bcarso
Well-known member
Side note: the resistor will always be quieter than the transistor-voltage ref-resistor current source, for the same available supply voltage. If this noise source is insignificant then not to worry---maybe the enhanced common-mode rejection is something you need and justifies the complexity and higher noise. But as PRR points out you may do just fine with little CMR.
But the noise analysis and straightforward measurements are perilous: the well-balanced diff stage will tend to make your zero-signal noise low, as it rejects the noise in the tail current. But at non-zero signal, the diff pair is perforce unbalanced and a variable fraction of the tail current noise gets through. So your noise is modulated by your signal. This is rarely accounted for and most often ignored, although it could in principle be decoded from THD+N sweeps with AP's and their ilk.
Another way to clean up currents is to use an inductor (the other white meat ;-) as part of the circuit...something rarely done for reasons of size, weight, parasitics, limited frequency range, potentially damaging transients, sensitivity to external fields, and cost, but still an option.
I did a preamp design once where, in order to minimize the first stage noise, I ran the current source load from a few hundred volts so the resistor to the transistor source could be large, and the transistor's voltage noise be negligible. That was wretched excess, although it worked well after one problem was finally solved. I learned my lesson about designing for low mains voltage the hard way, as the HV supply began to have onset of dropout and hence hum at very low line voltage, producing mysterious quasi-periodic spikes in the detector readout. Many long faces or worse at the telescope, while being reminded how many astronomers wanted the time we were using troubleshooting the instrument. This took months to understand and correct, and of course was obvious once understood.
But the noise analysis and straightforward measurements are perilous: the well-balanced diff stage will tend to make your zero-signal noise low, as it rejects the noise in the tail current. But at non-zero signal, the diff pair is perforce unbalanced and a variable fraction of the tail current noise gets through. So your noise is modulated by your signal. This is rarely accounted for and most often ignored, although it could in principle be decoded from THD+N sweeps with AP's and their ilk.
Another way to clean up currents is to use an inductor (the other white meat ;-) as part of the circuit...something rarely done for reasons of size, weight, parasitics, limited frequency range, potentially damaging transients, sensitivity to external fields, and cost, but still an option.
I did a preamp design once where, in order to minimize the first stage noise, I ran the current source load from a few hundred volts so the resistor to the transistor source could be large, and the transistor's voltage noise be negligible. That was wretched excess, although it worked well after one problem was finally solved. I learned my lesson about designing for low mains voltage the hard way, as the HV supply began to have onset of dropout and hence hum at very low line voltage, producing mysterious quasi-periodic spikes in the detector readout. Many long faces or worse at the telescope, while being reminded how many astronomers wanted the time we were using troubleshooting the instrument. This took months to understand and correct, and of course was obvious once understood.
)
