As a continuation to my previous article, I decided to make some more rigorous measuring. I grabbed 8 transistors from each type and measured hFE, and plotted all samples.
Tables
Raw data
| PN2222 | 2N3904 | BC547 | BC546 |
|---|---|---|---|
| 8 | 324 | 209 | 541 |
| 30 | 340 | 217 | 527 |
| 96 | 316 | 235 | 511 |
| 91 | 335 | 204 | 529 |
| 39 | 323 | 232 | 527 |
| 18 | 324 | 228 | 548 |
| 14 | 339 | 194 | 504 |
| 8 | 345 | 204 | 451 |
Statistics
| PN2222 | 2N3904 | BC547 | BC546 | |
|---|---|---|---|---|
| Samples | 8 | 8 | 8 | 8 |
| Min | 8 | 316 | 194 | 451 |
| 25% | 12 | 323 | 204 | 509 |
| Median | 24 | 329 | 213 | 527 |
| 75% | 52 | 339 | 229 | 532 |
| Max | 96 | 345 | 235 | 548 |
| Mean | 38 | 330 | 215 | 517 |
| Std | 35.885 | 10.306 | 15.023 | 30.335 |
| Variance | 1287.714 | 106.214 | 225.696 | 920.214 |
| CV | 0.9443 | 0.0312 | 0.0698 | 0.0586 |
| Skew | 1.1033 | 0.0059 | 0.0675 | -1.6881 |
| Kurtosis | -0.4912 | -1.5993 | -1.5749 | 3.418 |
Charts
Box plot
Here are the plots for each type of transistors as well
BC546
BC547
2N3904
PN2222
Conclusion
The PN2222 batch exhibits unacceptable hFE variation (CV = 94%), with a mean of 38, below typical application needs and near the lower spec limit. In contrast, the 2N3904, BC547, and BC546 lots show tight, centered distributions well within datasheet ranges.
While the suspected «fakes» (BC547, 2N3904) perform reliably in DC gain, hFE alone is insufficient. High-frequency behavior (ft), saturation voltage, and noise remain untested.
Recommendation: Source small-signal transistors from authorized distributors. The cents premium per part avoid hours of debugging.
In the next article, I’ll dive deeper into these transistors’ high-frequency performance by measuring RF gain (specifically, the small-signal current gain |h21e| or power gain in a 50Ω system).
Using a simple common-emitter test fixture with a VNA, I’ll sweep frequencies from 1 MHz up to 100–500 MHz (depending on each part’s expected ft). This will reveal if the «well-behaved» DC hFE fakes (like the BC547 and 2N3904) maintain consistent amplification at RF, or if they roll off prematurely due to poor internal capacitance matching or die quality.
The PN2222’s wild DC spread almost guarantees poor RF consistency, but the data will confirm it. Stay tuned for plots showing gain vs. frequency, ft extraction, and a clear winner for RF hobby projects!