Baxandall diode in quasi complementary design

Status
Not open for further replies.
I've seen some references online regarding the use of Schottky diodes as a Baxandall diode in quasi designs.
All were saying to use classical silicon diodes or transdiodes.
In my design Schottky diodes gave the best overall THD in LTSpice.
I've seen a reference like "Schottky diodes defeat the purpose" as Baxandall diodes.
It happens that I've read this after I've install a BAT41 Schottky diode in my quasi design.
So, should I change it to a normal Si diode or transdiode?
Also, why do Schottky diodes model best in LTSpice with a quasi output?
 
I had oscillations but I just had to add a 10uF capacitor across the gates of the drivers. That stabilized the amp. Drivers don't have any cap across the driver.
Although the original driver did have 500pF across bc
 
Ok. But shouldn't that be reflected in the result of the simulation? Like lower THD?
So far Schottky diodes gave the best results at 1W int 8 ohm output.
The appropriate part to use use is a transdiode as Baxandall recommended and wg_ski reminds us by saying the diode should have the same characteristics as the junction in the complementary circuit.

A transdiode is just a small transistor connected as a diode but I would not take a THD simulation alone as the best analysis tool. The sound quality or spread of harmonics is what you are after and looking simply at the lowest simulated THD with a fixed signal may not convey the whole truth.

Beware the temptation to believe that simulation programs are a magic bullet and every THD analysis with your program and its component models is going to be correct and appropriate, regardless of how precise the results appear to be.
 
You should be trying to place the same loading on the Emitters of the two complementary Drivers.

The added diode||capacitance||resistance on the PNP emitter is trying to replicate what the NPN Driver Emitter sees as it's load.

The output device has quite high capacitance, that needs a similar capacitor. The output transistor also has an effect that looks like a Power Diode. That too needs to be replicated. Neither a 1n4148, nor a Schottky can do that.
 
Last edited:
I am also looking at the FFT graph but is very similar between all diodes/transdiodes tested. As output volume increases the second harmonic increases and the third decreases starting with a higher 3rd at low level and ending with higher 2nd at higher levels. 4th and so on vary but little.
I don't actually see much change and I begin to suspect it is of little importance to the overall sound. As long as there's a diode it should be better than no diode.
What is the correct type of transistor used as a transdiode in this setup? I've seen both types recommended. Some say the same as output transistor, others the same as driver.
The only factor that seems to have a big impact on THD and harmonics is the driver emitter resistor. The larger it is the worst the output. I will revert to 39R stock.
 
Thank you all for the good info!
As I said, 2nd harmonics rise and 3rd harmonics fall as I increase the value of that resistor from 40R to 150R. Even with a bit higher THD I presume even order harmonics are to be desired to odd ones.
As I've seen recommended by Douglas Self in his schematic in the book the transdiode is made from a pnp transistor, like the driver.
Another interesting thing I saw in LTSpice is that the transistor model (part no) doesn't really matter for overall thd/harmonics distribution.
 
If H2>H3>H4>H5 in a monotonic fall, a la Jean Hiraga from the seventies, then the sound will be pretty good. Yes, evens are always better than odds, particularly from H4 and beyond. H3 is actually musical, if sharp. H5 higher odd orders confer a 'machine' sound, best to be avoided if possible.
The distortion profile is a function of output device, bias, drivers and operation parameters, and particularly the input and VAS stages. I agree, the transdiode doesn't seem to have much influence, and I use a 4001.

Ciao,

Hugh
 
Here are some screenshots of Flat Top window function FFT with different pnp driver emitter resistors at 1W output into 6ohm load at 15.5mA bias, with THD:
40R - 0.001819%
80R - 0.003466%
120R - 0.006514%
150R - 0.008897%
All transistor models I used in this output stage are Keantoken/Cordell models.
 

Attachments

  • 40R.png
    40R.png
    315.9 KB · Views: 318
  • 80R.png
    80R.png
    316.3 KB · Views: 298
  • 120R.png
    120R.png
    319.6 KB · Views: 298
  • 150R.png
    150R.png
    317.4 KB · Views: 305
And here's the same settings with 80R and 18W output.
THD holds nice at: 0.003502%
Also attached the circuit schematic.
 

Attachments

  • Screen Shot 2015-03-27 at 01.59.45.png
    Screen Shot 2015-03-27 at 01.59.45.png
    299.1 KB · Views: 400
  • Screen Shot 2015-03-27 at 02.00.41.png
    Screen Shot 2015-03-27 at 02.00.41.png
    399.3 KB · Views: 348
Yes, I messed up the connection of the transdiode :headbash:
Sadly when I connected it right the result is not that good as before. I get at most the same amplitude of 2nd and 3rd harmonics at about 70R. Lower or higher the 2nd lower and 3rd rise pretty much.
I used a PNP as I've seen in the book of Douglas Self. I attached a screenshot of the schematic he presented. I don't see a 0.5R resistor added in series, but maybe a more detailed implementation could include it. I'm not that knowledgeable as to assert anything.
 

Attachments

  • Screen Shot 2015-03-27 at 11.06.29.png
    Screen Shot 2015-03-27 at 11.06.29.png
    159.6 KB · Views: 325
Status
Not open for further replies.