The Blomley Class B amplifier


2010-11-18 9:12 am
Here's not the case as TR2 act on the 2n3904/06
bases before TR3-Tr6 conveyor act on the emitters so the speed of switching is determined by the Tr3-Tr6 delay mainly and they are clearly slower than 2n3904/06. That sequence is deliberate to forcefully discharge the base of the trz that needs to shut off, the circuit doesn't rely only on the underbias for that matter .As with the capacitances involved Low cob trz are usually low on beta too because they are mainly destined for common base common base connection miller's capacitance is just a fraction of cob as the voltage gain is below 1X so cob or rather ccb in this case is irrelevant .
I remember that the
UHF-VHF amplifiers I was building in childhood to catch the russian's channels barely had 10db gain per 1 trz stage made with some BFR90...91 trz used in common emitter connection so its CCE would be important while multiplied with the stage voltage gain which was like 1/5 of the dc current gain thus 10 db gain was kinda big deal at 800Mhz .
BF422/423 would be a more natural choice than BFR trz which I'm not sure how stable can be in audio while bf422/3 definitely work in audio, but I'm pretty sure VHF trz can easily be toast due to very low CE max voltages common to these types of transistors.Unless you're going to use high power RF trz you won't get big VCE ranges...and big power trz are usually used in common base connection or cascoded "two on a die" to reduce the influence of their inherently higher cob ...
I modelled this amplifier using 2N3904 and 2N3906 transistors. In the original Blomley design the pairing was 2N3904 and 2N3905. There are some subtle differences in the specifications for the latter two devices. 2N3905 has lower minimum hFE than 2N3906 and lower fT. T on is the same for both but T off for 2N3905 is less than for 2N3906. If it takes marginally longer for this to turn of it will take marginally longer for it to turn on again. One might create a .model for 2N3905 and substitute the difference using 2N3906 as a template for the common performance parameters and run a new simulation, however Mouser have listed a 2N3905 made by Central Semiconductor that could be worth trying.

Since the choice has to align with 2N3904 which has a low hFE the datasheet for this would need to be checked to see which is the best match out of 2N3905 and 2N3906 - see attached.


  • 2N3905.PDF
    384.2 KB · Views: 40
how do i properly connect the source? bypass the input bandpass filter? and what is the setting of voltage sources.

also is the variable f=10k the freq run?

You connect it same as you would a sine wave voltage. Filtering depends on what you are testing. Filtering is there to prevent a square wave from causing slew limiting so if you want to find the slew limit then yes, no filter. tr is the rise time which has to be >0. Vp is the peak voltage same as the spice sine wave spec. f is the frequency. The PWL (piece wise linear) formula is a series of line segments defined by the rise time, voltage and frequency. The problem with a normal pulse voltage is that it only has two states, the positive and negative peaks, so the initial simulation conditions are based on the negative peak and not zero. This creates a DC settling problem.
  • Like
Reactions: 1 user
I made a quick run f=20k vp=0.6 (max input amplitude to avoid clipping)..well it is quick indeed. The resulting graph looks fine I think.
BTW the psu rail voltage sources is set to zero amplitude. Is this the correct setting?

Thanks again!

You can test for PSRR (Power Supply Rejection Ratio) by adding a non-zero amplitude and frequency etc to the DC rail voltages. Typically, that would be 100 or 120Hz as from a full wave rectifier PS, but you may be interested in channel crosstalk at other frequencies. You can also simulate a basic power supply with an AC voltage source + rectifiers and bulk storage cap. This can be very interesting simulating turn on transients and provides a more realistic supply noise. Note that a floating AC voltage can lead to simulation problems, not a problem if simulating a grounded center tapped transformer using two AC voltages.
Last edited:
  • Like
Reactions: 1 user
Steveu Sir,
could you please share an .asc (spice file) for the PSRR method?


Sorry OP out of topic here.
I don't do this often, but I would do it if I was going to build the circuit. So, I just grabbed an old schematic and added a "stimulus voltage" to the DC power supply and ran the AC analysis. Note that 24VDC power supply is also the AC analysis stimulus. Then plotting the output, we see how much of the power supply noise appears on the output, as a function of frequency, the PSRR.


  • junkbox psrr.png
    junkbox psrr.png
    35.3 KB · Views: 35
  • junkbox psrr.asc
    5.7 KB · Views: 7
  • Like
Reactions: 1 user