Multiple amplifiers all acting crazy

Thread Starter

emigre459

Joined Jul 22, 2006
10
Hello all,

I apologize in advance if I have not done a thorough enough job of searching the threads for my answer. I am in the process of building a system that can amplify by six orders of magnitude. I have done this by simply using two 10^3 stages of auto-zero offset op-amps (AD857x). This is a simplfication, as my first stage is actually 4 quad op amps which I then average and amplify simultaneously in the second stage using a single op amp.

I currently have a problem where all 16 of my first stage amplifiers are giving me different outputs, even though I have the same input on all of them (currently just a simple two-resistor voltage divider that provides me with a 300 microvolt signal for testing purposes). The circuit is currently set up for an overall gain of only 10^4, since I have to worry about thermocouple problems when I get down to the microvolt scale and would thus have trouble with initial testing. In any case, I am troubled since none of the amplifiers is outputting anything like 300 millivolts: they're all outputting more along the lines of 700 or 800 microvolts, although some are actually lower than the input signal for some reason!

I'm currently having trouble getting my circuit diagram to attach, so I feel that I should mention that the circuit ends with a two-pole RC lowpass with a time constant of 0.1 seconds. Also, the circuit is setup so that (in its final form) it will use Reed Relays to switch between 10^5 and 10^6 (10^6 is the default) using a logic signal. Also, the amplifier circuits are all inverting amps with negative feedback.

One last thing should be noted: when I make any of the first stage amps into a voltage follower, it works perfectly. As soon as I try to amplify by an order of magnitude, however, things go bad.

Sorry for the long post. I need this project to graduate, so any and all suggestions would be welcome. Thanks.
 

JoeJester

Joined Apr 26, 2005
4,390
What format is your graphic?

If it's a bmp, open it in paint then save it as a gif and/or jpg and see which one is smaller.

You can decrease the color levels in paint to save it as a b&w, which should considerably reduce the size of the file.
 

beenthere

Joined Apr 20, 2004
15,819
Hi,

When you zero the input, do both stages give you zero out?. I've been building amps for many years, and prefer to sneal up on a total of 10^5 with three stages. Seeing your schematic would help.
 

Thread Starter

emigre459

Joined Jul 22, 2006
10
OK, my graphics are posted. The file Amp 1 shows one of the quads in the first stage (too much stuff to fit in the screenshot, so I can't show the whole first stage, but I just copied what's in that file 3 times, so you don't really need the other 3 quads). The file Amp 2 shows the seconds stage of the circuit along with the relays and the 2-pole RC. I hope the labels on the components are legible.

As for the question of what happens on zero input, I see rough voltage following (on the scale of hundreds of nanovolts) on the two amps I made to have unity gain, and about 22 to 28 mV on all the others (which are still set to 10^3 gain). By that I mean that two of the amps had 22-23 mV, while the other 12 had 26-28 mV. These AD8574's have about a microvolt of input noise, so I would expect about a millivolt of offset on the outputs, not tens of millivolts.

And the signal I see on the final 10^4 output (after RC averaging) is about -200 mV without a probe attached and about -500 mV with one attached (I am probing by hand using a Keithley 2000 and those values I just gave were off of a digital 'scope I have attached to the circuit output). The 'scope is taking the differential of my signal: I made it that way so that I wouldn't have to worry about the 'scope's signal ground being different than my own. I hope this all makes sense... Thanks for the quick responses!
 

Attachments

JoeJester

Joined Apr 26, 2005
4,390
emigre,

After looking at your second amplifier, it appears to act more as a buffer than an amplifier, irrespective of the gain switching relay. I was also curious concerning the lack of an input resistor on the negative input side of U1.

Is there a particular need for paralleling all the circuits or is that just for testing purposes?

I'm also curious why you choose R6 as 1 M. Even if you use a x10 probe or a typical digital multimeter, your liable to load that circuit down getting self induced errors.

