Is the simulation still accurate with Q5 and Q6 in heavy saturation (as Scott Wurcer pointed out in the other thread)?
Hi dkfan9,
Thanks for pointing that out!. I missed seeing that😳.
There is about 20mV Vce on Q5, Q6. So Q5 and Q6 are just acting as 20 ohm resistors collector to emitters and base emitter as a diode (or is it two diodes BE//BC diodes?). So maybe not Blesser pairs after all in this arrangement and under normal intended bias conditions?
Here's the circuit with Q5, Q6 collectors open and 20 ohm resistors across previous CE nodes. Change the bias resistor slightly to return to 2.2mA and the Gm at the idle point is almost the same as before these changes to Q5,Q6 and throughout the input range -- confirming Q5,Q6 are not active C to E but just act as base-emitter diodes.
Interesting, so the bases of Q3 and Q4 are tied together by R7 and R8 so they are not acting independently as I previously assumed, but they are still current driven.
We now have a sliding bias circuit with just 4 active devices. Has it been done like this before?
Thanks for pointing that out!. I missed seeing that😳.
There is about 20mV Vce on Q5, Q6. So Q5 and Q6 are just acting as 20 ohm resistors collector to emitters and base emitter as a diode (or is it two diodes BE//BC diodes?). So maybe not Blesser pairs after all in this arrangement and under normal intended bias conditions?
Here's the circuit with Q5, Q6 collectors open and 20 ohm resistors across previous CE nodes. Change the bias resistor slightly to return to 2.2mA and the Gm at the idle point is almost the same as before these changes to Q5,Q6 and throughout the input range -- confirming Q5,Q6 are not active C to E but just act as base-emitter diodes.
Interesting, so the bases of Q3 and Q4 are tied together by R7 and R8 so they are not acting independently as I previously assumed, but they are still current driven.
We now have a sliding bias circuit with just 4 active devices. Has it been done like this before?
Last edited:
After more simplifications I arrive at the circuit below (which looks familiar!):
The 3 steps to get there are in the attachment. First, Q5 and Q6 base emitters can be shorted and I1 and I2 removed. Then with Q1 and Q2 shorted base to collectors can be replaced by 1N4448HWT with 2 in parallel to get the same bias point (this diode has an NF=1 like a trans-diode whereas the 1N4148 has an NF of 1.7). Then combine R7,R8 into one resistor of twice the value. The temp.co also needed to be increased with the removal of the transistors. But we still see the same nonlinear gain profile as the original dadod circuit.
So again, has it been done like this before? The key difference to a standard output stage seems to be to drive the two bias diodes at the midpoint AND the bias diodes need to be carefully chosen to suit the CCS's current to get a sliding bias effect. The added resistors R4 and R5 appear to be essential to stop glitches at the output peaks.
Interesting. How does input impedance compare to the original? Do you have a model for the diodes you could share?
Hi dkfan9,
Sorry for not including the 1N4448HWT diode model. The attached has it with a plot for input resistance.
The input resistance is 4k ohms falling to 1k at the peak - a 4:1 change same as the Gm change from idle to the peaks. same as the original to a few %.
So the diodes are doing the spreading. The input voltage range in open loop is about 4Vt hence a 4:1 change in the diodes 're' (dynamic resistance).
Since the input resistance falls to 1k at the peaks the driving source needs to be less than 500 ohms to get the monotonic rising Gm with input swing (to minimise the higher harmonics at higher output levels). In other words we want voltage drive at the input, same as for standard output stages.
BTW With current source drive to the input there is a gain peak and then fall (see below with 4k source R). The fall at the top end may be from Beta droop of the output transistors (but I haven't checked to confirm).
The simplified version with diodes is looking like the Blomley approach with diodes driven by current - except now I am recommending voltage drive for the diodes for less high order harmonics.
----------------------
This thread is about the Blesser input stage in Post 1, so the Wurcer/dadod version, which turns out to not be a Blesser, could/should be discussed further in a new thread? Comments? Suggestions?
Sorry for not including the 1N4448HWT diode model. The attached has it with a plot for input resistance.
The input resistance is 4k ohms falling to 1k at the peak - a 4:1 change same as the Gm change from idle to the peaks. same as the original to a few %.
So the diodes are doing the spreading. The input voltage range in open loop is about 4Vt hence a 4:1 change in the diodes 're' (dynamic resistance).
Since the input resistance falls to 1k at the peaks the driving source needs to be less than 500 ohms to get the monotonic rising Gm with input swing (to minimise the higher harmonics at higher output levels). In other words we want voltage drive at the input, same as for standard output stages.
