| bbakota2000 |
| Long time ago I red an Accuphase's article on the net. It said that the phase (or degree of phase inversion) is of most importance for the correct reproduction of instruments timbre. In Nelsons writings I didn't saw anything about phase invertion at the ends of frequency response (high and low frequencies). One University did a research on how do we react on inverted phase below 100Hz, and they said we hear that as unnatural sound, even unrecognizeable. Can someone (maybe Nelson) tell me something more about that in Pass Labs amplifiers. I would also like to hear your opinions on this subject. |
|
|
| Nelson Pass |
Pass Labs products do not invert phase, although the
original Zen design (not a real product) explicitly does. |
|
|
| bbakota2000 |
| How do you achieve that? I mean, is phase invertion something you are trying to avoid in an amplifier with adding something in a circuit, or there just isn't phase invertion because of a good circuit construction? |
|
|
| bbakota2000 |
| Is "phase distortion" similar to "phase invertion"? Somewhere on this forum I red that some Pass amps and preamps have phase distortion of 5 degrees. Is that true? Can that be void? Is it important at all for the music reproduction? |
|
|
| mrothacher |
| Are you perhaps asking about phase shift rather than phase inversion?:scratch: |
|
|
| bbakota2000 |
| yes (I guess)! I don't know the proper word. What is the phase shift? Give me just brief explanation! |
|
|
| uli |
Hi,
an ideal amp with a frequency range of Zero to Infinite
wil have theoretically zero degrees of phase shift.
If you insert poles you get some degree of phase shift:
the input dc-blocking caps add some phase shift to the
lower range of the signal. therefore good designs start
either at pure dc (e.g. X1000) or at the lowest possible
frequency. this cap produces 45° phase shift at its pole
frequency. the same is true for the upper end of the ampīs
range. therefore designers try to make these poles about
the factor 10 away from the audible range thus reaching
phase shift which is very lo, about 5 degrees or so
An amp with ideal square wave reproduction (0 - oo Hz)
does not necessarily sound perfect. this means that in
real life you always have to accept imperfect solutions
the art of design lies in the perfect blend of those imperfections
just as e.g. a SE power amp with one 300B or an Aleph 2!
Uli
:nod: :nod: :nod: |
|
|
| mrothacher |
| Very lucid explanation, Uli:) |
|
|
| uli |
tx!:D
Uli
:nod: :nod: :nod: |
|
|
| bbakota2000 |
What about Aleph 5? I didn't find any info. Is there any way to minimise phase shift in this circuit without making it sound worse?
PS: Thanks for nice explanation! |
|
|
| Nelson Pass |
On the top end, we tend to roll off our amps - 3 dB at 100 KHz,
single pole. Someone else can calculate the phase that results.
On the low end, we roll the Aleph 5 off at -3 dB @ 0.15 Hz, single
pole, with unity gain at DC. |
|
|
| uli |
| quote: | Originally posted by Nelson Pass
On the top end, we tend to roll off our amps - 3 dB at 100 KHz,
single pole. Someone else can calculate the phase that results.
|
-> ~10° at 20kHz -> 3/16 inches ~ 5mm offset
| quote: | Originally posted by Nelson Pass
On the low end, we roll the Aleph 5 off at -3 dB @ 0.15 Hz, single
pole, with unity gain at DC. |
-> ~0,6° at 20Hz -> 1 1/16 inches offset ~ 28mm offset
Uli
:nod: :nod: :nod: |
|
|
| mrothacher |
| C'mon Uli share the calculation!:D |
|
|
| uli |
Hi mrothacher,
I calculated in the following way:
(maybe not quite correct, but the region is enough)
an rc - combination has one pole at 1/(2pi*r*c)
ok this means a 45° phase shift as the capacitive
impedance part is equal to the resistive one (= r) but 90°
phase shifted.
You now calculate the capacitive impedance x for the
desired frequency (x = 1/(2pi*f*c) and now:
phi = inv tan(r/x) here we are!
delay: (wavelength/360)*phi
wavelength is 340/frequency in meters.
Uli
:nod: :nod: :nod:
PS: dont use this calculation for filters with more than one pole!!!
as their phase depends heavily on the Q! |
|
|
| Nelson Pass |
| Thanks for sharing that. That's why I specified 1 pole. :cool: |
|
|
| uli |
:devilr: :devilr: :devilr:
Uli
:nod: :nod: :nod: |
|
|
| mrothacher |
| Thanks Uli. That helps!:cool: |
|
|
| GRollins |
Incidentally, I know I make some folks 'round these parts nervous when I build circuits with DC to light bandwidth. This is a good example as to why I do so. Phase shift is cumulative and by the time you string a phono stage, a line stage, an active crossover, and a power amplifier in series, you've got a bit of a problem.