The ideal spice model shows a -288 mV output of your 1st stage amplifiers, and if you erred by not realizing your scope probe was a x10 [dividing the 288 mV by 10] you'll get that 28.8 mV reading.
 

beenthere

Joined Apr 20, 2004
15,819
Hi,

If I have this correctly, you sum the outputs of the first quad op amp into the input of the second inverter.

The total resistance of your summing resistors in parallel is 40 ohms. With the 1000 ohm resistor selected, the gain is going to be 25. Pull in the 10K resistor, and you have 250. Switching R3 or R4 in the non-inverting input will have negligible effects. Might as well leave a fixed resistor of about 100 ohms there.

By the way, the filter action of R5 & C1 is exactly equivalent to R6 & C2. It's not clear why you are using the filtering on the output to begin with - those are good components that should not give a noisy output.
 

Thread Starter

emigre459

Joined Jul 22, 2006
10
In response to Joe:

I'm sorry, the way my graphics are segmented it is difficult to see exactly how I made the circuit. First off, the second amplifier is made up of U2, U1 is a buffer made for the Reed Relays (not sure if that was a misunderstanding).

Secondly, by "paralleling" I'm assuming you're referring to all the identical input resistors? This was done for thermal reasons. I'm not sure how familiar you are with thermocouples, but they can form even across tiny SMT resistors due to temperature gradients. These dummy resistors on the noninverting inputs are placed extremely close to the negative input resistors on the actual PCB so that the two will be in essentially the same temperature zone of the board, which means that each has the same temperature gradient and they cancel each other out via the differential inputs of the op amps.

Third, R6 is 1M due to something I read in Horowitz and Hill. If you'll notice, both poles do have the same time constant since C2 is 0.1 uF. In the second pole I've increased the resistance 10x and decreased the capacitance 10x. I was told that whenever one adds another pole, one should allows make sure that subsequent pole has greater resistance than the previous one. Perhaps I misinterpreted the advice and this is where the problems are coming from?

Fourth, the negative input resistor for the second stage is actually located at the end of the first stage: that's the 160 ohm resistors. You see, we have 16 amplified signals in the first stage, all with 160 ohms at the outputs. We then bring all 16 signals together (summing them) at the negative input of the second stage and have (on default setting) a 10K feedback resistor. Thus, we have 16 x signal as Vin and -10K/160 as gain (= -1K/16), which takes the sum, divides by 16 (an average) then multiplies by -1000.

Finally, I have checked my equipment, and sadly I do not have the probes on 10x, the scope is set at 1x for both channels.
------------------------------------------------------------------------
In response to beenthere:

Once again, the point to the noninverting resistors is for thermal reasons and thus I need to have the same value for those as I do on the inverting input lest my thermal voltages will not cancel one another out. Admittedly, this might be overkill since this is the second stage and not the first, but I was trying to minimize any errors that could arise. So much for that, eh?

As far as the RC is concerned, I was worried about the noise from the AD857x components as they zero their offset using a random clocking frequency. Unless it is due to other problems in the circuit, I actually have seen this manifest itself at the outputs of the first stage quads: put in a dc signal, get out a periodic but otherwise indescribable output.

Did I do the summing amplifier incorrectly? I looked it up in textbooks and saw that this was how to do it and that the values of all input resistors should be equal unless you want a weighted sum. Also, the sum of them in parallel shouldn't matter, should it? According to the golden rules of op amps (the zero current rule) my Vo should be -[(Feedback R)/(Any one of the input R's, assuming they're equal)]*(V1+V2+...+Vn).
 

JoeJester

Joined Apr 26, 2005
4,390
Finally, I have checked my equipment, and sadly I do not have the probes on 10x, the scope is set at 1x for both channels.
X1 should load the signals more than a X10. I would recommend you do your measurements again using X10.
 

Thread Starter

emigre459

Joined Jul 22, 2006
10
Well, is that possible using a Keithley 2000? I mean, I'm taking differential measurements with two probes, that's not quite the same as using a probe on a scope, right? While I am using a scope, it is connected to a coaxial cable which is soldered on to my output, thus providing me with yet another differential signal (measuring voltage right off the board instead of using a probe with a specific resistance inside it to generate current and measuring that voltage drop).

