summing large amounts of channels.

Help Support GroupDIY:

JohnRoberts

Well-known member
Staff member
Moderator
Joined
Nov 30, 2006
Messages
21,408
Location
Hickory, MS
I have never heard this before. Why is it called Fourier click?
I've never heard it called that either but perhaps a fourier analysis of a step function, or click (very narrow step function), identifies all the sundry frequency components generated.

Interrupting audio signals at other than zero crossings is a somewhat different problem from DC offset clicks. If you visualize a sine wave started or stopped away from a zero crossing, that introduces a DC component. Complex audio is more complex.

A great deal of work has gone into splicing unrelated audio samples together. With pitch shifters reassembling the processed samples after they have been streched out or squeezed together requires some finesse.

Synchronizing wiper changes in digital pots to be correlated to signal zero crossings reduces the energy from those switching perturbations.
===

IMO switch clicks are pretty low on the list of summing bus design problems.

JR
 

sahib

Well-known member
Joined
Aug 19, 2006
Messages
3,429
Location
Glasgow - UK
I've never heard it called that either but perhaps a fourier analysis of a step function, or click (very narrow step function), identifies all the sundry frequency components generated.

JR

The click is not a step function but rather delta (or unit impulse) function. In simple terms it is a blip. It happens at time zero, has zero width and infinite amplitude (and vanishes). So it has zero frequency.

However, I have just checked, the Fourier transform of delta function is a complex exponential.

I think it is one of those wrongly worded thing that is handed down from one person to another.
 

JohnRoberts

Well-known member
Staff member
Moderator
Joined
Nov 30, 2006
Messages
21,408
Location
Hickory, MS
I am not a tube guy or a mathematician... My understanding of turning some signal on or off suddenly is a gain step function, but I may have been calling it the wrong names all these decades.

Back in the 70s working with speech pitch shifters we did a lot of work splicing discontinuous samples together with the least objectionable artifacts...

This is getting pretty off topic.... don't get me started on tone burst generators and dynamic processing design.

JR
 

sahib

Well-known member
Joined
Aug 19, 2006
Messages
3,429
Location
Glasgow - UK
I am not a tube guy or a mathematician... My understanding of turning some signal on or off suddenly is a gain step function, but I may have been calling it the wrong names all these decades.
...................................
This is getting pretty off topic.... don't get me started on tone burst generators and dynamic processing design.

JR

I am not sure what has it got to do with being a tube guy but the basics of this does not require one to be a mathematician.

I do not want this go off topic and drag on unnecessarily either, but do not mix the click generated by turning a switch on and/or off with step function.

However, it would be interesting to know where the "Fourier Switch " originated from.
 
Last edited:

JohnRoberts

Well-known member
Staff member
Moderator
Joined
Nov 30, 2006
Messages
21,408
Location
Hickory, MS
definition said:
NOUN
mathematics
electronics

  1. a function that increases or decreases abruptly from one constant value to another.

Sure sounds like turning a signal on/off to me... but I am a college dropout so self taught about lots of this.

===
I already said that clicks/thumps from DC offsets, are subtly different IMO than starting/stopping audio signals at other than zero crossings.

I used to design my own tone burst generators and synchronized the burst start/stop not only for zero crossings, but to present full sine waves to avoid undesirable DC content when developing dynamics processing.

I recall an old studio trick to add edge to gated percussion by adjusting the gate to click on purpose when opening. That gate opening click would be an away from zero crossing click, and not a DC level step click (while there is some DC component associated with not switching on zero crossings).

JR

PS: Again my apologies, what else can we argue about?
 

Matt Syson

Well-known member
Joined
Dec 17, 2005
Messages
165
Location
France
I referred to it as a 'fourier click' partly as a way of differentiating a 'perceived; click that the 'ear' can hear as opposed to a click caused by fast switching between DC levels which can obviously be both heard and observed and measured on an oscilloscope. The BBC in their previous wisdom defined a step change in level of 'about 1dB' over a certain time period (millisecconds) that theyfelt the listening public would tolerate (not really notice) and thus the equipment should not exhibit such imperfections.
A 'sharp' cut or uncut of a low frequency signal can appear to click so in the workd I inhabit I need to demonstrate that operating a cut (mute) is an audio phenomenon OR a failing of the equipment design (a -40dB DC level shift perhaps!). It is of course signal level and frequency dependent.
The BBC 'definition used to be a 4dB peak measured above the usual noise (hiss/white noise) as measured using a standard BBC PPM meter with gain aded ahead of the meter such that it indicated 0dBu (mark 4). When switching a 'line level' signal whose constant noise level is lower than say -100dBu things can get interesting.
Matt S
 

Winston OBoogie

Well-known member
Joined
Jun 3, 2004
Messages
2,685
Location
UK.
Why all the agonizing over 100 inputs - or even 48? Although I haven't hung out in a studio for 30 years, I just can't imagine that many channels being used simultaneously for mix-down of any kind of music!
The discussion arose because Pucho works at a studio that has a big old classic analogue desk that's still very much used.
The busing in the desk is a modification by a past technician/caretaker of the desk and utilises the Jensen 990 op-amp for mix amp duty. The topology of the desk input channels isn't 990 based, but is an old 'British' design.
While the present scheme works, Pucho was curious if there is any benefit to updating the buses, possibly with devolved mixing.
The initial thread started as talking about an existing classic desk and, in that case, there probably isn't scope or interest in retrofitting current sources for feeding bus lines. But it's an interesting discussion nonetheless.

There are still quite a few studios that utilise big analog desks but, on the whole, you're correct that there's not much call for the big 'uns.

How are they mixing all that stuff down for final movie sound these days?
 
