So quite an improvement... obviously audible.
So trust your ears and try both configurations and report back with which you feel is the best audibly. That way you are not taking anything on trust but determining the result yourself.
The output from the middle stage (ltp with mirror load) drives a Q. That Q drives an NPN output device attached to it's collector. The same Q drives another NPN output device from it's emitter.
These two output devices are connected by a the pair of output resistors and a diode, (as you have stated above).
That looks like a quasi using a pair of NPN output devices in push pull.
Could you explain your viewpoint in more detail?
It would be easier if the Qs were labeled.
It's maybe semantics but I've always seen a Quasi-Complementary Output as a EF2 on top and a CFP on bottom biased 3 Vbe's apart (one for the CFP PNP and two for the EF2), in order to avoid a power PNP and use a small signal PNP instead (which is the one you wanted to avoid).
I don't see any of this in the 5534, actually there is no PNP involved in the output stage (apart from auxiliary stuff like clamps and protection). Rather there's an EF2 on the bottom (thus common emitter stage) and that "helper circuit" on top of it, but it's tied to the ouput side of the EF.
I don't see any of this in the 5534, actually there is no PNP involved in the output stage (apart from auxiliary stuff like clamps and protection). Rather there's an EF2 on the bottom (thus common emitter stage) and that "helper circuit" on top of it, but it's tied to the ouput side of the EF.
Attachments
Last edited:
Is the Output (sink) allowing operation as a single ended amplifier using the Output (source) as the single active device?
To me it looks like the Driver is using the upper and lower devices in push pull, but I am doing no better than guessing.
To me it looks like the Driver is using the upper and lower devices in push pull, but I am doing no better than guessing.
agreed. two complementary drivers and a pair of push pull NPN outputs.
Here we have only one driver but the same two NPN outputs.
what do we call it?
How does it operate?
Is there some kind of sliding bias?
or is it actually push pull?
Yes, that's when the diode comes into the picture. Assume a high output current to be sourced. Drive to the sinking NPN is reduced until it's collector (plus shifter on top) has raised enough to turn on the sourcing NPN.
Hence, when you bias it to class-A neg.rail there is one more transistor (the EF) in the chain as compared to pos.rail.
The attachment of the driver collector seems confusing but I think (guessing here, too) the base of the source NPN is mainly moved by the collector of the sink NPN, not its driver (whose current is much lower). Sim is needed to check that... but the 5534 has been disassembled many times, I think there's a paragraph on it in D.Self's "Small Signal Design" (don't have it handy right now).
The 5534 has access to the three pins, 1, 5 & 8
That makes it almost unique in opamps.
There are many "tweaks" described that make the opamp do the next to impossible (for other opamps).
The 5532 cannot be tweaked like this.
Now to D.Self's tweak: pulling pin6 to +ve.
Pulling pin5 to +ve effectively disables the Vbe and disables the upper NPN.
That leaves the lower NPN as the sole output device. It is driven by the Driver (in your pic).
Can you describe exactly what happens here?
It would explain why the 5534 breaks the usual "rule" for quasi NPN output stages needing a sink to -ve.
That makes it almost unique in opamps.
There are many "tweaks" described that make the opamp do the next to impossible (for other opamps).
The 5532 cannot be tweaked like this.
Now to D.Self's tweak: pulling pin6 to +ve.
Pulling pin5 to +ve effectively disables the Vbe and disables the upper NPN.
That leaves the lower NPN as the sole output device. It is driven by the Driver (in your pic).
Can you describe exactly what happens here?
It would explain why the 5534 breaks the usual "rule" for quasi NPN output stages needing a sink to -ve.
Still no answer what value that resistor from pin 6 to pin 5 should be.
I thought i knew every trick that can be done with the NE5534.
I use it since it came out in the 70th.
I thought i knew every trick that can be done with the NE5534.
I use it since it came out in the 70th.
It would be nice to know the before and after THD results and link it to perception thresholds.
THx-RNMarsh
D. Self has that measurement in his book about analog crossovers.
It really improves the treble to a point where he got the residual of the Audio Precission.
It really improves the treble to a point where he got the residual of the Audio Precission.
If someone can recommemd the approximate resistor value to be used with LT1363 this will be highly appreciated as well.
Do try harder to keep up at the back! The resistor being talked about here is specific to the NE5534. The LT1363 is a C-load type opamp and definitely should not be loaded with a resistor as the RC series network transports the output load back to the VAS stage. A current source is called for.
I mean, simple Ohms law :
The voltage difference between Pin 5 and Pin 6 should be measured.
The current in that resistor is then Vdiff / R.
The resistor value is R = Vdiff / I desired.
The voltage difference between Pin 5 and Pin 6 should be measured.
The current in that resistor is then Vdiff / R.
The resistor value is R = Vdiff / I desired.
ebay hints means you want to buy a ready-made CCS? Or you want to buy discrete transistors there?
Of course, but only after someone who hasn't the measurement capability has made their preferred choice on listening impressions alone.
I suppose I could Class-A one pair with JFET and the other pair with transistors, then I can compare the difference.
I have no idea how to proceed, I'm not an EE, I'm just trying to improve sound quality. =)
Suggest googling for some CCS circuits - the simplest kind has only a single JFET or bipolar transistor. For example - Design of a Constant Current Source - delabs
page389 starts the section on biasing the output stage of a 5532.
It has not got an equivalent to pin5.
He shows biasing to the +ve rail. 5.4mA from a CCS seems to be his optimum and reduces measurable distortion from 1.8ppm to 1.5ppm
Noise of the generator/measurement system prevents seeing any significant differences below that level. The plot on page393 shows the noise minimum at a level of ~3.7ppm.
Self does discuss the 5534 later, but I can't see the same output bias experiments.
Last edited:
Tek SG505's inverting output buffer uses 68R, as well a clever compensation scheme and an added gain stage.
- Status
- Not open for further replies.