Harmonic parrot food?
Having just completed my first amp and being astounded at my own brilliance that it works, i am now planning a new one of a rather different nature.
My first is a P-P class A ~8W 807 amp, designed for lowest distortion (I'll post some pics soon) and I am most pleased with it, it is undoubtedly the nicest bit of kit I have owned, not that it has much competition.
However; all this talk of valves amps sounding warm and sweet due to the even harmonic distortion has raised my curiosity and I am currently designing one of these parafeed amps (for my daughters of course ;) ). I have a nice NOS Tung-Sol Chatham 5998 to be used and a few 6H8C (Russian 6SN7) left over, so I am looking at a Beta-follower first stage with a parafeed + active load output.
I am currently debating whether the load should be pentode or MJE350 cascode any thoughts on that would be appreciated.
The main point of this post though is this distortion. How much is too much? i can get more power out of the tube by using a lower impedance OPT but with more 2nd harmonic. Also I've noticed in my spice sims that the asymmetry causes imbalance in the OPT which then oscillates at LF until settling to a new balance point.
I've found I can cancel most of this imbalance by increasing Rk of the lower valve in the B-follower, thereby shifting the operating points of the valves to create the opposite imbalance to that created by the 5998. (anyone still with me?:confused: )
I have a feeling though that this would create increased odd harmonics due to the imperfect opposition of the distortion created or something. Anyway the 2ndH is the point! To a certain extent...:rolleyes:
Also; While I understand that 2ndH is due to asymetric gain I'm not sure what mechanism creates the higher harmonics, anyone care to enlighten me?
Any thoughts about anything would be most appreciated.:D
I know next to nothing about this. I just wanted to say that I
too would be most interested in what the guru's have to say.
The objective we hope is not to distort the original music
within a reasonable dynamic range that can be reproduced
faithfully. Only to unfaithfully distort clipping into something
less obnoxious... 2nd Harmonic preferred slightly over 3rd,
and very little of any harmonics higher....
Pumping gently away from grid conduction on one
extreme, and squashing peaks mostly on the other.
The load line should be linear over most of its range,
only squishy near cutoff, and into A2 (where grid
conduction will prevent you going most of the time)
A bit of AC feedback from plate to grid though a cap and
resistor might bring the gain down, linearize, and lower
the effective plate resistance.
This doesn't rob as much headroom as a resistor under
the Cathode... Nor muck too much with the character of
triodes. Can even make pentodes, beam power tubes,
even certain sandy devices behave a bit like triodes
(without performace loss of triode strapping the screen)
Perhaps this will stabilze your situation by bringing the
plate impedance less than the load... Cathode resistors
INCREASE plate resistance??? Yep, go figure...
Read the chapter by O.H. Schade "Beam Power Tubes"
RCA Electron Tubes Vol1, 1938. I don't have the link
handy, but posted to this forum before. Quick search.
I am no Guru, could be spewing complete nonsense...
My understanding was that the even harmonics sound warm and odd harmonics sound dischordant as they bear no 'musical' relationship with the original frequencies. Also; Isn't 2ndHD caused by a smooth change of gain right across the voltage swing, rather than constant gain and then bunching at the ends? No guru myself either so :confused:
I'm taking feedback to the cathode of the B-follower to reduce sensitivity and drive signal. Of course this also reduces harmonic distortion, hence the initial question. I guess I am wondering if I should be using NFB to reduce gain or just attenuate at the input.
Attached is my current (very sketchy :whazzat: ) layout.
I'm thinking those current source loads could stand to
have a smoother rolloff into starvation. Let the far right
of the load line droop a bit... The object not to force
excessive linearity in the region of clipping.
All your clipping (or squashing) will be on the right end
of your loadlines, as grid conduction pushes you away
from clipping on the left.... Transients might clip on the
left occasionally, but not as often as the right....
2nd and 3rd harmonic both relate well to music.
2nd is the Diatonic Octave, 3rd is the Diatonic 12th
(12th being an Octave plus a 5th). Either plays to
well to the original note as a chord. Higher order
harmonics relate not so well to any music we tend
to be familiar with.
Intermodulation occours any time two notes are
struck in the presence of a distortion. Those can
can be completely unrelated to the original notes.
IMD is an unfortunate side effect of any added
I think, but I am not certain, that IMD also becomes
less related, the higher the harmonic that mixed it?
