Yes, but with 1,5mA quiescent in discrete stage I know for sure, that it will work in class B. And the only way to supress distortions is OLG in LT1363 , a lot of NFB.. And it is the same without discrete stage. Nothing gained with such underbiased discrete output stage (I bias 1,5mA) working with full swing in 600ohm load. Measured perfomance will be the same, I bet sure undistinguishable in blind test.
And that squarely returns us to the essence of thread. Something may measure better and look better on paper, but may still sound inferior to soething else that may not look as attarctive as a circuit.
But it works both ways. Maths may tell us 40 mA of bias current on a headphone amp is more than enough. Yet doubling that, even if it moves outside the limits of reasonable, may produce better sonic results.
When Thorsten told me to double the bias current, I admit I was sceptical, but I did it nevertheless. And true enough, he was right, the difference was not stunning, no light of day, but it is deifintely there, well in the domain of "unreasonably" high bias.
Because of effects like this, I believe we do not yet have measurement standards good enough to define the whole thing. With everything we've got, it's still our ears which must put the full stop with anything.
I ask myself: what's the point of adding a discrete buffer which operates in class B on a level below the op amp it is supposed to be supporting and boosting?
If I go through the trouble of making it work, then surely I want more from it than I get from the op amp on its own. My starting point would be to make it work in pure class A.
But that's just me, the weirdo who uses 50W devices to drive headphones. I only wish I could pass it on to you for an audition, a nice fully complementary, servo controlled, discrete amplifier.
The purpose of a line driver is not, omho, the same than an headphone amp.
It is just about never running out of current and stay stable whatever the load (very long cables in PA, 600 Ohms charge impedance).
The interest to add-it to an existing OPA is to try to keep the 'character' of this OPA (that you can like) when it is not able to satisfy those requirements. And keep the things simple. Discrete or IC, what the hell when it works?
For an headphone amp, i should prefer a low HD and fast class A amp, designed to feed a 25 Ohms load, with enough level to break a 600 Ohm headphone and kill his owner's ears.
I, of course, agree with you, bvv: "In audio, 50% (or more) of the sound is in the current" (on demand).
It is just about never running out of current and stay stable whatever the load (very long cables in PA, 600 Ohms charge impedance).
The interest to add-it to an existing OPA is to try to keep the 'character' of this OPA (that you can like) when it is not able to satisfy those requirements. And keep the things simple. Discrete or IC, what the hell when it works?
For an headphone amp, i should prefer a low HD and fast class A amp, designed to feed a 25 Ohms load, with enough level to break a 600 Ohm headphone and kill his owner's ears.
I, of course, agree with you, bvv: "In audio, 50% (or more) of the sound is in the current" (on demand).
Last edited:
BTW, those two Schottky diodes, as drawn, will do squat about the output bias. Add them in the drivers emitters and we can talk about.
EDIT: I mean, the upper diode is in the wrong place. C2 and C3, at 0.1uF, have absolutely no impact on a forward biased Schottky impedance or noise.
Last edited:
Supporting bad engineering decisions with subjective, uncontrolled, listening arguments.
The purpose of a line driver is not, omho, the same than an headphone amp.
It is just about never running out of current and stay stable whatever the load (very long cables in PA, 600 Ohms charge impedance).
The interest to add-it to an existing OPA is to try to keep the 'character' of this OPA (that you can like) when it is not able to satisfy those requirements. And keep the things simple. Discrete or IC, what the hell when it works?
For an headphone amp, i should prefer a low HD and fast class A amp, designed to feed a 25 Ohms load, with enough level to break a 600 Ohm headphone and kill his owner's ears.
I, of course, agree with you, bvv: "In audio, 50% (or more) of the sound is in the current" (on demand).
I would still consider a BUF634 set to 30mA quiescent for 12V RMS in 600 ohms class A.
Or an AD815 with 400mA per side Iout, at 15mA quiescent per side.
