:: many LNA design questions (1 at a time)

Hello n good day edaboard veterans,

i have many questions to ask to verify my curiosity into something i have been trying to play around in simulators, questions in other forums, etc etc, on low noise amplifier designing. (example this 1

https://www.eevblog.com/forum/metrology/characterising-low-frequency-noise/

and this most recent very interesting thread

https://www.eevblog.com/forum/metrology/ultra-low-noise-reference-2dw232-2dw233-2dw23x/100/)

in those threads, and other many PDF in the google jungle, i can understand that in the LNA, opamp is just to gain up the noise, using LPF + HPF as bandpass, then measure the amplified noise output. some have designed it as x1000 gain, some x2000, x10000, etc. i took some time to create simulations to try them, i have never made any active filters before using opamp. i want to understand more about the opamp criteria, i dont quite understand it. (here comes all the questions)

1) in intersil application note 1560. a wideband LNA, they use HA5147 opamp. it has 0.09uVpp input noise. in the application circuit, it has a gain of x101, does the output in theory carry forth the input noise as approximately 9uVpp?

2) with the above in mind, i can understand that they choose HA5147 due to its high bandwidth. however in theory, are these the main criteria to look for [1] low Vpp noise? [2] bandwidth? [3] input/output impedance matching? <-- i have no clue how to do this [3].

3) in design of the LPF and HPF, if say 0.01Hz to 10Hz, i have read some use 1st order, some use 2nd order, etc. i assume the ideal is nearly brickwall? so in practical we should design 4th order LPF/HPF? but in many discussions i have seen, many are done with up to 2nd order, i am assuming this is to simplify design yes?

4) by combining my bits and pieces of knowledge, i came up with this (simulation)






i inject 0.01Hz or 10Hz, and adjust the LPF/HPF so that the various measuring points show a -3dB drop, this give me the impression that the components are right to give a combined bandpass of 0.01-10Hz (2nd order)? is this the right way to design filter? for opamp i use OPA140 due to very low Vos/Ibias/noise, however, i assume this 1 is a very high input impedance opamp, i am not sure if i have matched the gain and input/output impedance in a right way. i seek the knowledge to know more about this, if anybody could enlighten me, thank you in advance.

using the above circuit as example, could i be right by choosing OPA140? @ 0.25uVpp noise? or due to the high gain x10000, should i forego low Ibias/Vos and go for primarily low noise density like AD4528/AD797/AD8428?
if to use AD797, due ot extremely high ibias, should the design focus then be to balance the input resistors so that the current noise will cancel each other? (im not sure if i understand rightly, but i assume is to equalize values of approx R2 and R3? i am not sure which "segment of resistance" to balance)

... and thank you all so much for reading such a long noob post

Design of LNA has nothing to do with filters in a first order. Signal filtering is done somewhere down the signal path. The LNA problem is
1. have a signal with given magnitude, occupied frequency spectrum and source impedance
2. find an optimal or at least sufficient low noise amplifier to pick up the signal

Noise related, amplifier input impedance, noise voltage and current density distribution are the relevant parameters.

Filter come into play if you want to measure noise voltage in a specific frequency band of interest, e.g. popular 0.1 - 10 Hz Vpp.

It's completely wrong to place RC filters in front of a LNA because they add noise, except for a RF block or well considered AC coupling high-pass.

Your post doesn't mention an actual LNA application (signal characteristic, source impedance etc.), respectively it's impossible to decide about an optimal amplifier.

i intend to design a LNA to find out the noise of batteries, zeners, low noise DC regulators. i think impedance under 10ohms.
your reply shed much light into my dis-array bits of thought, many thanks.
i am trying to take a peek into noise levels in the vicinity of 10 ~ 100nV, with x10000, this i hope to be easy to log as data in something like a 5.5digit DMM. which i guess ideally, it should resolve down to 1nV if the LNA is not too noisy? but what you said made more sense to me now as input RC adds noise = less signal, so i need to figure out a RC front end more than 1:10 signal noise ratio in order to resolve required "resolution" i am trying to find.

TYVM

FvM said:

Lowest noise will be achieved if the input is AC coupled with a cut-off frequency far below the actual frequency range. Or input DC coupled and high-pass roll-off in the feedback. The frequency window should be defined with active filters after preamplifier or digital filters.

