LM358 with Gain 30,000+ and single supply

Thread Starter

drkblog

Joined Oct 4, 2012
109
I need to amplify a very low signal (around 33µVp-p to 1Vp-p) and I have LM358 at hand. I already tried a couple of circuits and I've got a 4500 gain using a single supply with two stages. Using a circuit like the one I'm attaching, but with different values.
The circuit in the image is a variation from the one I've already tested. I have decoupled DC component and used voltage dividers to adjust the DC at the input of every stage. I need you guys to tell me if you think this will work.

I know the final output seems to be saturated but at first that won't be a problem.

I have tried using a dual (or split) supply so I won't have to deal with the DC offset in every stage. But I have another problem there (I will open another post for that). But basically I can't fix the virtual ground and the whole thing goes very unstable.
 

Thread Starter

drkblog

Joined Oct 4, 2012
109
This question is related to another post I just created here: http://forum.allaboutcircuits.com/showthread.php?t=75359

I need to create a very high gain circuit using LM358. So I tried using a split supply ±5V but the circuit was very unstable. And I couldn't find a good sample circuit to copy.

Which circuit would you recommend for high gain amplification using LM358 and split supply?
 

ScottWang

Joined Aug 23, 2012
7,397
You may try LF356 or LF357.
But the first thing that you have to reduce the gain for each stage.
You can seperate the total gain to 3 or 4 stages.
Exp:
Total gain : 1V/33uV=30303
Seperate to 3 stages :
1. 20 x 30.3 x 50
2. 12.12 x 50 x 50
3. 100 x 100 x 3.03

Seperate to 4 stages :
1. 10 x 10 x 10 x 30.3
2. 10 x 20 x 20 x 7.575

The first stage set to ≦ 20 is better.

LF356,LF357
http://www.datasheetcatalog.org/datasheet/nationalsemiconductor/LF356.pdf
 
Last edited:

bountyhunter

Joined Sep 7, 2009
2,512
If your input signal is 33uV (read other thread) an LF 356 is not low noise enough for you. In the stone age the NE5532 was a good super low noise audio pre amp gain block. There are probably lots of others. The LM358 is a terrible choice.

http://www.ti.com/lit/ds/symlink/ne5532.pdf

Understand your first gain stage has to be optimized for low noise and high input impedance then add at least one following stage for gain.

I think the total gain you need is about 88 dB so two stages should be enough.
 
Last edited:

bertus

Joined Apr 5, 2008
22,270
Last edited:

Thread Starter

drkblog

Joined Oct 4, 2012
109
Thank you all for your help! I just came back to electronics after fifteen years (or so) and there is a lot I forgot (it seems). Like pre-amplification for low signal. :)

I used LM358 because I saw a sample circuit for a sensor like the one I have to amplify. But now I need to extend the detection range and the documentation says I will need a high quality amplifier above 70dB. I think I will have to go near 85dB because I'm pushing it to the limit.

I've been reading the LF356 datasheet. Which parameter should I compare between LM358 and LF356 (or any other op amp) to know it is better for low signal? I mean, I know now LF356 is better because you tell me. But what if I have to decide by myself?

I'll buy a couple of these on Tuesday (tomorrow is holiday here). My provider has listed three options, one with a very different price. And has the OP177 and NE5532 also:

LF356N AMP.OP.JFET BANDA ANCHA 5 MHZ USD 0.62 FAIRCHILD
LF356H NAT AMP.OP.JFET BANDA ANCHA 5 MHZ USD 3.00 NSC
LF356M JFET INPUT OPERATIONAL AMPLIFIERS USD 0.51 FAIRCHILD

OP177GP AMP.OP. DE ULTRA PRESICION USD 1.80 ANALOG DEV

NE5532AP OPERACIONAL DUAL +/-22V 10MHOM USD 0.54 TEXAS INST

For the prototype I can buy one of each. But for the final version I have to choose the cheaper as long as it works. Any suggestion?
 

bountyhunter

Joined Sep 7, 2009
2,512
As I said before, the noise figure on an LF356 is too high if you are amplifying a 33uV signal. The NE5532 is a lot better and here may be newer "premium" audio pre amp devices which are a shade lower noise than a 5532.

If you look at the noise figure data or typical noise curves for the LF356, you will see what I mean. The equivalent noise for the LF356 is about 30 nV/root-Hz at a frequency of 50 Hz. The 5532 is about 1/4 as much noise at that frequency.

