Optimizing the Class B Emitter Follower

Is current drive better than voltage drive?

Introduction

In an ordinary three stage class-b amplifier, the emitter follower output is driven by a class-a stage. If the immediate load to the class-a driver is a simple resistor of lower impedance than impedance seen at the bases of the emitter follower, it is obvious that the emitter-follower is voltage driven. It occurred to me that the common alternative to load the class-a output with a current source creates current drive to the emitter follower circuit and therefore a push-pull current output.

If the current gain of the transistors were matched, it seems current drive would alleviate crossover distortion by the direct way the input current would be expected to drive an amplified current to the load. The input current would simply be driven as-is into the bases of the transistors to appear amplified without artifact to the output. The resistor load as an alternative would allow crossover distortion mechanisms full reign.

Presuming biases cancel at output and given input conditions:
i_signal = i_b1 + i_b2

and that current gains match (β_NPN = β_PNP):
i_out = β(i_b1 + i_b2)
i_out = β(i_signal)

Is this theory true?

To investigate I setup a SPICE deck modeling the class-a driver as a current signal source only, loaded as necessary to drive the emitter follower. This would eliminate all but the output stage as a source of distortion. I set the diode multiplier bias for a total of approximately 47mV across the two emitter resistors¹ and the input signals for an ac output of 5V peak.

Figure 1: Voltage driven emitter follower	Figure 2: Current driven emitter follower

SPICE Results for Voltage Drive

Fourier analysis for vout:
No. Harmonics: 16, THD: 4.63196 %, Gridsize: 1024, Interpolation Degree: 3

Harmonic	Frequency	Magnitude	Norm.Mag	Percent	Decibels
--------	---------	---------	---------	---------	---------
1	1000	5.30764	1	100	0
2	2000	0.118289	0.0222865	2.22865	-33.03916261
3	3000	0.208687	0.0393182	3.93182	-28.10812745
4	4000	0.00457183	0.000861368	0.0861368	-61.29622533
5	5000	0.0499479	0.00941057	0.941057	-40.52768141
6	6000	0.00538611	0.00101479	0.101479	-59.87247642
7	7000	0.0171109	0.00322382	0.322382	-49.8325843
8	8000	0.00432119	0.000814146	0.0814146	-61.78595413
9	9000	0.0040359	0.000760395	0.0760395	-62.37921495
10	10000	0.00188954	0.000356004	0.0356004	-68.97090245
11	11000	0.00154618	0.000291313	0.0291313	-70.71280269
12	12000	0.000202099	0.000038077	0.0038077	-88.38674552
13	13000	0.00321911	0.000606504	0.0606504	-64.34332661
14	14000	0.00127342	0.000239923	0.0239923	-72.39856234
15	15000	0.0028606	0.00053896	0.053896	-65.36886931

SPICE Results for Current Drive

Fourier analysis for vout:
No. Harmonics: 16, THD: 12.5555 %, Gridsize: 1024, Interpolation Degree: 3

Harmonic	Frequency	Magnitude	Norm.Mag	Percent	Decibels
--------	---------	---------	---------	---------	---------
1	1000	4.92209	1	100	0
2	2000	0.206496	0.0419528	4.19528	-27.54478097
3	3000	0.54593	0.110914	11.0914	-19.10027264
4	4000	0.0529415	0.0107559	1.07559	-39.36706488
5	5000	0.17918	0.0364032	3.64032	-28.77720877
6	6000	0.012936	0.00262815	0.262815	-51.60699703
7	7000	0.0724497	0.0147193	1.47193	-36.64225686
8	8000	0.00127415	0.000258864	0.0258864	-71.73856685
9	9000	0.0286507	0.00582083	0.582083	-44.70030169
10	10000	0.000680928	0.000138341	0.0138341	-77.18098179
11	11000	0.00867485	0.00176243	0.176243	-55.07776247
12	12000	0.000639199	0.000129863	0.0129863	-77.73029137
13	13000	4.62639E-05	9.39923E-06	0.000939923	-100.5381545
14	14000	0.00244699	0.000497144	0.0497144	-66.07035595
15	15000	0.00300727	0.000610974	0.0610974	-64.27954542

Final Thoughts

It is a normal outcome of design to conceive some outcomes that prove untrue. In this case the simpler resistor load wins out over the topology theorized to give a lower open-loop distortion result. Because current drive produces more gain than voltage drive, current drive would be expected to produce lower distortion after applying feedback however. Whether your design philosophy prefers the lowest final distortion or the most linear open-loop result will determine the final assessment of such results.

¹G. Randy Sloane, The Audiophile's Project Sourcebook, p.182. Author recommends 47mV bias across both emitter resistors as optimal bias apart from actual distortion measurements.

Document History
August 5, 2011 Created.