Simple 60dB discrete low noise amplifier (lna)

Douglas Self give the improvement with doubling of input devices as -1.8 dB.

I recall that from a misleading section where he keeps the total quiescent current in the paralleled transistors constant. This of course just reduces rbb' but not shot noise. To reduce voltage noise by 3 dB with doubling you'll also need to double the total collector current (i.e. keep it constant per transistor).

Samuel
 
Is that the $43 pre-amp? I agree on having the instrument compute the V/rt-Hz. It's tedious to compute the equivalent noise BW for each instrument state.

I think it's $4.30 -- move the decimal point.

The analyzer is a new toy. It will have a new home in September when classes start up. It averages very nicely, can take vibration measurements etc. (Did I tell you folks about the black-robe who mirthfully told us why we had to know about the motion of a spring?). I can get it to "do" V^2/Hz, all sorts of correlation analysis, but I am back in the ancestral homeland enjoying the long summer sundowns, and fresh caught walleye etc. at present.
 
Yes, he uses quite a low idle current and argues that the noise drop has a wide minimum.
Syn08 on the other hand with high quiescent current has got the noise down close to the theotetical minimum, voltage noise, that is.

WRT noise versus quiescent current, there are a few generalisations here but
data sheets in this case are your friend. For example using multiple 2sk170's,
not much point running more than around 2mA / device noise wise.

Same goes for BJT's, there is usually a sweet spot. Looking at MAT02 data
sheet, 2~3mA is the spot.

WRT this project in general, there's a lot to be said for making a fully
balanced IP design. It will allow much more flexibility for grounding and
getting good consistent real life measured results due to the CMRR, albeit
suffering the 3db increase in noise.

T
 
WRT noise versus quiescent current, there are a few generalisations here but
data sheets in this case are your friend. For example using multiple 2sk170's,
not much point running more than around 2mA / device noise wise.

Terry, but to get the highest gain possible in the first stage, would running at about 2mA per device not require larger drain and source resistors, or in the case of cascode, just a larger source resistors? And would that not add noise?

WRT this project in general, there's a lot to be said for making a fully balanced IP design. It will allow much more flexibility for grounding and getting good consistent real life measured results due to the CMRR, albeit suffering the 3db increase in noise.
T

This is a thread for me to explore ideas, so any suggestions for fully balanced input designs would be great.
 
I'm still fooling with simulations of the previous circuit and am having a heck of a time getting the DC operating point where I think it should be. I think it's fundamentally flawed in that there's no way the output can put enough signal through the feedback to the input to correct for device variations. OTOH, I ain't done yet.🙂

Conrad, are you referring to the circuit in post # 195?

http://www.diyaudio.com/forums/soli...e-low-noise-amplifier-lna-20.html#post2196633

Samuel has a properly designed circuit to which I posted a link in post #195. So does syn08.

Whatever I try here is nothing more than, as you say, fooling around. 🙂
 
Well, that's two of us! Who's to say why a certain approach appeals and another doesn't. Those are decent circuits, but I'm looking at others. Unfortunately my library and article collection has gotten so large that I can never find the original work for my vague recollections. I think there are more possibilities for a high gain ultra low noise amp.

CH
 
I think it's $4.30 -- move the decimal point.

The analyzer is a new toy. It will have a new home in September when classes start up. It averages very nicely, can take vibration measurements etc. (Did I tell you folks about the black-robe who mirthfully told us why we had to know about the motion of a spring?). I can get it to "do" V^2/Hz, all sorts of correlation analysis, but I am back in the ancestral homeland enjoying the long summer sundowns, and fresh caught walleye etc. at present.

Found it, Dennis with two n's not one. Realisticly you could get this to work with any number of parts. Topologically it's just half of the classic in-amp connection.
 
Quick question about measuring noise. The only device I have that can measure the noise spectral density is the 3585A analyzer. When I short its input with the marker at 1kHz it reports a noise level of 10nV/rtHz on average (3Hz BW, 2kHz span). I'm thinking that this is the noise floor of the analyzer. Let's say I connect to it a noise source and the device reports now 14nV/rtHz. Should I now subtract the noise floor of the analyzer from this figure, to obtain the noise density of the source? In this case it would be sqrt(14^2-10^2) = 9.798nV/rtHz. Could someone with more experience confirm or deny this please?
 
