The blameless Audio
Buffer
or a perfect headphone amplifier
A story about dancing with single electrons.....
APN 001
Date: 2010-04-10
Link to shematic (Buffer)
Link to shematic (Pre-LowPass)
Link to shematic (System/Grounding)
Link to measurements
part I
Link to measurements part II
Link to overcoming measuring limitations
Whats to do anymore:
- pictures of the breadboards
- deeper measurement (THD/THD+N)
- THD vs. freq
1. Long journey to find
the right.....
A long time ago, I started looking for a ultimate Headphone amplifier.
But all circuits I found were tradeoffs between distortion, noise,
complexity and cost.
Some weeks ago, I found the design idea of this buffer.........
First, I hacked it into a simulation programm.... and was impressed
about it's performance and that it would cost almost nothing.......
I started designing the stage so that a first prototype could be built
on a breadboard...... because that's the difference in real world and a
simulation... it's not the same !
2. Prototypes
The prototype was build up on a grey Tuesday morning in almost 1/2 hour
in my home-lab from components I found there (e.g. NE5532).....
First I set the current up to only 25mA ( not to destroy components by
too much current in the first step)
In the evening, I started measuring the prototype at the lab at work...
after the first results, I take into account, that the buffer has a
performance worth measuring into the deep wich means measuring things
right and not fast..... and that are the results I got......
4. Simulate again
After this results, I wanted to know wich was the limiting factor is
the circuit. After replacing the Opamps in the simulation it became
clear that the Opamp is the limiting factor in noise and distortion
domain... so a NE5532 is good, but not good enough to get best
results....
So there are some nice Opamps you could use.. some are best for Noise
but not for distortion, others for distortion but not noise and
some are a tradeoff......
|
Noise
|
THD
|
comment
|
Channels
|
Price
|
NE5532
|
122 dB max @ 1V
|
not more than 110dB
|
cheap
|
2
|
cheap
|
MC33078
|
120 dB max @ 1V |
not more than 115 dB
|
cheap
|
2 / 4
|
cheap
|
OPA134
|
120 dB max @ 1V
|
110 dB possible
|
THD good, Noise "bad"
|
1 / 2 / 4
|
affordable
|
OPA627
|
119 dB max @ 1V
|
130 dB possible (???)
|
too expensive and noisy
|
1
|
very expensive
|
OPA227
|
|
|
Noise good, THD ?
|
|
|
OPA211
|
|
|
|
1 / 2
|
|
OPA827
|
|
|
|
1
|
|
OPA1611
|
|
|
|
1 / 2
|
|
Opamp table (diff. input, 1Vout in 27 ohms)
3. More current
Next step was increasing the output current by reducing R1/R2 to 7 Ohms
and to mount some heatsinks on T3/T4 beacuse of Powerloss would be
approx 1W per transistor......
4. Adding a differential input
To get the best performance out of the buffer you must be able to get
it in (that sounds worried but that's it !).
With the differential input, you have a chance to make a single GND
point in your circuit and isolate GND noise from outside the buffer
circuit.
Keep in mind, that 120dB signal-2-noise ratio at 1V output level means
that the noise into the buffer must be lower than 1µV !
( And thats why the article is calles dancing with electrons)
On the other hand, the additional resistors make noise measurement
worse (about 3-5dB). So keep them low or remove them if not needed......
5. Measuring until the instrument
gives up
If you look at the results (here), you will see that you can reach the
measurements system limitations (if using the OPA2211 e.g.) .....
Opamp
|
Noise (2Vref)
|
THD+N
10mV/27Ohms |
THD+N
100mV/27Ohms
|
THD+N
1V/27Ohms |
THD+N
2V/27Ohms |
THD+N
3V/27Ohms |
NE5532
|
122,5 dB
|
68 dB / 0.04 %
|
87,5 dB / 0.0042 %
|
106 dB / 0.0005 %
|
106 dB / 0.0005 % |
105 dB / 0.00056 % |
OPA2134
|
119,5 dB
|
65 dB / 0.056 %
|
86 dB / 0.005 %
|
105 dB / 0.00056 % |
107,5 dB / 0.00042 % |
105 dB / 0.00056 % |
| OPA2211 |
125,0 dB |
71 dB / 0.028 %
|
92 dB / 0.0025 %
|
110 dB / 0.00032 %
|
113 dB / 0.00022 % |
112,5 dB / 0.00023 % |
AP-SYS2
|
> 140 dB
|
?
|
102 dB / 0.00079 %
|
114 dB / 0.0002 %
|
115 dB / 0.00018 % |
115 dB / 0.00018 % |
Also I had problems with the first breadboard version, therefore I had
to build a new one.....
6. Overcoming measurements limitations
Seeing point 5, I had to add a low pass filter to supress k2 and k3.
But this only makes sense with OPamps capable of generating lower THD
than the oscillator.....
Using OPA2211 should give the answer....
7. Perspective......
So long, here the project stops.... next step will be to build a better
oscillator (and even a complete system that's able to measure soooo
deep)
In the consumer/users point of view, I must say that the results are
much better than needed (and expected). I think a SNR of ~120 dB (or
even 100 dB) and THD+N ~ 90 dB should be sufficient for an headamp
amplifier. Even there's no source I know wich would deliver lower THD
than ~100 dB (Live concert with Headamp direct connected to
mixer perhaps ?) in home use (ok, my DAC has ~105 dB, but this is MINE and will not be borrowed :-)