|
< <
BACK TO DIY MAIN PAGE
Time To Get
Theoretical
I know some this will be a little heavy going for some, but don't
be put off, the conclusions are interesting.
There
are still quite a few things that have influenced the development of this
project, both the commercial JLTi and these two DIY versions – primarily the
Tube version.
Some may ask, what is the
differences between the JLTi and the DIY (Tubed) version?
For a start, the Tube Buffer in
the JLTi is much more enhanced, the so-called SLCF enhancement. For a
discussion of the SLCF and other good things, may I again refer you to my
on-line essay:
Tubes & the Gainclones
Also this enhanced Tube Buffer
needs higher voltages to work comfortably. Thus it is plus and minus 55V. It
is also regulated, using an adapted version of the SuperReg, developed by the
Vacuum State Electronics team headed by Allen Wright.
If you wish to regulate your
Tube, then by all means go ahead. I have stuck to simplicity and if you want
to go further, you have my blessing.
Let it rip...
Gazing Into The Dark
What follows may be theory and
attempting to gaze into the dark. Nevertheless I hope it Is of interest.
First Subject: Lynn Olson’s “The First Watt”
For a couple of years the
thoughts and writings of Lynn Olson has intrigued me, especially as they
agree with much of my own, indeed I wrote a series of articles for the
Audiophile Society of New South Wales in the 90's that turned out to
parallel many of his views. In March and April of 2000 Olson wrote two articles
in Glass Audio magazine. If you can acquire them, please do read them. I
have a PDF file containing both of them, but for Copyright reasons I don’t
feel right posting them on one of my web sites or that of
www.vacuumstate.com. Besides being 3.5MB in size it would make
for costly traffic costs. But if you have the opportunity to read these
articles, don’t miss out.
Lynn Olson has a web site:
http://www.aloha-audio.com
But I digress. From his site I
have borrowed this clip showing a distortion measurement of his Amity
amplifier:
Caption:
Amity PP Tube Amp. This is a 1Khz sine wave, 1.6W into 8 Ohm load.
The vertical scale is
10dB/division and lateral 1KHz/division. From this we can see that the 2nd
harmonic is-70dB (0.03%) and the 3rd –65dB (0.05%). This is quite
respectable, but what is noteworthy is the absence of higher-order
distortion. There is just a hint of 5th harmonic at –82dB
(0.008%). (The rest is either the noise floor or the limitation of
measurement). This is commendably achieved without feedback. Compare this to
more modern Push-Pull Ultra-Linear Tube amps (look at his web site as he
provides sample measurements) are much more elevated in higher order
distortions.
This is an important issue
because, in general, as the order of harmonic distortion increases,
it becomes more evident and displeasing to our ears. It's not the actual THD
figure that is important but rather how it is spectrally distributed. For
example, if 1% of 2nd
harmonic distortion is compared to an amplifier producing
1% of 7th harmonic distortion, both amplifiers
will measure the same THD reading - but will those amplifiers sound
the same? No way! They will sound very different. In fact 1% of 2nd
harmonic may well seem to improve the sound BUT
same amount of 7th harmonic
will sound very unnatural. So the spectral distribution is paramount and
especially avoiding high odd order harmonic distortion.
Another key to
understanding THD
is that while global feedback may well reduce the complete THD figure, of
what benefit is that if, in the light of above, the higher order distortions
become elevated?
Let's get back to Lynn
Olson, what Olson proposes is:
THE FIRST WATT is the MOST IMPORTANT WATT.
It kinda makes basic sense. After all
this is the part of the amp we are listening to virtually ALL of the time. Higher
power levels are more transient and shorter in duration.
Olson applies this to Tube Amps,
but what about Solid State or Hybrid Power Amps or Gainclones.... ?
Solid State do behave differently
but feedback still tends to reduce overall THD, but at the expense of higher orders being emphasized?
There is another difference; at 1
Watt virtually all Tube amps are still in Class A, whereas Solid State will likely be beyond its Class A range and switching
into B (certainly the case with Gainclones). So we expect Solid-State to increase distortion as we go DOWN in
power (getting closer to the AB switching point), whereas Tube tends to
reduce as level goes down.
Going the other way,
Solid-State likely decreases in distortion as power goes up (until it
approaches clipping) whereas Tubes increase. Tubes, especially pure Triodes,
are preferable in this, indeed both, respects.
So at 1 Watt crossover distortion
will not be a factor in tubes, but with
Solid-State Gainclone IC we would still hope for acceptable results at the
same power. But a big BUT! What happens when we go below 1 or 2 Watt
becomes critical with Solid-State. Distortion WILL increase; you can count
on it.
