CML driver design in SiGe BiCMOS process

Hi,
It's my first time to design circuit by using BJTs.
I am doing a CML driver as below

This is simply modified from CMOS CML circuits, but I found it doesn't work at this point, maybe due to the base current (ib) too large? but if I make the size larger to increase beta, the speed maybe affected.

And it's easy to get this error with spectre simulation, I think it's due to improper BJT sizing?


I read reference, there is a picture as below

it says for (a), the input and output capacitances are vastly different. use emitter-followers to drive the input of the current swtiching pair can bring the input and output capacitances closer, like (b). input-output isolation is improved by using common-base stage at the output of the current switch as shown in (c). (d) uses another buffer stage to make the input and output capacitances of the gaincell equal.
No idea how the input and output caps can be made better for equal with emitter-followers.

Anyone had experiences in designing this? or any comments, references?

Thanks!

This design is extremely popular since the input works below ground on a single supply and is very cheap to produce.

SunnySkyguy said:

This design is extremely popular since the input works below ground on a single supply and is very cheap to produce.View attachment 108199

Click to expand...

Thanks!
Yes, the one you showed is a BJT Opamp with pnp pair input so that the input can go below ground.

The common emiter ciruits has poor input/output impedance, so maybe you should buffer the output of each stage or use **broken link removed** amplifier as single stage?

Dominik Przyborowski said:

The common emiter ciruits has poor input/output impedance, so maybe you should buffer the output of each stage or use **broken link removed** amplifier as single stage?

Click to expand...

The Rload of my circuit is small, actually.
Yep, I may should try this. But may still have problem when operating in the large-signal.
There is a comment about this topology in Razavi's book: The benefits of feedback in a Cherry-Hopper begin to diminish as the signal amplitude increases and the diff pair experiences large-signal operation. This means that the technique proves more useful in the first few stages of a limiting amplifier than in the last few.