I have been working on a diy audio mixer based on ne5532 opamps with auxiliary output/input, panning and baxandall tone control.
Here is the circuit of one of the inputs:
(+-12v supply, 100nF on each Vpin of the opamps) (this "input module" will be duplicated for each chanel I make and all chanels get mixed in other circuit).
I made the PCB and everything worked fine (except the tone pots where inverted whoops
). However the preamp adds to much noise to the signal and makes it pretty unusable for instrument-level inputs.
The noise (hiss) gets amplified when i raise the gain. And gets worse When treble pot is boosted... It also has some hum but i think thats because i havent shielded it yet.
Maybe you, the opamp-gurus, could give me some design tips on how to improve signal/noise ratio on the preamp. Or modifications to improve the rest of the circuit??
Here is the circuit of one of the inputs:
(+-12v supply, 100nF on each Vpin of the opamps) (this "input module" will be duplicated for each chanel I make and all chanels get mixed in other circuit).
I made the PCB and everything worked fine (except the tone pots where inverted whoops
). However the preamp adds to much noise to the signal and makes it pretty unusable for instrument-level inputs.The noise (hiss) gets amplified when i raise the gain. And gets worse When treble pot is boosted... It also has some hum but i think thats because i havent shielded it yet.
Maybe you, the opamp-gurus, could give me some design tips on how to improve signal/noise ratio on the preamp. Or modifications to improve the rest of the circuit??
Your panpots short op-amp outputs when panned to one side and your mixing resistors for the auxiliary output are way too high.
What do you mean by "instrument-level inputs"? Do you want to connect an electric guitar to the input? As far as I know, those have a fairly high impedance, which makes a good FET op-amp better suited noise-wise than an NE5532. Mind you, I know next to nothing about electric guitars.
What do you mean by "instrument-level inputs"? Do you want to connect an electric guitar to the input? As far as I know, those have a fairly high impedance, which makes a good FET op-amp better suited noise-wise than an NE5532. Mind you, I know next to nothing about electric guitars.
Oh I though thermal noise was a problem only with very high values ( like 500k, 1M ).
I will try lower the resistors. But keeping R7 and R8 at their value to maintain high input impedance?
Resistor thermal noise voltage is proportional to sqrt(R).
So reducing a resistor by a factor of ten, reduces thermal noise by about three times (10dB).
What is the lowest possible input load resistor that will work? Use that value.
But use metal film resistors only. Carbon comp will have much more excess noise.
So reducing a resistor by a factor of ten, reduces thermal noise by about three times (10dB).
What is the lowest possible input load resistor that will work? Use that value.
But use metal film resistors only. Carbon comp will have much more excess noise.
I am no guru but I recently learned that carbon resistors are noisy and stacked ceramics distort. They are ok for power supply lines but signal circuits need more care.
Didnt see that! fix: (stereo signal will be stronger at full left or right than in the center, I cant figure out how to change that).
Levels less than a volt, I plan to to make the mixer flexible and use it for various types of inputs.
yeah I used carbon, lol didnt there was a diference between carbon and metal resistor other than the price. I used electrolytics on the input and ceramics on the LPF of the preamp
47k?. Thanks, any sugestions with the model of the pot? I used the 16mm ones. But i know those are not the best...
For all the other resistors and potmeter values in the circuit, reducing them will reduce noise, but R7 and R8 are an exception.
Assuming you are not going to listen with nothing connected to the input, R7 and R8 need not be reduced. In fact reducing them will worsen the noise.
The reason for that is that they are in parallel rather than in series with the signal path. Reducing them will reduce their noise voltage, but it will reduce the voltage division between them and the source impedance even more, so in the end their contribution only gets worse.
Another way to look at it is by describing their noise with a noise current rather than a noise voltage (Norton equivalent).
Last edited:
Of what order of magnitude is the output impedance of whatever you connected to the input when you found you had a noise problem? What was the thing you connected there?
Regarding resistor noise, resistors have two types of noise:
Thermal noise. This depends on the resistance and the temperature, but not at all on whether it is a carbon composition, carbon film, metal film, wirewound or any other type of resistor.
1/f noise, also known as flicker noise or excess noise. This manifests itself as random variations of the resistance. It depends a lot on the resistor type and model:
Carbon composition: very bad
Carbon film: far better
Metal film: better yet
Wirewound: so good it is almost immeasurable
There are a couple of reasons why wirewound resistors are usually impractical, so we often end up using metal film.
However, as the excess noise consists of random variations of the resistance, you will only hear it when there is a voltage across the resistor. In your case, there are no large DC bias voltages across the resistors, so I would not expect your circuit to be very sensitive to resistor excess noise. It certainly won't do harm to use metal film and when there is a signal, it reduces 1/f noise sidebands around the signal, but unless you use ancient carbon composition resistors now, I would not expect it to do much for the noise floor.
Thermal noise. This depends on the resistance and the temperature, but not at all on whether it is a carbon composition, carbon film, metal film, wirewound or any other type of resistor.
1/f noise, also known as flicker noise or excess noise. This manifests itself as random variations of the resistance. It depends a lot on the resistor type and model:
Carbon composition: very bad
Carbon film: far better
Metal film: better yet
Wirewound: so good it is almost immeasurable
There are a couple of reasons why wirewound resistors are usually impractical, so we often end up using metal film.
However, as the excess noise consists of random variations of the resistance, you will only hear it when there is a voltage across the resistor. In your case, there are no large DC bias voltages across the resistors, so I would not expect your circuit to be very sensitive to resistor excess noise. It certainly won't do harm to use metal film and when there is a signal, it reduces 1/f noise sidebands around the signal, but unless you use ancient carbon composition resistors now, I would not expect it to do much for the noise floor.
Related to Jean-Paul's remark: according to an old application note I have somewhere, in the long run, any DC current flowing out of a carbon potmeter's wiper causes anodization of the wiper and results in a rotten contact. It is therefore good practice to either AC couple a potmeter, especially the wiper, or not to use carbon.
You might want to have a look at Douglas Self's "Small Signal Audio Design".
I connected my phone in the input (as i don´t own a signal generator). I dont know the output impedance of it, but its probably several ohms low.
I don't understand why all amplifier designs have fixed gain, with attenuation pots at the input
What are the advantages/disadvantages of variable gain vs fixed gain with input attenuation? At first sight, variable gain seems more "logical (?)"
Ok. so in this case, using lower values is more important than the resistor´s material.
OK, then the op-amp's equivalent input noise current won't have made any difference worth mentioning.