It's not a CA3096 but...

In message , bitrex writes

I used one of these for a Gilbert cell .

There's also these:-

Brian

--
Brian Howie

Whew, such poor transistor specs!

--
 Thanks, 
    - Win

Looks like the same link twice, there.

They're like 2N3994s/2N3906s in a package but every important spec (fT, saturation voltages, hfe) are a little worse than the discrete device, lol. The PNPs are fast-ish, though, and one would hope the Vbes and gms fairly well-matched to their same-polarity device?

fT of the PNPs in the CA3096 as specced in the 1997 datasheet is only

6.8 MHz typical @ Vce 5V, Ic 100uA. The NPNs in the 3046 are 550MHz typical @ 3V, 3mA

Mediocre specs, and _nothing_ about matching. No wonder bipolars are fading :(

Well, it's not an IC. That means no substrate, and totally independent parts. But the individual pairs are likely from nearby locations on the wafer, probably pretty well matched. The transistors might actually not be too bad; in part it's the manufacturer's specs that are horrible. AoE III has measured beta values for a sample of these transistors, see Figure 8.39, page 504, curves 3 and 31. NPN 2N3094 beta was 195 to 190 for 1uA to 50mA, PNP 2N3906 was 70 to 210 at 10mA. There's Early-voltage V_A data in Table 8.1a on page 501.

--
 Thanks, 
    - Win

Why not the HFA3101, which is already wired as a Gilbert cell?

I had to separate off one of the base pairs to unbalance it slightly for the particular application.

Missing link

Brian

--
Brian Howie