Well normally in a balanced mic preamp you use a PNP followed by a NPN on each leg ... Something like this: http://sound.westhost.com/project66.htm ... A variation of this topology is what I currently have in my test bed
But we don't need that kind of gain, those circuits are a lot like the INA and SSM preamp chips (In fact the INA/SSM chips are basically this very circuit integrated onto an IC), they run the most quiet at high gains (40Db or better) and the noise really starts to pick up with a 20Db gain .... In a pro level mixer you need to amplify a mic to a level of a couple of volts P-P or more, this is why everything is run off of +/- 15V bipolar supplies, in the case of an SDR we only need enough amplification to 'trip' the lowest few bits in the AD converter So I'm basically experimenting with circuits using a single PNP on each leg (PNP's are almost always quieter than their NPN complement) leaning with a common emitter circuit, If the impedance was 50 ohms instead of 200 ohms each I'd go with the common base ... most of the gain comes from the transistors, the opamps main job is a differential amp to summing the two legs together, balance is the key here so a bipolar supply is almost a necessity, it certainly makes things easier from a noise and balance standpoint ... but since we have some amplification before the opamp, the opamp's noise is swamped out by the signal and also balance (For CMRR) isn't as critical as it is when you have 20 foot of mic cable between the source (mic) and the preamp inputs so a single supply variation is something I'm looking into ... other 'tricks' are to put 2-4 PNP's in parallel, two in parallel cuts the noise in half (Or lowers the NF in half) 4 in parallel cuts the noise by 4 ... after 4 you reach a point of diminishing return and 2 would probably be adequate as well as cost effective Here's a neat little "Compound" opamp circuit with uses a transistor on the input AND in the feedback path and takes an otherwise mediocre opamp and makes it nearly as good as the most expensive ones ... The 4136 is one of the orginal low noise audio opamps but it's only good up to 2v p-p before distortion becomes a problem ... Think what this circuit could do with a NE5532/4 and adapted to a balance configuration ... Somewhere in one of my hundred or so noebooks I have several circuits like this but I still haven't found them When I finalize a design don't worry I'll share it JR JR JR --- In [email protected], Grant Hodgson <[EMAIL PROTECTED]> wrote: > > JR wrote :- > > --snip-- > > >> Even a pair of 10 cent 2N4401's in front of a NE5532/4 > will stomp about any of the expensive 'designer' low noise op amps ... > Just ask the engineers at Soundcraft, Mackie, Yamaha, or Beringer who > all use a pair of transistors ahead of the opamps in their > professional balanced microphone preamps ... > > Do you have any more details of circuits for this PNP buffer? Common > base? Common emitter? The Janus board in the HPSDR project uses > LT1128s, which are single-eneded low noise op-amps. I was thinking of > using AD4841s or the even lower noise LT1115s for the same function in > the uWSDR project, both of which are somewhat cheaper than the LT1128. > It may be the case that the noise of the op-amps is not too important, > system noise figure being dominated by the noise of the RF front-end > devices - I haven't done a system analysis yet. > > > >> Problem is I'm at a > standstill because my DDS's are too noisy to take advantage, hopefully > the 995x based DDS Controller (David Brainerd's design) will be quiet > and spur free enough to explore further > > The new AD9910 and especially the AD9912 are the current 'best in class' > - by a big margin. I've done some spur measurements of the AD9910 and > it looks very good indeed - no spurs above approx -70dBc or so, phase > noise measurements are a little more involved but I hope to have those > done before Christmas. For now, I think it's safe to rely on the > 9910/9912 data sheets, which show exceptionally low levels of phase noise. > > >> JR > > regards > > Grant >