BTW With current source drive to the input there is a gain peak and then fall (see below with 4k source R). The fall at the top end may be from Beta droop of the output transistors (but I haven't checked to confirm).
The simplified version with diodes is looking like the Blomley approach with diodes driven by current - except now I am recommending voltage drive for the diodes for less high order harmonics.
----------------------
This thread is about the Blesser input stage in Post 1, so the Wurcer/dadod version, which turns out to not be a Blesser, could/should be discussed further in a new thread? Comments? Suggestions?
Here's Post 1 sim, without caps:
Biased for optimum (60mA). Distortion at 10Vpk out is 0.05% THD.
Biased for optimum (60mA). Distortion at 10Vpk out is 0.05% THD.
Hi Ian,
Thanks for the model. I would be interested in that new thread. Seems like a useful biasing scheme for driving from an opamp.
As I understand high current spikes will always happen in output devices (and other devices nearby if local feedback is used). This is unavoidable I believe. But overall feedback can produce high current spikes in the input devices, mostly due to the fact that the overall gain of the amplifier is very low during the crossover situation. The error signal that the input devices have to process can then be extremely high, and high current spikes can happen. This can be verified probing current of input devices on Spice simulations.
As I understand the circuit, you are correct. In the circuit shown that happens for ac signals because of the capacitors connected to ground. The input impedance for dc signals is very low (common base) because the lack of local feedback.
Leaving the
Not sure I understand all the concepts, but when resistors R8/R7 are not connected to ground there is no local feedback and the input impedance is what corresponds to a common base connection.
Holimar. Transistor were not that expensive compared with similar bipolar devices (in terms of power). Anyway 80% of the cost was assigned to chassis, transformer and filter capacitors.
Not sure if it is of any interest but my design of the input stage derives from the uA741 operational amplifier design. The npn/pnp transistor compound used on that design makes possible to have complementary outputs eliminating the need of a double differential pair as used in some audio amplifiers to make the circuit symmetric (also my own goal) . From that circuit I started to eliminate superfluous devices not needed for an audio amplifier frontend, and ended up with the final schematic shown in my original post.
Are you sure it was the 741? I'm struggling to see any similarity there. Looks like a fairly standard differential input stage.
Not really.
Post #1 shows +/-50V PSU (not +/-15V) and 3 pairs of output MOSFETs (not 1 pair),
It does make a difference, doesn't it ?
Ah BUENO!!! 🙂 🙂 🙂
Palabras mayores!!!!!
Felicitaciones!!!
......................
Holimar?
World class!!!!
Best of Argentine Audio for decades 🙂 🙂 🙂
Congratulations 🙂
Hi juma,
My sincere apology. I forgot the PS volts and 3 pairs! 😳. Here's an update. "m=3" means 3 in parallel.
Yes, a difference: THD 0.006% at 1W and mostly 3rd harmonic👍 and 0.023% at 95W.
Idle current trimmed for about 45mA each pair. I also changed the lateral model to ones with subthreshold conduction.
Hi tucura,
Re Post 31 and R7,R8. That refers to the dadod circuit which is now known to be not a Blesser arrangement. What we need to look at are the same resistors in the "holi-org" circuit of Post 1 circuit - coincidentally the same resistor numbers for the circuit above (Post 37).
R7,R8 "holi-org" circuit act like emitter degeneration resistors in the input stage of a CFA. The mid node of R7,R8 is the equivalent of the inverting input of a CFA. In the circuit above the mid node of R7,R8 I added R18 to common to monitor the current to common. This resistor corresponds to the resistor for a CFA feedback network but it is not used this way for this amp. Instead, the feedback is applied to the emitter resistors equally with two feedback resistors of twice value to one Rf to R18. This trick gives the same feedback effect as a CFA but with a little more feedback depth than the usual CFA method.
Re Post 31 and R7,R8. That refers to the dadod circuit which is now known to be not a Blesser arrangement. What we need to look at are the same resistors in the "holi-org" circuit of Post 1 circuit - coincidentally the same resistor numbers for the circuit above (Post 37).
R7,R8 "holi-org" circuit act like emitter degeneration resistors in the input stage of a CFA. The mid node of R7,R8 is the equivalent of the inverting input of a CFA. In the circuit above the mid node of R7,R8 I added R18 to common to monitor the current to common. This resistor corresponds to the resistor for a CFA feedback network but it is not used this way for this amp. Instead, the feedback is applied to the emitter resistors equally with two feedback resistors of twice value to one Rf to R18. This trick gives the same feedback effect as a CFA but with a little more feedback depth than the usual CFA method.