Discrete Audio Products for Audio Production and Live Recording
The basis of the entire Tonelux designs centers around our two new discrete op-amps, the TX-240 and the TX-260. Why did we call them that? Who knows.
In designing the TX-240
op-amp, I came across some very interesting things. The op-amp
sounded better without a power output stage, plus it required quite a
bit less internal compensation for stability. Of course, this
raises the problem of how to drive a transformer. This is where
the second op-amp came in. It is not actually an op-amp as such,
as it has no gain and is only a non-inverting power stage.
Starting with the TX-240
op-amp, I began with some parameters, such as overall open loop gain,
output drive and bias.
OVERALL OPEN LOOP GAIN
My feeling is that most of the op-amps designed today have more gain
than they really need. The increase in gain makes the distortion
lower, but in all applications, the op-amp never really sees more than
50dB of actual gain. The rest is fed back. These op-amp
designs can be in the range of 120 to 180 dB of gain. For the
TX-240, I chose 85 dB. The op-amp stays very stable, and requires
little internal compensation to keep it that way. This also seems
to add to the "open depth" that some engineers have said about the
Tonelux products. To me, too much fed back gain is like pressing
on the gas and the brakes at the same time.
OUTPUT DRIVE
For proper internal use and limited external use, I decided that being
able to drive 60 ohms was plenty. This is more than enough to
drive level pots, panners, solo/mix busses and sends. It will not
drive a transformer, but keeping the drive abilities more conservative,
I was able to use small signal transistors for the output stage instead
of larger power transistors. This has one big advantage.
The dies in larger output devices have an increased capacitance and are
generally slower. This forces the designer to make sure that the
input and gain stages that come before it are slowed as well, which
insures stability. Although this compensation is not in the audio
spectrum, it appears that the op-amp sounds more open with the smaller
devices and less internal compensation. Currently, the TX-240 has
only ONE capacitor for compensation. There are only SIX
transistors for the audio and one for the current source. This
simple design insures that the op-amp does little but amplify.
BIAS
We all know the terms CLASS A, CLASS AB, etc. The older Neve
op-amps were CLASS A biased and the 2520, 990 and variations are CLASS
AB biased. Class A biasing simply means that the output device(s)
are biased in such a way that the device swings the entire
waveform. This creates little if no distortion because there is
only one device and it is not switching from one device to another as
the signal passes from negative to positive. The down side of a
CLASS A op-amp or power amp is that when it is idle, it is drawing 50%
of its power just sitting there. It has to so it can swing
positive and negative. If it is powered by a single power supply,
it will also need a large output capacitor that can handle a constant
voltage on it.
CLASS B is a type bias where you use TWO devices, one for the positive
going part of the wave form and one for the negative going wave
form. The advantage of a CLASS B amplifier is that when idle, it
is drawing almost zero current, so it runs cool. The problem is
that because of the nature of a transistor junction, it has to have
around .6 volts before it conducts, which means that there is a dead
gap as the signal approaches zero and starts to go in the other
direction. This is called "crossover distortion". To
minimize this problem, the designer will "leak" a small amount of
current into each transistor, biasing them into the range of the other
transistor slightly, which causes both transistors to overlap slightly
and there is no longer a dead gap. This is called CLASS AB
because it is a B amplifier with a little A bias. There is still
some distortion, but most or all of it is eliminated with a lot of
negative feedback in the op-amp. It gets canceled.
What I have done in the
TX-240 is to bias the devices more into CLASS A than is usually
done. It is not 100% CLASS A and not 100% CLASS B. In
experimenting, I found that there is a point where the distortion
really doesn't go down any lower as the bias goes up, once you reach
that point, the rest is just current and heat. The TX-240's
distortion is less than .007% with the bias which I call Differential
CLASS AAB, since it also uses a pair of output devices so there is
minimal offset when it is idle and the output cap can be smaller and it
is biased harder into CLASS A than a normal CLASS AB amp.
THE OUTPUT DRIVER
STAGE TX-260
The TX-260 is the power output driver. The idea was
to design a discrete power stage that could power a transformer and all
the abuse that goes with it. It is a typical push-pull power
output stage, similar to most of the power op-amps out there.
With a pair of input transistors, the TX260 has a total of FOUR
transistors. The op-amp is biased the same way as the
TX-240, so the distortion is very low. Because there is no
gain, the slew rate is also blistering.
There are two advantages
with this design. Having the gain stage separate from the output
stage allows a level control to be put between the two amps for gain
control, like a fader, without having to put it across the transformer
secondary, which changes the tone as it is adjusted because the load is
changing. Another advantage is the power op-amp stage is not in
the feedback loop of the gain stage, so the reflected distortion of the
transformer does not become part of the feedback path. I have
found that the distortion of the transformer is much more consistent
from low level signals to high level signals. There is a good
amount of low end distortion, which I prefer to call "love", and it is
very even. The graphs under the EQ4P info page show the effect of
this. The distortion of the transformer is a very "familiar" kind
of distortion, with more stability.
The whole signal chain in
the Tonelux product line is not a copy of anything that is out there
now. Both of the op-amps were fresh, ground up designs based
around my theory of splitting them up and keeping the transformer out
of the gain loop, and keeping the slower power devices out of the
feedback loop also. This idea draws from much older designs,
where each stage was it's own and there was no overall loop that gets
fed back. Because of this, I have been able to design a very
clean circuit with all the harmonic tones coming from the transformer
in a very controlled manner. It all sounds very warm and very
consistent over all levels.
And of course, the sausage:
