NE5532 Preamp Noise Level High With 47nF/10nF Feedback Caps, Gain 50, Scope Plot Attached

Hi Guys thanks for replying.


I did the following as you mentioned.
1. Inserted 47nf and 10 nf across the 10ohm feedback resistor (i didnt have 22nf)
This worked and reduced the RF sensitivity  drastically when i moved my hand close to the input components the output was not affected.
I also inserted 4.7nf across both 100k resistors. this possibly helped too.
The output had alot of white noise content and hissed alot through my cyrus power Link amplier into my speakers. I have attached a scope plot of the amplifier output. The scale is 20 mv/div, amplifier gain is 50. As  a test i shorted the + opamp input directly to ground as a test to see if i got less white noise but i didnt. The scope plot is with the input shorted to 0V.
Using 80mV peak  to peak as the noise / amp gain 50 = preamp noise output is 1.6mV peak to peak very bad.
I am assuming this is due to my bad layout. I try again with the component layout and keep the components as short as possible.
Any advice appreciated.
Has anyone made a low noise preamp using NE5532 opamps on vero board without the hacky upside down on vero board method because i am starting to think its impossible without a proper ground plane on a custom PCB.
Regards
Heath

Heath,The NE5532 is a good opamp and you only have a gain of 6, so there can't be too many problems.- does the noise stop when you short the input to ground?- are there any sources of noise nearby (an AM radio will help detect these)?- did you trim all unused part s of the veroboard tracks - an inch or more  of unused track will act like a good antenna and may either pick up noise or feedback and cause oscillation)?- are you sure your component values are right and you have not inadvertently made the gain much higher?- you are using +/-15V regulated supplies, so why do you need the input and output capacitors?  You  will need them if the amp is fed by or feeding to something with a bias on it, not otherwise.- what is your input? Could that be picking up noise (espeically if it is a coil of any sort?I'll think a bit more in the meantime, but let me know ref the aboveCheers  David

Here is a noise calculation for your preamp input, with the input shorted to ground:10 nV/sqrt(Hz) = NE5532 voltage noise density
30 nV/sqrt(Hz) = NE5532 current noise density * 10K ohms resistor13 nV/sqrt(Hz) = thermal noise density of 10K ohms resistorThese add up RMS'ly: Square each, sum the squares, take square root, giving total: 35 nV/sqrt(Hz) total noise density at preamp input.You state that the gain is 50. Then the noise at the output is 0.035*50 = 1.75 uVrms/sqrt(Hz).
The shunt capacitors are all at the amplifier input, so the gain and output are wideband. Your oscilloscope will show this wideband noise.

For the NE5532 GBP = 10 MHz, your gain-of-50 has bandwidth of 10 MHz / 50 = 200 kHz, so sqrt(BW) = 447 sqrt(Hz).
Multiply the output noise density by sqrt(BW) to get 780 uVrms = 0.78 mVrms wideband noise at the preamp output. Triple this to get an approximate peak-to-peak value 2.3 mVpp output noise. Your oscilloscope sees 1.6 mVpp, good agreement, so you seem to have a clean amplifier.
What sweep speed was your oscilloscope set to for the photo? This would tell you if most of your noise is indeed above the audio range. I see a signal waveform in the noise. What was the oscilloscope triggering on?
I suggest you listen to the output, and also run the output into your sound card and run spectrum-analysis software, to evaluate the noise in the audio band.
Cirrus Semiconductor uses NE5532's as input amplifiers on their demo board for their CS5368 eight-channel 24-bit audio ADC. Cirrus states that the PCB is two sided, so it might give you layout hints.

Heath, Peter has given you a better analysis than I could, thanks Pete.However, Pete, the gain of the amp is only +6 (1+10k/2k) - refer to previous post.I wondered if he might have inadvertently used a wrong resistor and got a lot more gain?(I have done worse things previously :-)And putting caps across the input and the feedback resistor should have reduced the bandwidth and thus the noise?The "Signal" may be 50/60 Hz mains pickup, and it would be nice to expand the trace to see if the noise is just noise or if there is a fairly obvious frequency (your spectrum analyzer would show this Pete).Heath - if you're only dealing with Audio you can fossick around on the net and get some half-decent instruments out of your PC sound card - signal generator and scope are common but I have seen a spectrum analyser as well.  Very limited - not much above 20KHz.  National Instruments and Analog Devices also make some USB muti-purpose devices that you can get fairly cheaply if you can prove you're a student.One other (very clutching at straws) thought.  You have a non-inverting amp.  Any feedback (due to layout etc) will be more likely to cause oscillation.  maybe try an inverting configuration instead if you don't ahve a good reason for this?  Swap your + and - inputs  and use a 1.8K or 1.5K from the - to ground (non inverting gain equation is just Rf / Rg without the +1.  (ideal would be 1.667 K for gain = 6)

Hi thanks for replying i am trying to build a low noise audio preamp. The gain of the preamp is about 6.
I could not measure the noise output of the opamp acurately so i put the output through my power amplifier  which has a gain of 50. Scope plot is that of the power amplifier output. I then divided the scope measurement by 50 to get the opamp noise output. The power amplifier noise was very low when i tried to measure it i could not acurately and its bare audible through my speakers with my ear shoved against the high frequency unit,
Regards

Hi I have built another preamp using ne5532 on an old veroboard and used the same type of turned pin socket for the I.C. This worked much better.
i used used the opamp from the failed board on the new one.The output white noise on the new board is far lower and a completely different bandwidth (very low when i connect the power amp and speakers to the preamp output in comparison to the previous board). I stuck to original circuit except I have a 1k ohm pot between the opamp input and ground to vary the volume. I suspect the problem is with the turned pin socket i used because the opamp ic seems to slide in much easier on the failed board whereas on the new board the IC has to be firmly pushed into socket. I suspect some of the pins dont make proper contact with the socket.Could this be the problem??. I am going to swap out the turned pin sock on the failed board and see if it make a difference.At least I  know its possible to build a vero board preamp using the ne5532.
RegardsHeath

Heath, glad you are sorted out.  depending on which pin was not making contact on your old socket, it may have been the problem (one of the input pins would have done that - they are very high impedance so would have picked up any noise in the area).
Your 1k Pot on the input - if you only need a low input impedance, why the two 100k resistors in the original design?  I trust you wired it as a pot, ie picking off a portion of the input voltage, not so it shorts the input when turned one way?This may have something to do with your problem - as above, higher impedance inputs pick up more noise. Most microphones are fairly low impedance so don't need high impedance inputs.The golden rule for scientific experimentation is - only change one thing at a time.  That way, if you get a change you know what's caused it.  Can I suggest you remove the 1k pot and see what happens? Or change it to a 10K or 100K?Cheers // David

In my noise-calculation post, I misunderstood that the gain from amplifier input to scope input was gain=50. Now I see that gain from amplifier input to scope input is 6*50=300, so my calculations should be corrected for that. Also, if you have a bandwidth-limiting capacitor across the feedback resistor, the calculations should be corrected for that also.Cirrus (on their CS5368 evaluation board) uses 634-ohm input resistor, instead of your 10K-ohm, to reduce resistor thermal/Johnson/Nyquist noise. Calculate the feedback resistor accordingly.I'm glad you found a bad socket.