Squashing a sine wave single endedly with a square
law produces 2nd order harmonic. Squashing a sine
wave equally on both ends produces a 3rd Harmonic.
Clipping a sine wave flat on top, bottom, or both will
produce 3rd and higher high order distortions, as will
most types of distortion in the zero crossing...
A square wave can be thought of as the is a sum of
3rd, 5th, 7th, 9th, etc... Odd harmonics in diminshing
Even harmonics sum to a sawtooth? Is that correct?
Thats what I remember reading, but seems unlikely
as I think about it. Get that doublechecked before
accepting as fact.
Wow, Most informative! Many thanks.:)
Back to my original question though, how much distortion would a nice sounding amp produce, 1%, 10%? I've heard that it sounds nice on instrumental stuff but makes vocals sound terrible.
Have found and printed that Beam Power Tubes article, will read tonight.
Whole Number/Whole Number harmonics of any pure single
sinewave usually sound musical. Its when two or more notes
get mixed... Things turn ugly.
The is one notable exception, the "Power Chord". To play
the Diatonic Root and 5th at the same time. That would
correspond to the 1/1 and 3/2 harmonic in our world...
Almost any distortion will produce musical intermodulation
given a pure tone or power chord as the only input. Any
thing else put through the same amp likely to sound like
Guitar amps explained in a nutshell...
Anyways there can be a lot of "bad" harmonics and the
human brain may choose to ignore it. Perhaps related to
the way we isolate a single voice in a crowded room....
When our signal processing capability is overloaded, we
seem to prioritize sounds we hear loudest, earliest, and
at the lowest frequency.
Low order harmonics are lower in frequency that the
"nasty" high ones, thier presence can be very effective
in distracting us from noticing higher orders and IMD...
When distortion becomes unavoidable (at clipping) the
root should preferably be significantly louder than
2nd harmonic, and this louder than the 3rd, and this
louder than the rest of the unrelated products...
There is no particular reason to want a lot of distortion
at signal levels that do not threaten clipping, unless
you are building an MI amp to play only power chords....
Thanks, that and the Schade stuff have much improved my understanding of distortion.
However I am intending a hi-fi amp rather than guitar :rolleyes: I should have mentioned that from the start.
My thinking on the required gain, NFB, output resistance etc. has resulted in the design changing a bit (See attached) It's now a parafeed cathode follower, DC coupled to the first stage. I really like the simplicity of this topology. :cool: I'll probably use a pentode sink (EL34? Suggestions?)
Has anyone seen this topology elsewhere? I did a search but all I could find was this
I think I'll go for a dual 9V secondary mains toroid as the OPT as this would give me the option of putting secondaries in series for 940ohm Zp giving about 6W or parallel for 3K8 Zp giving about 1.5W but better Zout/damping and distortion.
The advice I gave applies equally to either Hi-Fi or Guitar.
Only with guitar you may want distortion most of the time.
And with Hi-Fi, distortion only the minimum necessary to
Your latest amplifier circuit is extremely linear, and may clip.
Really depends how soft those constant current sources are
in starvation, as they approach the rail. Thats where good
or bad harmonic effects of clipping will be determined.
Some music theory for Techies:
There was this jerk named Pythagorus or something...
Always playing with string and triangles instead of doin
something useful with his life.
Anyways, he was probably not the first to notice that
a string of half (or double) a reference length at equal
tension would play a highly related note. The "Octave",
or to non-musicians, either the 1/2 or 2/1 harmonic...
What he did that was unique, was try strings of 2/3
or 3/2 length. Discovering the "Fifth", I'll get the the
screwy name later as this harmonic has nothing to
do with five.... at least not mathematically.
So now he had two different type of notes that played
good together (discovering or at least documenting the
first "Power Chord"). But he didn't have TAB to write it
He continued this way, shortening his string 2/3 over
and over. Lengthinging by an octave if it got too short.
every time he did this, a new note was formed.
This nonsense went on untill the 13th time. When he
discovered that he was back almost (but not quite)
the same length of string where he had started. He
had defined twelve unique notes, and played them
in "Circle of Fifths" sequence.
Sorted from lowest to highest by frequency, he had
given us a "chromatic scale" of 12 notes that had a
logarythmic relationship. Like marks on a slide rule,
or the frets on a guitar.