Jan
This would be extremely difficult, as even with Rbias = 0 idle current is 15mA.
Attachments
Forced in class A through the output ? (or flooding it with liquid nitrogen, should be able to make 20mA at -200C)
Last edited:
Look guys, what is important is that you run as high quiescent current as practical.
I think that 1.5ma, although most likely an improvement over what TYPICAL IC's have, is on the low side. 15ma is better. However, when I design a headphone amp, I use 100ma Iq.
My colleague Ron Quan just presented a paper (AES 9197) that denotes the crossover distortion in many popular IC op amps. This is because most the IC's are virtually starved in the output stage, in order to keep within the current spec. demanded by 'management', sales pressure, etc. This is the main source of any higher order distortions, AND even a serious contributor to PIM.
I think that 1.5ma, although most likely an improvement over what TYPICAL IC's have, is on the low side. 15ma is better. However, when I design a headphone amp, I use 100ma Iq.
My colleague Ron Quan just presented a paper (AES 9197) that denotes the crossover distortion in many popular IC op amps. This is because most the IC's are virtually starved in the output stage, in order to keep within the current spec. demanded by 'management', sales pressure, etc. This is the main source of any higher order distortions, AND even a serious contributor to PIM.
Nobody mentioned the Allison topology: it is true class A, and is vastly superior to most alternatives for a wide range of applications, from line driver to headphones, or even power amplifiers.
It has the same transistor-count as a diamond buffer, but performances are like night and day.
Circuit and application example here: http://www.diyaudio.com/forums/analog-line-level/224128-volume-control-evil-mad-scientist-style.html
It has the same transistor-count as a diamond buffer, but performances are like night and day.
Circuit and application example here: http://www.diyaudio.com/forums/analog-line-level/224128-volume-control-evil-mad-scientist-style.html
The purpose of a line driver is not, omho, the same than an headphone amp.
It is just about never running out of current and stay stable whatever the load (very long cables in PA, 600 Ohms charge impedance).
The interest to add-it to an existing OPA is to try to keep the 'character' of this OPA (that you can like) when it is not able to satisfy those requirements. And keep the things simple. Discrete or IC, what the hell when it works?
For an headphone amp, i should prefer a low HD and fast class A amp, designed to feed a 25 Ohms load, with enough level to break a 600 Ohm headphone and kill his owner's ears.
I, of course, agree with you, bvv: "In audio, 50% (or more) of the sound is in the current" (on demand).
Of course on demand, pure class A amplifiers of any reasonable power output soon become very big and very impractical. Typically, assuming 10% efficiency, one would be burning like 2.2 kW of power per hour for a nominally 100W/8 Ohms class A amp. And again as much in summer for the high power air conditioner, to prevent overheating.
That's why I always say "high current capability". I cater for it by calculating my overcurrent protection circuits for the output stage 50 mS current delivery specs with a little less delay, say for 45 mS activation time. Combined with my output stages, typically 4 pairs per side, each rated at 15A continuous, 30 A impulse, 230V, 200W, that takes care of all dynamic poudspeakers I have ever encountered under MUSIC conditions (as opposed to pure sine wave continuous on the test bench - do that with music and you'll be clipping in no time).
The only catch here is the time interval of that current capability. IEC standards assume 20 mS pulses, while I feel that's too short to be truly useful. I know that some manufacturers (e.g. Harman/Kardon, Electrocompaniet, etc) publish out of this world specs, quoting even hundreds of Amps of current. The only way I could ever fit that into my framework of real life is to look at 1 mS impulses, which is ridiculous in my view. That way, I could say my amp will deliver 120 A of current, at 30 A impulse of 1 mS. Instead, I prefer to say it delivers 28 A during 50 mS, and on a continuous basis, it will do something like 12 Amps for a minute or two, until the heat sink overheats and the overtemperature protection circuit fires (thershold is 65 deg. C).
- Status
- Not open for further replies.