As we approach the point at
which Class AB switching occurs (crossover) the reduced amplitude will not
suppress these distortions (so they may not show up as well at 1 or 2 Watt
and even less at 8 Watt etc). Will very low power levels see an increase in
high order distortions?
This is what we will investigate
here.
How does our project fare with these factors in mind?
We start with 1 Watt:
Caption:
This 1W into 8 Ohm load.
Distortion is primarily EVEN
order, the only odd order showing up is the 3rd, and this is
about 80dB down or 0.01% out of the total THD (Total Harmonic Distortion) of
0.114%.
Caption:
This 2W into 8 Ohm load.
Much the same. Distortion is
again primarily EVEN order; the 3rd has barely changed. Almost no
difference except the note noise floor has been slightly suppressed as would
be expected.
It is apparent that at these
power levels, the First and Second Watt, our amp is very well behaved. There
is a notable absence of ODD higher order distortion.
Before going on, I decided also
to do a IMD (Inter-Modulation Distortion) test at 1W using 1KHz and 2KHz
combined test signal:
Caption:
This 1W approx into 8 Ohm load. IMD = 0.046%.
Again this looks pretty good,
high order irregularities minimal.
BUT...
But what happens
when we REDUCE power. This is a Class AB amplifier and so is prone to
crossover type distortions. We need to examine THD at much lower power levels.
Caption:
This 125mW into 8 Ohm load.
This level means still Class AB
(thus switching). As expected the noise is being suppressed less (this is a
combination of the amp Signal To Noise Ratio and that of the test equipment)
but around –90dB at this low level quite acceptable. No major increase in
THD, indeed the slightly higher THD percentage is probably caused by the
higher S/N. This is better than I would have expected. Some pushing up
of odd orders was anticipated and it has hardly happened.
Let us go much lower still:
Caption:
This just 25mW into 8 Ohm load.
I also did 60mW and it was
comparable to 125mW. But at 25mW it is most likely operating in Class A.
Again S/N will be contributing to THD = 0.151%. An extremely small increase
in 2nd and 3rd orders, and 4th unchanged.
Caption:
This only 10mW into 8 Ohm load.
Definitely in Class A now.
This seems confirmed by almost total lack of odd orders, but THD figure
definitely influenced by S/N. But at this very low level indeed so no reason
to complain.
I am very happy with these
results. Especially when you consider the fact that, as the noise floor
increases it potentially mask higher order distortions. This makes the 10mW
more difficult to interpret, but at 25mW and 125mW problems should have
showed up already because here the noise floor is still well suppressed.
Yes, I must say I'm happy with these results.
I have a question and I'm not
sure what the answer is? The effect of the noise floor acting as dither if
higher order distortions remains buried at low power levels. The other
thing, if the amp is well behaved after feedback (indicating good linearity
before feedback i.e. good open-loop characteristics) and that these higher
order distortions remain constant in absolute level while the dynamic level
(caused by the music), then dither may have a beneficial effect not just at
low level but all levels. Keep this in mind, and I know it may be
difficult to grasp if one isn't familiar with these types of measurements,
that the noise floor is constant in all the above measurements even if they
look like they are not. The reason is simple, the reference 0dB represents a
different voltage in each graph. Since the main culprit (I suspect) is
crossover distortions, then these may well be effectively masked by dither
because these crossover distortions are as constant as the noise floor? I am
not sure about this but the thought has to be entertained?
Pseudo Single-Ended Operation?
But that
is not all. It is apparent that this amplifier outputs little in Class A,
hence I would like to make this statement, one that is open to argument (I
welcome debate) but nonetheless needs to be stated: If we can control
potentially nasty crossover induced distortions, especially high order, then
it becomes apparent to me that either an amplifier should output high level
of Class A and must at the same time have near perfect symmetry (very very
difficult) OR it should have very little Class A so that once it
switches into Class B the amplifier is largely operating single-ended
during those moments in time (and thus avoiding symmetry problems since it
isn't symmetrical). This thinking
would be considered heresy by some.
So I
conclude that Class AB amplifiers must be either highly symmetrical with
significant Class A output power OR have very little Class A provided
crossover distortions are superbly suppressed (and possibly dithered) giving
a pseudo single-ended operation.
(Added 27th May 2003.)
All Power Measurements
I did a series of these
measurements as high as 47W and as low as 10mW and the consistency was
remarkable. Here are the THD figures from all of them.
