I finally called up a schematic of the DMM reissue, and now understood what you're attempting to do. Much to my surprise, EHX uses the gain recovery circuit between BBDs that you showed, although why I don't know. It seems superfluous to me. Although, now that I think of it, I guess if they have one half of a 4558 for the gain recovery after the 2nd BBD, that means there is another unused op-amp sitting
nearby the first one, so......
The datasheet for the MN3008 shows there is usually no insertion loss, but also shows as much as +4db output or as little as -4db (relative to input, same specs for the MN3005), so I gather that is why the op-amp stage between the BBDs is capable of added gain as well as reduced output. In other words, it is there to compensate for part-to-part variability so that the S/N ratio and max delay level is the same across all production pedals. I guess when you crank out as many of these as EHX does, you can expect variation in BBD lots over time. That approach may provide some benefit to you, but personally I don't see it as essential.
You will also note that while the first BBD uses a pair of fixed resistors for output mixing, the second one uses a trimpot. If I was the one making this pedal, I might opt for trimpots on all BBDs, or alternatively use a pair of 1% mixing resistors (2k4) on the output of each BBD.
It is worth pointing out that one of the most highly praised analog delays out there is the Maxon AD999, and if you've ever seen gut shots, you'll immediately note just how many trimpots are in there. There is a bias AND a balance trimpot for each of the 8 BBDs it uses. So, while some manufacturers find they are satisfied with using a matched-value resistor pair for mixing, and using a single trimpot to derive a single bias voltage used for all BBDs in the circuit, the Maxon approach would suggest that maybe the difference in sound quality resulting from a few more adjustable parts is worth the builder's time and effort.
Datasheets indicate separate bias sources for pin 7 and 8, however the DMM simply taps the bias circuit at one point for pin 8 and a different point for pin 7. If it works for them, it can work for you. Essentially, what you would be doing is using the bias circuit that EHX uses for the first BBD, and applying it to all subsequent BBDs in the same fashion.
Does all of this make sense?