Click to expand...

when you say high pass roll off in feedback, i assume is this C6 ? just for sanity check, if R3+R6 ~ 220k and C6 = 68n ~~ the high pass in this case is 10.6Hz?


when you say cut-off extends far below actual, as in the 0.01Hz needs to be around 0.001? wow that needs very big capacitor.

OP with lowest noise is LT1028 with 35 nVpp 0.1 to 10 Hz, But you need to keep total source resistance below 10 Ohm, your AC coupling filter has to be designed respectively. If you choose an amplifier with lower noise current density, e.g. said OPA140, resistor noise will still affect the overall noise. SPICE noise analysis will show you the contribution of each component to overall noise.

Lowest noise will be achieved if the input is AC coupled with a cut-off frequency far below the actual frequency range. Or input DC coupled and high-pass roll-off in the feedback. The frequency window should be defined with active filters after preamplifier or digital filters.

- - - Updated - - -

No. C6 must not be there, it causes instability of the amplifier. I meaned a series capacitor for R5, but you don't achieve the intended frequency range with feasible capacitor values.

oh ops.
what if
gnd -- 68u -- 22R --\
gnd ----22R ------/

does this work?

on the other hand, i took a peek with LT1115 and AD797. it seem to make more sense now that a lower voltage density noise is needed as the input source impedance is so low that it will be negligible to create more current noise --> is my deduction correct?

there is this which i roughly did to try and see where does the lower end noise could go as it is not on the PDF. does this make sense? esp the noise density convert to pp noise? (x sqrt of freq range? and x6.6 for peak peak?)



- - - Updated - - -

FvM said:

OP with lowest noise is LT1028 with 35 nVpp 0.1 to 10 Hz

Click to expand...

oh blind me, yes i see it now.

i have some time to look for more info, and i found this site.
http://dicks-website.eu/noisecalculator/index.html

i managed to plug in some numbers



i am not completely sure, but it seems like the OPA140 does perform better? can anyone have a look?

edit **
OPA140 values : Rs ~ 2.4M, 30C, 0.01nA, 9.99Hz. 12nV/Hz noise, 0.0008pA/Hz noise, 0.0008pA/Hz noise, R1 68k, R2 0.68R
LT1038 values : Rs ~ 0.68R, 30C, 0.01uV, 9.99Hz. 1nV/Hz noise, 1pA/Hz noise, 1pA/Hz noise, R1 68k, R2 0.68R, R3 0.8M

OP140 ~ S/N 31.5dB, noise factor ~ 0.0159
LT1038 ~ S/N 9.9dB, noise factor ~ 19++

so in fact, i should use OPA140? the Rs source in OPA140 could double as the first AC coupler. or did i miss something obvious about the LT1038?

Hello again edaboard veterans

i wish to understand, the math involved in comparing the voltage noise of these 2 opamp. (yes still learning to understand them). this time, i read this app sheet called SLYT237 from TI.

in OPA140 pdf, the voltage noise is simply given without reference to gain involved, which is assumed then to be gain =1 (@ 1 Hz, Vn ~~ 16nV). in the AD8428 pdf however, it is a different opamp package, and it specifies its voltage noise AFTER a gain of 2000. @ 1Hz Vn ~~ 3.5nV.

what should i do to either the spec of OPA140 / AD8428, so that i can compare both their noise at same GAIN? divide the AD8428? or multiply?

thanks in advance

Hi,

If find the "Design note 15" from Linear Technology very useful.
"Noise Calculations in Op Amp Circuits – Design Note 15 by Alan Rich - Linear Technology"

https://www.linear.com/docs/4201

Klaus

KlausST said:

Hi,

If find the "Design note 15" from Linear Technology very useful.
"Noise Calculations in Op Amp Circuits – Design Note 15 by Alan Rich - Linear Technology"

https://www.linear.com/docs/4201

Klaus

Click to expand...

thanks Klaus, in the case of the OPA140 where it operates in current mode, R3 is not series to produce EN2/ER2, but in parallel with the input. so in this case i assume EN2 = ER2 = parallel of source with the parallel R3? (= very low value i assume) ~~ am i right?

i created this tabulation based on 2 app notes about noise (especially the eye opener for me, calculating 1Hz noise). by shuffling different supporting resistors, at the end, we find the usable signal over noise output. it really puts some insight into opamps that is used purely for current and opamps purely used for voltage mode. and opamps used for AC, and opamps used more for DC applications.
and thank you again for all tips guys.