Ultra Low Noise - Ultra Low Impedance - Ultra big Problems!

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hello out there!

I tried hard to bould up an amplifier which is capable of:


  • gain around 40-60db
  • ultra Low THD >108db @10kHz
  • low enought noise to seperate k2,k3,k4 from the noise floor
  • instrumental Amp desing with high CMRR
  • source Impedance around 500mOhm
  • source Voltage around 40mV with a high common Mode voltage

For my desing i read the book written by duglas self and made the Decision to finally use opamps in a parralel desing.

First Testsetup:

I started with the good known Instrumental Amp topology and lm 49990 Opamps. Because of the low voltages, the high current ability of lm49990 i used "low" valures for all resistors.

As a Sinus Generator i used Frex Super Oszillator (which has got a THD less than 110db!!!).

The Signal was pushed down to around 40mVrms by an Lpad attenuator with a output resistance orund 4Ohm and was feed into my preamp.
Everything worked better than expected the K2 line was 110db under the fundamental, the K3 line was 118db under the fundamental (10kHz).

Second Testsetup:

I desinged a new Board with dual Input Opamps to lower the noise of the Instrumental amplifier and added a DC control loop.
The Amplifier worked as fine as the first one with a the "single" desing.
But when i started working with the real Source, the problems began!
I never Expeced such a difference between a 4Ohm Source and a 20mOhm source !
In the moment the input Opamps try to be a FM transmitter with a center frequency of 20Mhz.
I think I need an Input filter to improve this problem, but i don't whant to add a "big" resistor to the path.


What kind of Input filter would you suggest?

In the the next days i try to upload Scope screenshots and pictures of the PCB's.

Greetings out of Berlin, Stefan



APPENDIX:


Resistors in the First test:
R1,R2, zero Ohm
R6 (gain) 10 Ohm
R5,R4 (feedback)470 Ohm
R7,R8,R9,R10 470 Ohm
R16 0Ohm
R13 0Ohm
R18 47Ohm
IC5 not used


The second Testsetup inclouded a "slow" overvoltage input protection. For the tests it was not activated!
All Resistors were chosen simmilar to the first run (values on the chematics are wrong!)
There are a few caps and resistors choosen infront of the input as dummys for the input Filter
 

Attachments

  • SecondTestsetup.pdf
    40.3 KB · Views: 182
  • FirstTestsetup.pdf
    17.9 KB · Views: 115
If it's oscillating, it's got a (small) negative input impedance. Possibly your ground impedance is too high (did you use a groundplane?), or you got unlucky with power supply decoupling and effective rail impedance has a nasty resonance peak in the 20 MHz vicinity from interaction between capacitance and trace inductance (all of these caps have low ESR, too).

If as little as 4 ohms instead of 0.4 ohms cures it, I'd consider just including that bit of input series resistance. Each opamp has an input voltage noise equivalent of about 47 ohms, and you're adding about 5 with your resistor values. Adding 3.9 ohms, or 7.8 ohms per opamp, noise increases by a whopping 0.6 dB. Not terribly much. It's even less of a problem if you put a small inductor in parallel.

Alternatively, I'd try slowing down the first stage. The whole thing seemingly has more bandwidth than you need anyway.
 
Hi,

You may add a parallel LR network in series to the Inputs.
See also the Datasheet of BB/TI INA103, section "Applications Information", "Input Considerations" on page 8 and Fig. 2.
See also Figs 4 and 12 for the DC-Servo connection.
I assume that You are using the 3-OP amp topology, where each input feeds into a highgain noninverting OP amp.
The gain of both beeing set by a single common Resistor.
The outputs of which feed into a lowgain (Av~0.5-1) differential Input OP amp. Adding the DC Servo makes a fourth OP Amp.
I don't know if that is Your topology when You talk of "well known topology" and using "Dual OP amps to reduce noise" (??) and "IC5" .
I'm also not sure if 'slowing down' the first stage will help here, as the first stage pair already has a much lower bandwidth than the second, the differential stage.
Any impedance mismatch in the first stage will also spoil the CMRR.

Going the low impedance road is certainly a good idea regarding noise, but the limitations of the OP Amps have to be taken into account.
The datasheet of the LME49990 (see page 10) suggests to keep the load impedance, feedback Rs included(!), above 600R and recommends feedback resistor values of 1k-2k.
In Your circuit the OP Amps will see considerably lower values (470R/2).
I'd rather choose around 1k in the furst stages an around 2k in the differential stage
At the same and due to the high gain of the first stage and under "high common mode Voltage" conditions (not specified) each first stage's output has to put out large voltage values.
40mV (is that rms?) could mean ~120mVpp, hence 11.4Vpp at a gain of 95 (1+2*R4/R6). Adding some input common mode voltage and You end up close to or already above the drive limits of the input stages, or the input common mode range of the differential stage.
Btw: is the common mode part a DC value or does it contain AC also?
If it were DC only You might think of AC coupling the Inputs.

Since the large output voltage swings feed into a low (feedback) impedance, the supply decoupling caps might be of insufficiently large value.
Page 10 of the DS suggests 10uF + small ceramics and gives hints regarding layout and grounding.
The 10uF might not be enough for Your low load impedance condition.

So adding LR networks in series with the inputs and taking care of the supply decouplings should be the first measurements.

jauu
Calvin

ps: have seen the attachments just after typing :rolleyes:
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.