lme49990 noise problem

Hi David,

I don't believe that we have an issue with the noise spec on this part; I haven't heard of any other complaints or issues. Can you quantify the noise level that you are measuring over what bandwidth and how you are measuring it? Additionally, providing an application circuit would be helpful.

Best Regards,

JD

JD;
Thanks for your response to my enquiry.  The "noise" is just that, what I hear on my speakers even when there is no input to the preamp.  It is a hiss, or "white noise" with a broad spectrum.  When I look at it on a scope, I see just a lot of hash, with no particular frequency standing out.  When I replace the 49990 with another type of chip, such as the opa2132, the noise is reduced about tenfold.  The 2132 is an acceptable level, but the 49990 is much too high, and I have had many complaints about it when demonstrating my prototype to customers.
I researched the internet to find out what could be done, and there were some who thought that the chip is just more sensitive to rfi, so I added ferrite beads at the signal inputs, and an rf choke at each power pin. I reduced the feedback resister from 10k to 500k, and used a super low-noise resistor type (metal foil) at this position, which was also suggested.  I also use triple-bypass caps at each power pin, as suggested in the data sheet, and a complete ground plane extending over the back of the board.  I have also had stability problems, but I used a 10k ohm isolation resister at the inputs, and this seemed to solve that problem. 
In short, I have done everything I can think of, as well as what has been suggested by others on the internet, but the noise is still there, and at an unnaceptable level for my potential customers.
My option, at this point, is to use another type of opamp (such as the opa2132) which sounds pretty good, but the 49990 is the best sounding of any opamp out there (in my opinion) and I would sure like to be able to use it.

There is also the issue that I have mounted two of the lme49990 on an adapter board, so that they may be plugged into an 8 pin dip socket, and emulate a typical 8 pin dip dual opamp chip.  The surface mount chips are mounted back to back on a small circuit board.  I don't know if this could create a problem.  See picture below.  Also see attached schematic.

I really want to know if there is nothing more to be done, or if the lme49990 really does have 10 times the white noise of another typical chip.  And if you can think of anything else to try, in order to reduct the noise, please let me know.

Thanks

picture of chips mounted on adapter board:

Schematic:

JD:
OK, thanks for the input; I think you have put me on the right track.  When I short the input, as you suggested, and swap chips, the 49990 has no discernible noise, while the 2132 does still have some slight noise.  So, I guess you are right that the 49990 is really low distortion.
So then, when I replace the 10k input resistor with 1k, the noise does reduce quite a bit, but the 2132 is still slightly lower in noise, with the same setup.  I also tried 100 ohms, and also 0 ohms.  I tried this with both my boards, one that has all the ferrite beads, and one that does not, and this does not seem to make any difference.  So, it seems that the 2132 is still less sensitive to whatever is going on at the input.
So, overall, my design is much more satisfactory, but if you have any further ideas of what can be done to reduce the input noise even more, I would appreciate it.

Thanks;

David

Hi David,

It is good  to hear that shorting the input to ground measurement confirms that there isn't an IC noise related problem. 

I believe you are on the right track with utilizing an input filter to limit the amount of input pickup, but I would discourage using a series resistor directly to the input. The input current noise will wreak havoc.

Typically a lowpass filter is created with a small valued series resistor (100 to 470 ohms) connected just after any coupling caps, with a termination resistor to ground (10k to 47k) with a parallel small valued cap to ground. The pole of the filter can be selected to your desired input bandwidth, say 20kHz or more. Some people also use two stages to get a 40dB/dec rolloff. With this configuration, you will be limiting noise pickup and without the series resistor connected directly to the input your noise from the input current noise should be kept low.

Please let me know if this type of configuration works for you.

Best Regards,

JD