Interesting output stage idea

[keantoken]

I was looking at a book and messing around on my sim and managed to throw together an interesting idea. The biasing mechanism is Q3/4, and there is a feedback path connected between Q3/4. R5 is adjusted to put output closest to 0DC at no input, and C6 acts as a highpass filter to increase NFB at high frequencies to cancel out parasitic oscillations and such. This is pretty much designed as a variation of the SymAsym output stage, as it is the only amp that I have looked at for a very long time . This is a class A output stage on my sim, but my sim seems to be a bit weird lately so you may not exactly get 20mA peak output current on your sim.

I know that injecting feedback between two diodes in an amp is pretty much like suicide, because of the natural unlinearity of diodes, but since I am still not up to date with all this technostuff I decided to get a second opinion. Who knows, maybe the nonlinearities cancel out each other when they are combined in the output track. It seemed not to give too bad results on the FFT, though. If you want to know what my FFT looked like, don't be surprised if it doesn't make sense as LTSpice seems to have problems with generating suitable sine wave input (on my system at least).

Here's my schemea:

[Wavebourn]

Symmetrical Darlingtones, with OE, biased using Zeners in feedback, work in A class, with parallel feedback by voltage. Very critical to power supply, since output stage bias current is the function of power supply voltage minus break-down voltages of reverse-biased base-emitter junctiuons of Q3, Q4. I already mentioned today break-down of reverse-biased base-emitter junction voltages that are nos specified by manufacturers, but for DIY or in monolytic IC it should work well.
However, distortions will be very low, especially on low output power that is especially good.

[lineup]

hello again!

so, what is the voltage cross R1 and R2
At 5+5 volt I would imagine it to be like 4.2-4.3 volt.
This means the idle current in output is 4.3/22 ~ 0.195 A = 195 mA

The voltage gain is maybe more or less like nothing.
Rather less than -1.
Have you measured putting 0.5V RMS at input
and see what the output will be into 8 Ohm?

lineup

[keantoken]

As far as I know there are two main uses of feedback:

1: compare against input and apply result to correction point to correct nonlinearity.

2: accurately tune important currents in a circuit to give accurate output.

I was meaning the feedback in this circuit to be #2, but does it also function as #1?

In theory, using the transistors as diodes will more accurately bias the important transistors, which is why I used this arrangement. Do you think that this output stage is suitable for a HIFI amp? You seem to be telling me that my idea will work well. Thanks!

Here is the FFT I took:

[Wavebourn]

I forgot to mention that I supposed resistors to be connected to opposite rails, in such case it will work well, if powered from voltages more than Vbe-reverse-breakdown+2+Vbe-forward+drop on emitter resistors, otherwise Darlingtons will be always nearly saturated.
Also, load resistance must be at least 10 times more than 8 Ohm.

[keantoken]

Lineup!
Welcome back. Seems you were missing for a while!

The voltage across the resistors is 4.2V-4.3V (exactly as you guessed!).

Anything above .3V on input gives massive distortion (may be a slight problem).

Humor: I am such a goof . I accidentally unplugged my computer right after posting this. (for real)

Here is the output at .5V input:
My toaster would probably sound better than this.

[Wavebourn]

Originally posted by keantoken
As far as I know there are two main uses of feedback:

1: compare against input and apply result to correction point to correct nonlinearity.

2: accurately tune important currents in a circuit to give accurate output.

I was meaning the feedback in this circuit to be #2, but does it also function as #1?

In theory, using the transistors as diodes will more accurately bias the important transistors, which is why I used this arrangement. Do you think that this output stage is suitable for a HIFI amp? You seem to be telling me that my idea will work well. Thanks!

Here is the FFT I took:

As I said before, if to connect resistors to opposite rails it will be one idea, if to leave them as is it is another idea.
In first case your "diode-transistors" work as Zeners and the amp is biased to give high output swing. However, for high output swing your feedback resistors (resistanc of source and feedback resistor) must be selected for voltage gain at least 5.

If to leave everything as is as lineup mentioned output transistors will be close to saturation and maximal output voltage swing will be very low.

But I like the idea of symmetrical common emitter stage with deep NFB by voltage (your amp has 2 feedbacks, by voltage and by current).

If to develop it a little bit it will be a very good amp.

[keantoken]

Increased load to 100 ohms... Or was that just to help prove a point?

What values would you suggest for those resistors?
edit: Decided to do some math (sorry...). Would 10 ohms work? ~3V across R1/2. With these values, current into an 8 ohm load goes to 15mA peak with almost no 2nd harmonics whatsoever.

I have discovered a flaw with this arrangement: The base currents of the output transistors aren't defined as much by the output resistance, so increasing the load will actually cause overdriving. Am I right?

[lineup]

Originally posted by keantoken
Lineup!
Welcome back. Seems you were missing for a while!

Missing, maybe ...... maybe not

People without ideas, will not contribute to much audio improvements.
People with new ideas, will not contribute to anything really better than known ways.
At least Not very often. That's a fact.


Lineup Estimation.
Out of 50 new audio ideas a few people get:
--------
25 ideas will not work at all!
15 will work, but not as good as best known stuff.
04 will be working something like best known before
01 IDEA will be an improvement in some way
--------


lineup - a man of new ideas .. but not many hits, so far

[keantoken]

Ah... Yes... You are somewhat deceptive as the grin in your avatar forshadows...

agent_keantoken signing out. (got to go to bed) Bye.