Thanks Bill!
Having '70s "vintage" metal base, grounded Protoboards makes a huge difference.
The plastic based ones are unusuable and the later metal ones don't ground the baseplate to the black binding post.
I should probably do a thread on how to modify modern ones.
I found a simple bias current-reduction trick that was used in the AP System 1 that I may implement here. (*)
With a typical ZTX851 pair the bias current, Ib is about 30 µA and the offset current less than 1 µA.
A differentially-connected cart, which is a floating connection, only sees the 1 µA Ios.
Ib however floats each input down to about -15 mV with a typical input pair. (30µA * 499Ω.)
In some minds this might limit using the preamp with unbalanced input connections.
Connecting the cart unbalanced misses the whole point but that doesn't prevent someone from wanting to.
The leakage current of an electrolytic coupling capacitor could easily exceed the Ib seen here for several minutes but that's another story.
Lowering the bias current, if it can be done without a noise penalty or more devices seems worthwhile.
To satisfy most of the bias current I propose to add a 10Ω common mode resistor from the junction of the input bias resistors (499Ω each) to ground and then bias the 10 Ω Rcm to +15mV.
Most of the bias current is then satisfied by the +15 mV placing the input bias current, Ib, into the same range as the Ios.
The net result is an Ib that is much lower and an input with a common mode voltage closer to 0V.
The +15 mV is derived from the low noise +11V shunt regulator, is attenuated by more than 50 dB and then appears in common mode.
It's not a differential noise source.
I tried it.
With my super-good ZTX851 pair having low Vos and Ios I got Ib to around 1µA.
In the sloppy pair with an approx 1 mV Vos and 3µA Ios Ib was lowered to about 5µA.
This is not bias current cancellation its more like bias current assistance.
For the price of two resistors I think its worth it.
* (In the AP System 1 the equivalent bias resistor is 100KΩ. The typical 5534 bias current is -500 nA so what AP did was inject 500 nA into the input. The -50 mV offset that would have developed is made closer to 0 mV.) https://proaudiodesignforum.com/forum/p ... 852#p12852