drive the basis current to get a low Rpi (small signal resistance)
(rpi = dDvbe/Dib)
so that the Rload of the photodiode is small, i didn't see at the
noise figure of the Transistor yet , anyone tried this ?
thanks for the help soo far,
Yannick
It seems amplifier input noise voltage combined with detector capacitance is
the limiting factor, wich is why i went for very low noise discrete stage.
its a shame you cant go for tuned input stage, however the key to getting
rid of the noise is to use a narrow pas filter, or if you converting it to a
dc voltage you can just use a low pass filter with low cut off frequency
wich should reduce your noise greatly.
low noise discrete mosfets have about 1db noise wich i think equates to
about 1nv/hz-1/2. however i found one of the largest problems to be EMI
pickup.
I roughly calulated that at 3 meters with 5mw of laser power reaching the
target and 10% being reflected back in a 90' cone and a 1mm receiving area ,
that 50pw would be received by the photodiode wich at about 1A/W = 50pa.
with a 15 mm lense that increases to 12.5 nw, and an APD shld get at least
1.25 ua or much more, but my optical system is a bit of a lash up (lenses
glued to the pcb etc) and so i think i am seeing a lot less than this as its
very critical to focus it precisly, and it doesnt stay focused. i need to
make a machined housing to mount the lenses, detector and act as heatsink
for the laser etc.. but this probably isnt going to hapen.
AC coupling the signal from the photodiode is one way to reduce daylight
signal, or using an inductor in the feedback path, or as you sugest to bias
it, however dont forget as Win reminded me that inductors have resistance at
ac depending on their Q wich will add to noise, as well of course as well as
any paralel capacitance.
you could look at this circuit for ideas ..
http://www.imagineeringezine.com/PDF-FILES/fetamp0.pdf i think your idea of
using a curent mirror is a good idea, and would be easy in this circuit, you
might need to use a negative bias voltage for the mirror or bias the input
transistor a bit higher. you could easily arange the curent mirror so at the
frequencies you want to amplify it only fed back a smal fraction, thus
avoiding the need for the inductor (wich by the way are horible at picking
up EMI). you can get narow band light filters to filter out most of the
daylight and just alow your laser light through, but i dont think this wld
be necessary.
however stil the best way would be to mix it opto-electronicaly so it is
converted to a lower frequency before it is converted to an elctrical signal
wich is what im working on trying to do now and its looking quite good.
i gave up trying to get proper 455 khz ceramic filters or using resonators
as filters, and i found that the 20khz filter i made out of 4 inductors was
horible becuase its phase varied with signal strenght, i used toroids to
avoid interfernce but these are ungaped and so a bit variable and also means
the hysteresis of the feritte cuases a non linear efect wich i didnt
consider. Also ive gone for two IF now, wich makes it more complicated but
was necesary. it means i am not so limited to what high frequency i could
use. also the 2nd stage is now driven at constant amplitude so linearity
isnt an isue, obviously if i was to redisgn i would use an inductorless
filter circuit.
Colin =^.^=