A 33uV signal is extremely low, the gain blocks noise must be low enough to not add significant noise.

You might want to ask Audioguru about a low noise amp, he knows a lot about it.
 
Last edited:

crutschow

Joined Mar 14, 2008
34,283
For a low noise signal, you should also limit the bandwidth with a filter (two or more poles preferred) to no more than you need for the highest signal frequency. That will minimize the noise bandwidth (thermal noise is proportional to the square-root of the amplifier bandwidth).

For a low frequencies you also should look at the 1/f noise spec. A low noise amp, such as the OP-07 or OP-177 which have low thermal and 1/f noise, should work well for your first stage. If the first stage gain is sufficient, you don't need such low noise amps for subsequent stages but you could, to keep all amps the same type.
 

Thread Starter

drkblog

Joined Oct 4, 2012
109
Ok, I'll buy the NE5532 and the OP177 and test them in the lab. If both of them work I will use the cheaper for the final product. One thing I forgot to mention is the bandwidth I pretend to cover which is 100 Hz to 2600 Hz. I could even reduce it to: 220 to 2600 Hz if I stick to a specific sensor.

Which kind of filter should I choose?
And where should I put it (before the first amplifier, after it, at end)?

I've been reading some of the documentation you guys referred and its great! But I think I will keep asking questions here in the meantime. Because I have a lot of work and I need this project working ASAP.

The simple supply configuration for AC amplifiers seems to be the most used. Why is that? Specially with high gain, the DC level created in the input gets multiplied reaching the maximum voltage of the output. So, in theory at least, using split supply should make things a lot better. Anyhow, when I tried that way, I've got a very unstable circuit. But I don't know if that is a characteristic of the circuit, or is something I did wrong...
 

bountyhunter

Joined Sep 7, 2009
2,512
Ok, I'll buy the NE5532 and the OP177 and test them in the lab. If both of them work I will use the cheaper for the final product. One thing I forgot to mention is the bandwidth I pretend to cover which is 100 Hz to 2600 Hz. I could even reduce it to: 220 to 2600 Hz if I stick to a specific sensor.

Which kind of filter should I choose?
You could just add a parallel cap across the gain feedback resistor to set the upper -3dB point at 2600 Hz. That will give you a -20 dB/decade gain rolloff. You can set the lower -3 dB point with a cap looking into the input impedance of the first stage.

Most of the noise content is typically down at the lower frequency caused by the 1/f corner. Above 3k the noise content is pretty low.
 
Last edited:

Thread Starter

drkblog

Joined Oct 4, 2012
109
I've been reading the applications notes for the OP177 and it doesn't seem to be used for AC amplifiers commonly. So I will start with the NE5532. Here is a sample circuit for 19.3 gain with two filters like bountyhunter suggested. I calculated the low frequency for -3dB at 220Hz and the high frequency for -3dB at 2.6kHz using the formula C = 1 / (2 π R f)


Simulation works with -3dB of the theoretical gain at both limits. But using a central frequency like 1200Hz I still don't reach the maximum gain of 19.3 that I get when no filtering capacitors are there. Any clues?

What is the criteria for choosing R1 in this circuit? Should I take it into account for the Q of the RC circuit of R1 and C1?

BTW: I built the dual power circuit with the LM358 and it worked now. In the first attempt I missed the resistor between positive input and ground (R1 in this circuit).
 

ramancini8

Joined Jul 18, 2012
473
I was involved with the TI app notes mentioned here, and I don't think they will solve yor problem. You have two problems that have to be solved with the same circuit, so this problem is complicated.
As mentioned your problem requires a low noise first stage or the amplified noise will swamp out your signal. I don't believe any of your data because I suspect that noise infested your design. Where to get a low noise preamp; scour TI's web to find an app note written by Tom Kugelstadt a few years ago and you will learn the basics. If you need more info call Tom at TI (lost his number but info will have it) and tell him that Ron Mancini promised you that he would help you. Please do your homework before you call.
Next ask Tom for his recommendation for filters; he wrote the filter section of Op Amps for Everyone. Maybe you can include low noise, filtering and significent gain in one stage---maybe not. Now be tricky; look at the open loop gain plots for op amps and realize that the gains starts falling off at 20dB/decade at very low frequencies. This won't be much of a problem for 50Hz, but it must be considered. To achieve a cloosed loop gain of 10 @ 1% accuracy the open loop gain at the frequency of interest must be at least 60dB; I think the design you presented fails this test because the LM358 has very low gain-bandwidth (see figure 12 of the TL071 data sheet where the open loop gain curve is shown). If you are interested in the error analysis of decreasing loop gain see an article in the TI journal that I wrote last decade.
This may sound like a lot of detail that you may not want to sift through, but learning never hurt anybody.
 