Quick question about measuring noise. The only device I have that can measure the noise spectral density is the 3585A analyzer. When I short its input with the marker at 1kHz it reports a noise level of 10nV/rtHz on average (3Hz BW, 2kHz span). I'm thinking that this is the noise floor of the analyzer. Let's say I connect to it a noise source and the device reports now 14nV/rtHz. Should I now subtract the noise floor of the analyzer from this figure, to obtain the noise density of the source? In this case it would be sqrt(14^2-10^2) = 9.798nV/rtHz. Could someone with more experience confirm or deny this please?

Yes, you "un"-rss it. I would average a lot before taking the answer too literally. Don't forget you can't take the three decimal places seriously either. I would say 10nV and 10nV rss'd is 14 that's about all the resolution you get at this level.
 
Last edited:
Thanks guys!

OT: Conrad, I see you like old cameras. I got a couple of IIIc for parts, you can have one of them. The missing parts are the meter, and the film winding ratchet (breaks often on those cameras). I also have a minty IIIc that I use occasionally. 🙂
 
OT- yes, I've got an ok IIIc too. The meter door hinge is broken and no good way to fix it, but the images are 1st rate. I've also got a parts camera- maybe we could make a single good one with yours? No photo of the whole collection, but I've also got a nice miniature graphic, various TLRs and an Ikonta. The darkroom is still set up, but I haven't done any wet processing in a few years now.

On this low noise amp, I'm sure you saw the LT app note where they parallel several LT1028s. It also occurs to me that paralleling my favorite opamp, the National LME49710 might have some merit. They're already very quiet and I'm pretty sure they're less expensive so one could use more of them to advantage.

That flashing smilie thing to the right is really annoying!

CH
 
As Scott mentioned averaging -- in the "Why didn't I think of that?" department -- if you average the noise, and multiply by 6.6 you get the peak-to-peak noise within 6.6 standard deviations. The o'scope will probably shows you some of the outliers way out on the tails.

Speaking of stats -- Cameron Iron Works has 2,500 blowout preventers in operation and only one failed in a number of years...

Edit -- is it a Retina iiiC, or a Leica?
 
Kodak Retinas.

Since this amp is AC coupled anyway, I've been thinking about some designs that incorporate an interstage transformer. After all, a reactance (in theory) generates no noise- voltage gain for free. So, SK170s for the front end, a transformer impedance transformation, then a discrete amplifier circuit after that. Not sure if there's an advantage or not.

CH
 
OK, I'm back.

Conrad, with a transformer would it be not difficult to get good frequency response from 10Hz to past 100kHz?

Meanwhile I've tested syn08's circuit but with 2sk170, then I built Samuel's circuit exactly as per his schematic, not modified. Then I tried to modify it by paralleling more 2sk170 in the first stage. I had trouble with both circuits with oscillations, sometimes when shorting the input, other times all the time.

I ended up with this circuit which is nothing but a trimmed (you can read dumbed down) version of syn08's circuit. Removed all the frills; it proved to be stable for various tests, including 3, 5, or 9 paralleled 2sk170BL. The version with 3 jfets measured about 0.7nV/rtHz at 1kHz, with 5 jfets about 0.62nV/rtHz, and with 9 jfets about 0.5nV/rtHz. Perhaps it's laughable and bad engineering, but I'm OK with this version and with the result.

I used two 12V sealed lead acid batteries in series. The build is "dead bug style" in a cookie tin box. If the batteries were not freshly charged the noise went right up; which confused the heck out of me at first, because I didn't notice it and couldn't explain why the same circuit that measured low noise before now had double the noise. The circuit is very sensitive and just removing the lid adds lots of noise. I had to move the box physically away from the oscilloscope and all other devices to get the noise to its lowest measured level.
 

Attachments

  • amp-noise-dccoupled-6aa.png
    amp-noise-dccoupled-6aa.png
    10.8 KB · Views: 357