Still, if you think it will help to switch over to a regular probe on 10x, I'll gladly give it a shot. I'll let you know what I come up with.
 

JoeJester

Joined Apr 26, 2005
4,390
emigre,

I thought you were using an oscilloscope not a DMM. So it was my err in understanding. Now that I understand what TE your using I'll review.

I redrawn your circuits in a classical schematic diagram. That is why I posed the questions.
 

Thread Starter

emigre459

Joined Jul 22, 2006
10
Well, I got the university's electronics guy to come by and take a look at my circuit. I thought you all would like to know what I learned.

First, it turns out that I'm not very bright. I looked at the datasheet for the ad857x and saw a maximum supply rating of 6 V. I took this to be +/-6V (i.e. V+ is 6, V- is -6). Nope, not so as I'm sure you know. So, I'm putting in a total of 10 V (-5 to +5) into chips that can't take more than 6 V without blowing up. Somehow I'm guessing that's the start of it.

Second, my relays were a brain fart. First off, he said that the relay in my feedback on the second stage should be put on the other side of the feedback resistors (something to do with low impedance paths and high impedance paths = didn't quite follow that part). Secondly, he pointed out that my dummy resistor relay is pointless, since I should only need to have a resistor there equal to the parallel combination of my 16 160 ohms, thus making it a single 10 ohm resistor (I think one of you guys mentioned that earlier and I didn't grasp what you meant until now).

Finally, in regards to the relay on the feedback of the second stage, he suggested a wiring that would make it so that even when the relay was open, I still didn't have an amplifier with open feedback.

I have a lot of work to do, obviously. My plan is to tackle the power supply issues first so that I can see if all my components are, in fact destroyed. Boy I hope not. I'm planning on doing that using the LM 317 and LM337 voltage regulators from National so as to take my current +/-5V and make it +/-2.5V. Obviously, if my choice of components sounds stupid and/or you think any of the above info is incorrect, I'd love to hear about it. Well, back to work...

EDIT: Changed voltage regulators from 2937 to 317 and 337 because I couldn't find a fixed negative regulator at 2.5 V and would like to reuse the components for other power supplies in the future.
 

Thread Starter

emigre459

Joined Jul 22, 2006
10
Well, I got my new power supply built, and things certainly have improved. However, now I'm experiencing a different problem. Whenever I measure the output of the whole circuit using a scope, I see a voltage drop after the RC filter at the end. Before the filter, I see the signal I would expect, after it I see a smaller signal. So I thought that perhaps this was due to scopes' notoriously problematic input impedances (can't remember, are they too high or too low?), and so I switched over to a much more effective measuring instrument: the Keithley 2000 DVM. I thought this would make it so that I didn't see that drop anymore. Instead, now all I ever see in the output is 85-90 mV. No matter what signal I give it, it always is the same. I've tried measuring using a Keithley 182 DVM and a LabVIEW DAQ. Both got 85-90 mV as well. How come everything is looking bad in the same way? I'm so confused...

Also, in case I didn't mention it, I've changed the feedback resistor on the second stage so that the circuit has overall gain of 10^4. In addition, I've removed the relay marked RL2 and just short circuited to the 1K resistor, which is now a 10 ohm resistor. Finally, I've rewired the relay marked RL1 so that the diode's input is connected to the buffer output, not DGND (and DGND is now connected where the buffer's output was).
 

Thread Starter

emigre459

Joined Jul 22, 2006
10
Oh no, I messed up, didn't I? I shouldn't have rewired that relay...that diode is a tamping diode, isn't it? Putting digital ground (DGND) on its input is the proper way to hook them up, isn't it? Could somebody please verify this?
 

Thread Starter

emigre459

Joined Jul 22, 2006
10
Hey, if anybody out there is still reading this, I just wanted to let you know that we are going a whole other way with the circuitry now since the chips are much more limited than we originally thought. Thank you for all your kind help, though! BTW, how do I close a post? If I don't hear anything from you guys, I'll just assume it will close itself. Thanks again!
 
Top