Thread Starter

drkblog

Joined Oct 4, 2012
109
I have two goals, the first one is making this work. The second one is learning as much as I can about this. Because it's the first stage of a large project. So every detail counts.

I know LM358 won't work but is the one I have at hand until tomorrow. It's worth to notice that this circuit is only a simulation. My real circuit has the LM358, and I know for sure (right now) I have a reading for a 360µV signal (with a lot of noise). But yet I know this signal is a valid reading from the sensor. Here is it, this is the reading after a single stage of gain 19.3 (times) with no filter. The circuit is mounted over a breadboard. Yes, I know it looks pretty ugly:

 

Audioguru

Joined Dec 20, 2007
11,248
You should not use a breadboard for high gain low noise circuits because the tangled mess of wires are antennas that pickup mains hum and other interference.

Your 'scope photo shows 100Hz which are the pulses every 10mS (from the power supply?). Nearby fluorescent lights also produce 100Hz interference that might be radiated as electromagnetic pulses or the light pulses are picked up by clear cased diodes.

Your 220Hz highpass filter and your 1200Hz lowpass filter are very simple each with only one R and one C. The simple 220Hz highpass filter reduces frequencies as high as 2200Hz the the simple 1200Hz lowpass filter reduces frequencies as low as 120Hz.
Then the middle frequencies around 530Hz are attenuated.
You need filters with more RC stages (an active Butterworth Sallen and Key filter type) for sharper cutoffs.
 

bountyhunter

Joined Sep 7, 2009
2,512
What is the criteria for choosing R1 in this circuit?
High as possible to keep from reducing input impedance too much (which loads down the source driving the input), but must be low enough so that the input bias current doesn't cause a significant voltage drop across the resistor.

Should I take it into account for the Q of the RC circuit of R1 and C1?
You have to take the input impedance into account as to loading on the source signal device. That's the tradeoff: low noise amps don't like having a high source impedance but a low impedance will load the thing driving the input.
 

Thread Starter

drkblog

Joined Oct 4, 2012
109
You should not use a breadboard for high gain low noise circuits because the tangled mess of wires are antennas that pickup mains hum and other interference.

Your 'scope photo shows 100Hz which are the pulses every 10mS (from the power supply?).
Yes, it's twice 50Hz (line frequency here).

Your 220Hz highpass filter and your 1200Hz lowpass filter are very simple each with only one R and one C. The simple 220Hz highpass filter reduces frequencies as high as 2200Hz the the simple 1200Hz lowpass filter reduces frequencies as low as 120Hz.
Then the middle frequencies around 530Hz are attenuated.
You need filters with more RC stages (an active Butterworth Sallen and Key filter type) for sharper cutoffs.
I suspected that both, the breadboard and the filter problems. So I'll change the filter later. And I'll build the circuit over an experimental board, as I don't have the means for building a properly designed printer circuit right now.
 

Thread Starter

drkblog

Joined Oct 4, 2012
109
From the manufacturer of theses sensors I know for sure I have to build a high gain, good quality, amplifier. As far as I learned here, I need to find a good op amp (I'll buy NE5532 and TL071 today) and build a filtered amplifier over an experimental board (at least). If anyone can think of another IC for low noise amp I'm open to suggestions. Thank you again guys...
 

Audioguru

Joined Dec 20, 2007
11,248
A Texas Instruments OPA134 single, OPA2134 dual and OPA4124 quad opamp has a noise level almost as low as the NE5532 dual opamp. The noise level of the TL07x is 2 times to 3 times higher.
 

Thread Starter

drkblog

Joined Oct 4, 2012
109
Thank you audioguru! You recommended an active Butterworth Sallen and Key filter. This will require additional opamps in my circuit, isn't it? I guess I should use the same low noise option as for the amplifier. So let's say I buy an OPA2134, I could use one for the filter and the other for the first stage. I should put the filter before the first amplifier (I'm guessing here)
 

Audioguru

Joined Dec 20, 2007
11,248
I think you should amplify the tiny signal before filtering it so that you filter out some of the noise of the amplifying opamps. An OPA2134 will do it all.
 
Top