>>
103 noise is around 0.2 microvolts, 134 noise around 7 microvolts
>
how do you get that?
I am taking the maker's word. And I am foolishly looking at "Typical specs", when I know that only the Guaranteed Maximum Specs are to be trusted (and not too far or too literally).
But I'm also using common sense. The 103 has naked high-current transistor inputs. The 134 is wrapped up in resistors.
This is input noise, what we usually look at when dealing with weak sources.
This is "output noise", but it is unity-gain so if we refer it to the input we get the same noise.
The "1KHz spot noise" is claimed to be: INA103=1
; INA134=52
The total noise depends on the bandwidth you care about. In audio this is usually 50Hz-15KHz or 20Hz-20KHz. The bass-end makes little difference if the noise is "flat": 15,000 and 14,950 are "the same". The noise usually is not flat, but then again our ears are very un-flat for low-level bass. Since so many speakers and many listeners are fairly flat to 20KHz, assume 20KHz.
Total noise increases with the square-root of bandwidth. The square-root of 20,000 is 141.
So the total 20KHz noise will be about: INA103= 141*1
= 141 nanoVolts (0.141 microVolts); INA134= 141*52
= 7332 nanoVolts (7.3 microvolts). The 134 spec actually says 7uV.
Passive 200 ohm microphones will tend to have 100-200uV (0.1-0.2 microVolt) self-noise, so the preamp's noise should be in that area.
"Line level" is usually 20dB-60dB higher than pro-mike level, so 20dB-60dB (10X to 1000X) more noise is tolerable. The INA134's 7uV of noise is not the very best we can do, but really good enough for almost any line-level use.
>
I´m a little confused I saw the datasheet of the ina103. and I see the noise voltage v/s frecuency curve the noise with a gain of 100 or more is less than with a gain of 1. I can understand that.
"can" or "can not" understand? Well, I mentioned the concept in another thread but didn't beat it to death yet.
B-B's official explanation is:
Since they cite the specs with ZERO source impedance (Rg), they say total input noise is the sum of input-stage noise and the output-stage noise divided by input stage gain. They also say the 103's output stage (which is very similar to a 134) has more noise (it should be less because lower resistances).
I am looking at the gain-set feedback resistor. Its thermal noise sums with the source noise and the transistor noise.
The expected source resistance is about 200 ohms. That also seems to be about the optimum source for these transistors.
At gain of 1,000 or 100, the gain-set resistor is 6 or 60 ohms, much less than 200 ohms. It adds nearly no noise.
At gain of 10, the gain-set resistor is 660 ohms, much more than the source resistance. It adds a lot of noise.
At gain of 1, the gain-set resistor is "infinity". Actually the added noise resistor includes the two 3K series resistors, so the noise source is 6K. This is much-much higher than 200 ohms, and should give about 8 times the noise. More because all those 6K resistors in the second stage. Even so the noise is higher than you figure from simple resistor thermal noise... the input transistors are working way off of their optimum (200 ohm) impedance, and shouting-out their current noise. If we only used it at lower gain, we would bias the input transistors to a lower current to reduce their current noise.
Mackie has several papers about low-noise mike amps featuring low-impedance gain-set networks, and several trademark names implying "Very Low Impedance". They are certainly correct: there was a whole generation of transfomerless mike amps with large feedback resistors. When used at high gain, the gain-set resistor is smaller than 200 ohms so the noise is low. But when used at low gain, the gain-set resistor value is high, total noise resistance seen by the input is high which degrades noise, plus at some point the current-noise in this resistance rises directly as gain (instead of square-root of gain). And in some designs the noise got quite high at moderate gains.
As I said, Mackie is on the right track, but implementation of
VERY low feedback resistors for all gains is tough. See my "novel preamp" thread. To get really low resistance noise at moderate gains requires resistance values you can not do on an affordable variable resistor, and probably a capacitor too big to fit in a modern channel strip. And more drive current than you can get from affordable chips. Mackie's designers probably sketched-up the "right way", then multiplied by 8, 12, 16, or 24 and esitmated the cost, and went back to the drawing board. Their "Very Low Impedance" mike amps are better than many, and really quite good in many situations, but for practical reasons are not all the way down to the noise floor at all gains.