The wider the bandwidth the less the phase shift, in the manner indicated above. Granted, people who live in more RF prone areas may have problems, but those of us who live in parts of the world that are barely out of the stone age have at least that one aspect going for us.
Grey |
|
|
| Nelson Pass |
Many years ago I learned not to send really fast amplifiers
out to customers. I don't like for them to have problems due
to crappy cable and poor grounding, so 100 KHz is my default. |
|
|
| GRollins |
Yeah, I know. The real world is a bummer sometimes. But that's one of the glories of the DIY mindset--you can tailor the circuit to local needs and conditions. Not to mention oddball prejudices such as my DC to light conviction. It's not that I ask anyone to come along; it's just something I like to do for my own purposes. I figure you can always cap the thing off at any arbitrary bandwidth that you might want, but it's not so easy to stretch the bandwidth out after the fact. At least, not so easy for those who don't want to cheat by using NFB to do the job.
Grey |
|
|
| uli |
| quote: | Originally posted by Nelson Pass
Many years ago I learned not to send really fast amplifiers
out to customers. I don't like for them to have problems due
to crappy cable and poor grounding, so 100 KHz is my default. |
BTW IMO its best practice to slow down an amp with a single
pole to avoid SID in any case!
Normally you encounter SID in multistage amps somewhere in
the driver or predriverstage where global feedback does not
influence the open loop gain, this stage is overloaded.
To get rid of this phenomenon people try to use NDFL
(nested differential feedback loops) but there you build up
multiple pole systems which tend to be not that stable as a
single pole system. Even an Aleph amp has at least 4 poles
which influence phase and thus the point where NFB switches
to PFB which leads to oscillation.
By carefully adjusting the major lopass pole you set the point
where amplification drops down that gain is below 1 before the
point of PFB occurance is reached.
Uli
:nod: :nod: :nod: |
|
|
| AuroraB |
Distortion is whole family of strange beasts....
In signal theory, the term "Linear Distortion" is used on a circuit that does not have flat amplitude response and/or constant phase.
The easiest way to envision linear distortion is to draw the first 3 ( fundamental, 3rd and 5th) of asquare wave on a piece of paper, and then try to alter the amplitude relations, or phase shift one of the harmonics....
Nature is a cruel beast.......;) |
|
|
| uli |
| quote: | Originally posted by AuroraB
...Nature is a cruel beast.......;) |
But I love her... ;)
Uli
:nod: :nod: :nod: |
|
|
| jh6you |
| quote: | Originally posted by uli
... Even an Aleph amp has at least 4 poles which influence phase and thus the point where NFB switches to PFB which leads to oscillation. By carefully adjusting the major lopass pole you set the point where amplification drops down that gain is below 1 before the point of PFB occurance is reached.
|
Even if I do not fully understand what you say, it is interesting. Do you mean that Aleph amp has many poles, but Aleph is internally compensated so to have a single pole? |
|
|
| uli |
Every gain stage creates a pole caused by the non infinite speed
of the gain device!
In case of Aleph you get one pole formed by Q12/C6, one by
Q5/Q6/C7 one byR7/C8 and the last is formed by R3/R5/R7/C5.
The major pole (which determines the frequency range) is R7/C8.
Uli
:nod: :nod: :nod: |
|
|
| jh6you |
uli
Thanks for your response.
Which Aleph are you referring to? |
|
|
| uli |
...but the circuit is the same for all
Uli
:nod: :nod: :nod:
PS: I personnally tend to like Aleph2 most! |
|
|
| jarek |
| quote: | | By carefully adjusting the major lopass pole you set the point where amplification drops down that gain is below 1 before the point of PFB occurance is reached. |
Uli,
In reality, what is a simplest method to check that this point is reached? |
|
|
| uli |
Hi Jarek,
the simplest method is putting a probe of an oszilloscope
to the output and watch the amp oszillating - now you
know that you have reached (or passed) the point.
Uli :devilr: :devilr: :devilr:
:nod: :nod: :nod: |
|
|
| Nelson Pass |
That's the simplest way, but from experience I can say that
amps which are near to oscillating, but not actually oscillating,
can sound just awful, like some op amps run at unity gain.
It is reaonable to expect that in order to get good sound, you
need a phase margin quite a bit greater than that which merely
prevents outright oscillation. |
|
|
| jarek |
Hi Uli, Nelson| quote: | | ...you need a phase margin quite a bit greater than that which merely prevents outright oscillation. |
Can it be the point where we get clean, regular and without overshoots square wave at 1kHz?
Regards |
|
|
| Nelson Pass |
| You would imagine so, but I think it's not necessarily the case. |
|
|
|