10mW 0.177%
25mW 0.151%
60mW 0.139%
125mW 0.125%
250mW 0.120%
500mW 0.113%
1W
0.114%
2W
0.111%
4W
0.110%
8W
0.109%
As expected, distortion continues
to drop gradually with increasing power, yet harmonic distortion pattern is
always the same. The distortion harmonics largely ‘cascade’ like a waterfall,
something that was recognized by insightful research decades ago (and
subsequently forgotten) as sonically desirable.
The name Jean Hiraga comes to mind.
Conclusion
Indeed, looking at all THD graphs
(up to 32W) the results are consistent, 5th order or higher
harmonics are well suppressed. In all cases the 3rd harmonic
is equal or lower than the 2nd (cascading) and the latter is
repeatedly the dominant factor in the percentage of THD. This lack of odd
order distortions is very welcome and the predominant EVEN order only tends
to add very little ‘sweetener’ to the sound.
Max Power Test:
What if we push it to the limit,
like more than 40 Watt?
Caption:
47W & THD = 1.324%
WHOA! What has happened
here? Looks kinda ugly but actually not all bad news. The scope shot next
explains all:
Caption:
19.5V RMS into 8 Ohm = 47W.
Notice that the scope shows that
the amp is NOT clipping? So what is happening to the top and bottom of the
waveform? It is actually 100 Hertz ripple intruding (double the Mains
frequency) when the power supply is required to deliver peak current. You
can see this modulating the 1KHz waveform and if you look carefully the
lines are lateral and low frequency in content. Because we are using smaller
than usual 1000uF reservoir caps, these caps are depleting, creating the
above effect. But music signals are NOT continuous like our test signal. So
we would expect that short music peaks might well be left unaffected. So the
amp is easily capable of 35 Watt RMS (continuous) and short-term music
peaks of 50 Watt.
The use of such small reservoir
caps may seem controversial, yet using higher values may well be lowering
the time constant (slower) caused by the combined AC source impedance of the
power transformer and rectifier bridge. The problem is that this is a
complex dynamic inter-face as it is governed by both the charging cycles and
interfered by the DC current requirement of the amplifier itself. I would
love to see someone model this mathematically, could make for some
interesting reading.
This is not exactly new thinking.
I recall some fifteen years ago that Martin Colloms (of HFN & RR) writing on
this very subject and he demonstrated to himself, as I recall it, how he
could make sonic changes by varying the wire gauge between the Bridge and
Reservoir caps. As far as using small caps in our Gainclones, I do know
this, it has a great agility in the mid-bass that makes it as good as
anything when playing acoustic or even electric bass, and other instruments
that require good sense of pitch, pace and tonality.
Final
So here are by reason some good
pointers to the more than acceptable sound quality of these little wonders.
I have said to myself many times over that they don’t deserve to sound as
good as they DO. After all, this is a feedback amp, indeed an Op-amp device
where Feedback is THE controlling element as to gain. It is also Push-Pull
(derision from certain quarters), Class AB (more derision) and not much of
it is Class A (probably about 0.03W) as the idle current is only slightly
more than 30mA.
So while it breaks the mold in many ways there are also some
good indicators. The use of Current Sources (six minimum, probably seven) and Class A in all stages
up to the Quasi Output Stage (NOT complimentary). The Constant Current Class
A Driver (to the Quasi out) used internally must be very good in absorbing
crossover artifacts as they look so well suppressed, no real increase with
descending power, in 3rd order but also 2nd order
which continues to dominate. At all levels it is the 2nd and 4th
combined even orders that define the measurements. Excellent!
I came
across this quote from another manufacturer who uses the 3886 version as
stating:
One common question is whether the
output section of the amplifier is biased into Class-A operation.
Technically, the answer is “yes,” as the exceptional linearity of the output
section design achieves the advantages of pure Class-A operation. Using the
“Intelligent Power Transistor with Gain”
system keeps the output section properly biased at all times to eliminate
crossover distortion caused by the output transistors during turn-on and
turn-off, as opposed to the brute force Class-A biasing techniques used in
other inadequate amplifier designs to compensate for poor linearity that
result in low efficiency, excessive heat, electrical waste, and the need for
power management.
Imagine calling
the Gainclone IC
“Intelligent
Power Transistor with Gain”.
What a Hoot!
Added 17th May 2003.
But measurements are not
everything, it is the SOUND that is the final arbiter! That is why the
commercial equivalent is called
JLTi. Guess what
that stands for?
So all I can say
is Just
Listen
To
it.
Join Us On
The DIY AUDIO FORUM
< <
BACK TO DIY MAIN PAGE |
