[circuithology] White emitter-follower

[RoGeorge]

Found a mention about White Emitter-Follower on a time-nuts mailing list, while researching about DCF-77 receivers.  Didn't know the term, searched and found it doesn't have to do with time keeping.  White emitter-follower is a circuit that behaves similar with the single transistor emitter-follower (common collector), but with many of the pitfalls of common collector fixed by a second transistor added and some feedback.

- it's an improved emitter follower, patented by Eric L. C. White in 1944, for tubes, AKA White Cathode Follower (WCF)
- advantages in comparison with emitter-follower
        - symmetric response
        - better slew rate
        - amplification closer to 1.0
        - higher input impedance (about 30 times)
        - much smaller output impedance (about 100 times)

Patent, origins and circuit analysis:
https://www.pearl-hifi.com/06_Lit_Archive/02_PEARL_Arch/Vol_04/Sec_19/991_White_Follower_Optimization.pdf
https://trace.tennessee.edu/cgi/viewcontent.cgi?referer=&httpsredir=1&article=3833&context=utk_gradthes


Simulated schematic is from the thesis in the second link.

Notes:
- normal emitter-follower
        - asymmetric push/pull current capabilities (asymmetric raise/fall slew-rate)
        - asymmetric amplification for each alternance (transistor's beta is bigger at bigger Ic)
        - output impedance in the range of few ohms
        - voltage gain almost 1, but worst than the White emitter-follower
       
- White emitter-follower
        - might oscillate for Rgenerator 50 ohms or smaller (so add some extra Rseries ~600 ohms at the input)
        - for Rload between 50ohms-5kohms, the Rinput was between 0.83M-9.1Megaohms
        - output impedance less than 0.5ohms (vs. ~15 ohms for a normal emitter-follower)
        - gain voltage was 0.998 (vs. 0.984 for a normal emitter-follower)
        - Rgenerator and Rload controls overshooting



At an AC simulation, the output impedance is less than 0.1\$\Omega\$ for a very broad frequency range.  All these look almost too good to be true for one extra transistor, plus the circuit is not mentioned very often, so I wonder what's the catch.

Anybody tried using the White emitter-follower in practice, how well does it work?

[42Khz]

It's a fairly popular circuit in (CMOS) chip design, although there it is typically referred to as a "flipped voltage follower".

[T3sl4co1l]

Note that the advantages are smaller because BJT gm is so much higher than FETs (vacuum type or otherwise).

The push-pull drive may be advantageous in some applications, though it is still class A of course, no efficiency savings.  Well, efficiency over a plain load resistor yes, just to make that clear.

Tim

[PCB.Wiz]

At an AC simulation, the output impedance is less than 0.1\$\Omega\$ for a very broad frequency range.  All these look almost too good to be true for one extra transistor, plus the circuit is not mentioned very often, so I wonder what's the catch.
Anybody tried using the White emitter-follower in practice, how well does it work?

That's a lot of added parts, including some large capacitors, so today's opamps would make that redundant.

[mawyatt]

There's also a Super Source Follower (what we called it back then). We utilized such in a custom Low Noise LDO specified specifically created for integrated microwave & MMW VCOs in a SiGe BiCMOS process (IBM7, 8 and 9HP).

Much of the close-in phase noise in many VCOs is attributed to the low frequency supply noise since most VCOs have poor low frequency supply rejection, and we needed something better than what typical LDOs could deliver. This specialized LDO utilized the Super Source Follower and a unique bandgap reference decoupling strategy, later patented (8692529).

https://www.researchgate.net/publication/302724491_Low_noise_low_dropout_voltage_regulator/fulltext/579677ae08ae33e89fad8862/302724491_Low_noise_low_dropout_voltage_regulator.pdf?origin=publication_detail


Edit: Here's a sketch of the Super Source Follower concept. An interesting note about the noise filter is since no DC current flows thru the resistor other than M2 gate leakage current, a relatively large resistance can be employed to keep the shunt filter capacitance small.

Vaguely remember an experiment where the R was replaced by a gated pinched FET which would be "gated off" when the VCO was active in burst mode, the charge on C would hold for the burst duration. Even more extreme was the use of gate charge trapping which we had used on a unique RF Spectral Imager, where the charge was captured and stored on an effective floating gate for M2, this would hold long enough for the burst VCO use. Anyway, all in the interest of getting Lower Phase Noise integrated VCOs.

Best,