Perhaps a possible hint from the results of my recent traces:
Tested parts were 2SA933 'R' and 2SA1037 'Q'. The A1037's were purchased later than the A933's (probably a period of three or four years?). These parts are actually the same die in different packages (the older data-sheets actually show each package version alongside one another).
I found that the beta*early multiple between the two versions to be close at about 21,000. Of course to be true, the lower gain graded A1037 Q also produced higher Vaf (about 89 vs 56).
Not as if this is a huge sample, but certainly raises the supposition that the beta*Vaf multiple may be relatively constant for a given process.
I know this is old but I didn't see anyone else pointing it out. The SPICE model you show is for the BC550C, not the BC560C. And the Fairchild (Now Onsemi) BC550C performance is verified in the performance of my Kmultipliers and in other tests I've done. Onsemi BC550C (replaced by Fairchild parts when they were acquired) on the other hand is no good for Early voltage. So Bob is not who's making a mistake here.
BTW, Early voltage is not constant. It changes with Vce. It usually has a peak somewhere in the middle of the Vce sweep and then falls again as Vce approaches the max for the transistor. Samuel Groner's comments on Doug Self's book shows this quite clearly in the open loop plots when just the VAS transistor is changed. In the plot I randomly stopped at the difference between the VAF peak and the VAF at the ends is about 4db. That is about a 60% difference. So your measurements could be significantly different simply because you had a different measurement method. One could instead of using a trivial number, plot VAF over the Vce range with interesting results. Groner's tests are here:
http://www.nanovolt.ch/resources/power_amplifiers/pdf/audio_power_amp_design_comments.pdf
Last edited:
Beta and Early voltage are highly variable, both depend on the effective base width which
is measured in microns and subject to process variation. Early voltage is also related to the device voltage as that affects the doping profile.
The slope section of the response is not totally flat, BTW, it depends on the doping profile and so estimating the intercept as the "Early voltage" is just a crude approximation. In devices where punch-through happens before avalanche-breakdown of the base-collector junction you'll see the current curve up and then go asymptotic as punch-through is reached.
Some devices achieve avalanche breakdown before punch-through and some don't, again depends on the doping profile.
is measured in microns and subject to process variation. Early voltage is also related to the device voltage as that affects the doping profile.
The slope section of the response is not totally flat, BTW, it depends on the doping profile and so estimating the intercept as the "Early voltage" is just a crude approximation. In devices where punch-through happens before avalanche-breakdown of the base-collector junction you'll see the current curve up and then go asymptotic as punch-through is reached.
Some devices achieve avalanche breakdown before punch-through and some don't, again depends on the doping profile.
Please feel free to put your own collection of PNP transistors through your own set of test procedures and measure whatever data interests you. The work I did here and the data I measured & presented, was more than sufficient for my own needs. Me, I'm content. Perhaps other DIYers have other requirements; that would not be too surprising. To get the data you seek, measured exactly the way you want it measured, doing it yourself ("DIY") is a guaranteed pathway to satisfaction.
Perhaps that means the ZTX757 is even better than the attached curve. Which would be nice.
Zetex specifies Theta-JA = 175 deg C per watt for the "E-Line" TO-92 thanks to its use of zippety doo-dah silicone epoxy. That Theta-JA is 12% lower than Fairchild's spec of Theta-JA = 200 deg C per watt for their regular old TO-92.
_
Zetex specifies Theta-JA = 175 deg C per watt for the "E-Line" TO-92 thanks to its use of zippety doo-dah silicone epoxy. That Theta-JA is 12% lower than Fairchild's spec of Theta-JA = 200 deg C per watt for their regular old TO-92.
_
Attachments
early voltage Vs max Vce rating
early voltage Vs max Vce rating
im not sure about early voltage rating being higher than a devices max Vce rating and how this effects the choice of a device.
i want to use 2N5401/5551 for VAS duties at 2mA and sims indicate the voltage across the device will vary from approx 40v to 80v.
early voltage Vs max Vce rating
im not sure about early voltage rating being higher than a devices max Vce rating and how this effects the choice of a device.
i want to use 2N5401/5551 for VAS duties at 2mA and sims indicate the voltage across the device will vary from approx 40v to 80v.
Most SPICE-beginners here on diyAudio, trust and rely upon the modeling parameters published by Cordell Audio here. They do this for two reasons:
I recommend you look up the Cordell Audio SPICE model of the 2N5551 discrete NPN transistor. You will discover that its (simulated) Early voltage is a factor of 3.7 times larger than the transistor's datasheet maximum Vce rating. If this makes you uncomfortable, you could ask Bob Cordell about it.
1. Bob Cordell is a well respected author (his encyclopedic and authoritative audio power amp book is available on Amazon), and SPICE-beginners are more comfortable trusting a known guy with a big name.
2. Everybody else uses the Cordell models, thousands of diyAudio members, so how terrible can those models really be?
2. Everybody else uses the Cordell models, thousands of diyAudio members, so how terrible can those models really be?
I recommend you look up the Cordell Audio SPICE model of the 2N5551 discrete NPN transistor. You will discover that its (simulated) Early voltage is a factor of 3.7 times larger than the transistor's datasheet maximum Vce rating. If this makes you uncomfortable, you could ask Bob Cordell about it.
