Alternative topologies for balanced to unbalanced amplifier

[Elvee]

I built a quick prototype of the circuit I gave in the first post.
Here is the amended schematic after "real-life-corrections".

On the dynamic aspects, it does deliver: it is capable of putting out a clean 20MHz, 5Vpp squarewave into a 50 ohm load, which is quite impressive.

Its Achilles heel is the DC accuracy/stability.

[MarcelvdG]

Originally Posted by Elvee

OK, here is the unbalanced-output, wide(r) bandwidth version.

It is good to ~10MHz. Going higher would require real HF transistors.

And as a bonus , a slightly improved version of the original circuit with a more accurate 0V level at the output.

If C1 is omitted, the outputs will behave like a grounded center-tap transformer's secondary.

I deliberately gave the original a common-mode loop that is both weak (low transconductance) and slow (filtered by C1), so that the output behaves more or less like a floating voltage source at audio frequencies, that is, like a normal transformer secondary rather than a centre-tapped transformer secondary. That way you can use the output in single-ended mode without the common-mode loop needing to do anything with the signal, but you also keep the flexibility of using the output in differential mode. The accuracy of the DC common-mode level was rather irrelevant because I could use AC coupling. I designed it for a local radio station that uses both professional and consumer gear, and there it really comes in handy.

Anyway, your single-ended version looks nice and is definitely more suitable for your measurement application.

[MarcelvdG]

Originally Posted by Elvee

I built a quick prototype of the circuit I gave in the first post.
Here is the amended schematic after "real-life-corrections".

On the dynamic aspects, it does deliver: it is capable of putting out a clean 20MHz, 5Vpp squarewave into a 50 ohm load, which is quite impressive.

Its Achilles heel is the DC accuracy/stability.

How about the low input impedance and the resistor-matching-dependent common-mode rejection (which is probably not too bad because you have gain>>1)? Are you only measuring signals from very low impedance sources?

[Elvee]

The low impedance is not a problem, it is a 50 ohm environment.
Having the common-mode dependent on resistor accuracy is not a problem for me (although I would prefer freedom from that constraint).
Anyway, I could also add a second Rush cascode for the feedback, very much like your circuit.
More annoying is the fact that the ultimate low frequency CMMR depends on the matching of Early effect in complementary transistors, mainly Q5 and Q6.
The 2N2222/2N2907 pair is quite good in this respect, they achieve more than 70dB up to 500KHz.

The poor DC stability is annoying too, and adding a DC servo would add a layer of complication.

All this is manageable, but I would like to find something friendlier

[MarcelvdG]

Originally Posted by Elvee

The low impedance is not a problem, it is a 50 ohm environment.

50 ohm is a characteristic impedance that isn't used much in audio applications, except in broadcast transmitters. Are you sure you are on topic?

[Elvee]

Originally Posted by MarcelvdG

50 ohm is a characteristic impedance that isn't used much in audio applications, except in broadcast transmitters. Are you sure you are on topic?

Yes, because as I said, I am interested in topologies, and I remember having seen some unusual ones on this forum.
Some might be totally unusable above several hundreds of KHz, but others could perhaps easily reach tens of MHz by lowering resistance values, etc