Anyways, the note played first in a sequence seemed
to strongly influence how the others were heard. Not
all the other notes had a simple fractional relationship
to the "root" note. We hear fractions, not logarithms...
Seven (including the root) sounded good. The other
five were duds (the black keys on a piano, if you
happened to start with "A" as your root note.)
Start with any other key on the Piano and the spacing
between the good "Diatonic" notes are the same, but
the black keys don't just up and re-arrange themselves
to accomodate that you started other than "A"
So the seven "Diatonic" notes were numbered I-VII,
and we are stuck with that crazy number system for
harmonic notes instead of meaningful fractions.
If Pythagerwhatsisface had only gone to see "The Sound
of Music" he woulda known "Do Re Me Fa So La Te" and
spared us a lot of grief.
But there's more than one way to skin a cat, and them
ballywooders like Ravi Shakala-de-da had something
else up thier sleeves (did they have sleeves back then?)
If probably goes wayyy back even futher than Peabody.
If you take all the simple fraction fractions from 1/1 to
1/8 and all inbetween. Knock out the ones that are
duplicates, like 4/6 is a dupe of 2/3. You have some 22
fractions that survive the culling. I think they are called
Sarutis or something. Others might call them "Just"
intonation. I may got my factuals slightly in error so
get a grip on another source right quick like...
Anyways, this system ain't logarythmic. Frets on a Sitar
is all over da place. You can't just up and change "keys"
and have the other fractions line up the same way.
They even made some of the frets movable, perhaps to
help correct this oversight on the fly.
Then you gots those freaks who has to be different and
try other logarythmic scales that do change keys easy
and with frets that stay put where you expect em...
10, 19, 26, and 31 evenly tempered notes per octave
have a lot of useful "hits", but get harder to play the
more dud notes they seem to accumulate inbetween.
Then you got Microwave Engineers that throw octaves
out the window and create a scale of only 13 unique
notes (the same every 14th), but the pattern repeats
upon the Tripling of frequency (Third Harmonic) instead
of the old fashioned octave (Second Harmonic).
You can blame Bolen and Pierce. I coulda mispelt either.
One or the other might even be alive to give me grief...
There's also a spacey chick named Elaine Walker that
plays some rockin tunes in that scale.
Good ratio of hits to duds, maybey better than what
Pythagerwhatsit's crummy old system gave us.
Anyways, if your amp can't play what you want cause
of some physical limit, like the power supply rail. It
should at least play an aproximation of MUSIC instead
of freaking out and playing unrelated garbage.
It might help a bit if you got a clue what music is...
From the standpoint of math, and what the human
mind can process. Not merely from a perspective of
personal taste. As I clearly have none...
So ends this fractured fairy tale!
I am now armed to the teeth with new-found knowlege of what music is (and about to listen to some Elaine Walker found here. Sounds intriguing.:cool: Always love finding new music)
And the design has progressed a little:
Using an EL34/6L6/KT88 type pentode CCS (specific suggestions appreciated) and LM334 for the B-follower.
My problems now are:
1 - Power supply rejection.
When designing this initially my theory was that a cathode follower with active load would have almost all power supply ripple at the output due to the potential divide formed by valve (low impedance) and load (high impedance). This would be great, as the output is taken between cathode and +HT therefore the ripple would common-mode across the OPT and not coupled to the secondary. This does not seem to be the case in the sim shown above. Here the AC across the OPT is about -3dB from full HT ripple. In an attempt to address the problem I have fed some of the ripple to the B-follower cathode through C6. This works but is not ideal and any better suggestions would be most welcome.
2 - Stereo separation
The amp will be built on a single chassis with single PSU (probably RCRC type) Both stages are (theoretically) constant current so crosstalk through the PSU should be avoidable with careful bypassing, however all the triodes are in double triode envelopes with both channels running through the valves. I have no experience in running valves like this and crosstalk through the C(hk) worries me. Tung-sols 5998 datasheet shows internally paralleled heaters so I figure I can measure C(hk) and shunt the signal to ground with a significanatly higher cap (100x ?) between each heater pin and the chassis. Unfortunately the 6H8C datasheets I can find show only 1 heater so I'm just hoping the same strategy will work. OPTs will be kept apart so this shouldn't be a problem. Is there anything more I could be doing?
|All times are GMT. The time now is 11:29 PM.|
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright ©1999-2014 diyAudio