Charlie,
those photos are 1365 pixels wide and blowing-out the board layout. Annoying to read text when I have to scroll left-right-left on every line. (No, there isn't any practical forum-side fix.)
In addition to moaning about it, I've taken copies, down-sized and enhanced them, more suitable for forum use:
one -
two
Andy (or Admin or Jakob): do you want to
edit #12730 and use my copies instead of the 1365-wide ones?
>
"N... saw ours and ripped us off".
Oh, bosh. This general plan has been around forever. I agree with Alan: 1975 at least. I think I had seen a related topology,
"pseudo 'long-tail-pair' setup, with the pot between the emitters", used as a hi-Z preamp even before that. But it was early/mid 1970s before transistors were reliably good enough to work this way at microphone impedance. The PNP-NPN pair Alan mentions is certainly older than that, well-known in 1965. Using a cascade instead of a single has good and bad points, and is a matter of taste: good preamps have been built both ways. The nice feature of the cascade is that you can use low-Beta transistors, an important point for 1975. So, like most audio, it is all old bones in new soup. Alan's recollection of putting these bones together in this way meshes with my memory of the history of the time; a lot of designers were doing things like this then.
Or if they mean brand-N ripped off the idea of omitting the gain-set resistor.... everybody should be ashamed.
>
I don't fully understand why all of the pre's use a discrete diff pair either...even the SSL 9k does this also.
For low noise in low impedance you need LARGE-AREA devices. Chip-makers buy silicon by the square-foot, but sell it per functional-block, so their habit is to use small-area devices to get the maximum number of function-blocks per wafer. So historically there have been no chips that gave low noise in 150 ohms. Transistor makers also buy silicon by the square foot but are happy to slice out and sell you any size you want: 10mA or 10A. And just-plain transistors can be used many ways and are thus sold in vast numbers, and much cheaper than the same area of silicon on a chip. So with few (mostly recent) exceptions, you can get lower voltage noise from a pair of "high current switch transistors" than any chip, and possibly cheaper than the few lo-noise lo-Z chips on the market.
>
I am looking for rough gain number here.
Input stage Q1,Q2,U2A has a gain of ((8K2+8K2)/((4K7+4K7)||4K7))+1= 6.2
U2B has a max gain of around 14 (don't overlook R24).
Cross-coupled output stages like U1A,U1B suck. I'm going to pretend it has a gain of 2, but it may be more like 1.2.
Total gain is then 6.2*14*2= 174 or 45dB... is that what the spec claims?
Input stage has gain of about 6.2 and max out about 8V, so maximum input level is about 1.3V. That's reasonable, assuming you pad hyper-hot inputs like big-capsules on loud drums.
I don't know why they bothered with Q1 Q2. The noise level is determined by the
second U2B stage, NOT the input stage. The input stage may really come close to "no noise" in 150 ohms. But the second stage sees source impedance of 5K-15K at most settings, so the resistance noise alone is 5.8 to 10 times higher than the self-noise of a 150 ohm mike. The first stage only has gain about 6.2, so in most settings the main noise source is the second stage. (And that assumes that U2B is low-noise: I know the NJM4558 is not as bad as the old 4558s, which in turn were not the worst chips of 1975. But I'm sure U2B adds some noise beyond the resistance noise.)
C100 is too big for the 4K7+100 feedback resistors, and probably pointless at such low gain. It may be a leftover from a proper gain-control design. Short it and see what smokes.
If the output noise is high (I estimate 80dB below 1V, which isn't stellar), and you want to hack, put a proper gain-set network (with big cap) in place of R19, and bypass the whole R27/U2B stage. With that affair you won't be able to fade to zero, but you can't have everything.
C13/R23, C15/R22, C14/R26 are not doing great things for 20KHz response. In fact I'm not sure I like C11, C12, C13 or C15 at all, and C14 sure could be 22pFd.
Back-to-back C10 and C5 does not make sense to me.... could be a mistake? I suspect no cap at all is needed in front of the U1 output stage. None of the bass cap values look small to me, but there are so MANY caps that the total bass-droop may be bad. 2 or 3 DC-blocks per box should be plenty.
Wonder if we could get the first post with the messed up schemo deleted?? Thanks a bunch for posting my schemo ANDY!