Last edited:

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
13,456
Location
Marcelland
How are they mixing all that stuff down for final movie sound these days?
I don't think there are many analog film console in intensive use today.
Anyway, this dicussion cannot be really put in perspective with the current state of affairs.
Quite often in music mixing, although there are many tracks, sometimes more than 100, a lot of them are pre-mixed in stems, which can be mixed with an analog mixer of a small number of channels. Often no more than 16. So it's a kind of bucket summing, with the first stage digital.
 

moamps

Well-known member
Joined
Sep 24, 2005
Messages
1,906
Location
Croatia, HR
The click is not a step function but rather delta (or unit impulse) function. In simple terms it is a blip. It happens at time zero, has zero width and infinite amplitude (and vanishes).
The delta function is only a theoretical function, it does not exist in practical terms.

So it has zero frequency.
So why then do we hear a click:)?

Since audio systems are mostly AC coupled, the step function is integrated by decoupling capacitors and gets a impulse shape similar to the "delta function".
 

Winston OBoogie

Well-known member
Joined
Jun 3, 2004
Messages
2,685
Location
UK.
I don't think there are many analog film console in intensive use today.

Yes I think that's probably true.
A good proportion of a friend's income now comes from mixing music for film and tv, and I don't believe he ever sees/hears the dialogue/folley etc. tracks beyond a basic draft to work to.

He's 100% ITB. Just him and his rig. It seems such a lonely and impersonal job though if, like him, you came up working behind a desk with a band to record.




Anyway, this dicussion cannot be really put in perspective with the current state of affairs.
For sure. It's nice to dream and think about things though :)

Quite often in music mixing, although there are many tracks, sometimes more than 100, a lot of them are pre-mixed in stems, which can be mixed with an analog mixer of a small number of channels. Often no more than 16. So it's a kind of bucket summing, with the first stage digital.

That would certainly be the logical way if you wanted to sit behind an analogue desk.
Personally, I'd love to see more new music out there that was recorded with WAY fewer tracks but more folks simultaneously playing together in a room.
 

sahib

Well-known member
Joined
Aug 19, 2006
Messages
3,429
Location
Glasgow - UK
The delta function is only a theoretical function, it does not exist in practical terms.
Of course. Nothing starts and stops abruptly. But it helps us describe and analyse.

So why then do we hear a click:)?

Since audio systems are mostly AC coupled, the step function is integrated by decoupling capacitors and gets a impulse shape similar to the "delta function".

I do not mean to drag this on. But my understanding is that impulse function is just another name of delta function (or dirac delta), which is the derivative of (a perfect) step function.

The reason I brought up zero frequency because I was not sure how one would apply Fourier decomposition to a signal that is described as having zero frequency.

However, Matt has explained why he referred to it that way and that is the end of the matter for me.

Edit: Sorry to miss to answer your question why we hear a click. This is because an impulse applied to one terminal of the (coupling) capacitor creates a charge with opposite polarity on the other terminal and this is naturally amplified. A blip, which is manifested as pop or thump due to speaker action.
 
Last edited:

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
13,456
Location
Marcelland
The reason I brought up zero frequency because I was not sure how one would apply Fourier decomposition to a signal that is described as having zero frequency.
Because Fourier transform does not consider a single signal to have no frequency. Zero Hz is a perfectly valid notion for Fourier, as well as negative frequencies.
 

Brian Roth

Well-known member
Joined
Aug 20, 2005
Messages
1,903
Location
Salina Kansas
Personally, I'd love to see more new music out there that was recorded with WAY fewer tracks but more folks simultaneously playing together in a room.
Now we're on a veer <g> but these guys did stuff on a 4 or 8 track (horns were overdubbed as I recall) 50+ years ago and it still sounds semi modern. Four guys/girls in a room....you can hear and feel the sound of a converted movie theater in that room....and also the smell of the cig smoke in the room! <g>



(On that last one, you hear Otis moving in and out from the mic!)

I should start a thread about Why Stax IS The Best vs. Motown! <G!>

Bri
 

Schnitzel

Member
Joined
Jul 10, 2017
Messages
14
While reading through this thread I got curious if I can do further improvements to the summing amps of my Midas XL200 desk. I attached the schematic of one summing amp and apparently it's the classic virtual ground approach. What I already did a few years ago was to replace the NE5534 with an LME49710 because it hast 140 dB of open loop gain. As far as I can tell, it definitely made an audible improvement in the sense that the mix is a bit clearer, not as "dirty".

I had a look at different designs of various classic consoles. For example the summing amps of a Soundcraft 6000 (schematic attached) seem to look a lot like the trans amp mentioned earlier, if I am correct. I also attached a schematic of an SSL 4000G summing amp. Maybe someone can elaborate more on this design? I don't really get the idea behind that approach...

To sum things up, is open loop gain the major performance parameter for a virtual ground summing amp? Would there be further improvement to replace the LME49710 with a trans amp for summing? I'm thinking about creating a kind of drop in replacement (or at least something with minor modifications to the desk). IMO the current summing approach might be feasable with a new desk that is designed from the ground up. But when it comes to upgrading an existing console this would be way too invasive to me.

All the schematics for the mentioned consoles are easily found online so I hope it's okay that I post these snippets here.
 

Attachments

  • midas_sum_amp.png
    midas_sum_amp.png
    377.8 KB · Views: 38
  • soundcraft_6000_sum_amp.png
    soundcraft_6000_sum_amp.png
    31.2 KB · Views: 43
  • SSL_4000_sum_amp.png
    SSL_4000_sum_amp.png
    95.7 KB · Views: 45

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
13,456
Location
Marcelland
For example the summing amps of a Soundcraft 6000 (schematic attached) seem to look a lot like the trans amp mentioned earlier, if I am correct.
Correct. Almost textbook.
I also attached a schematic of an SSL 4000G summing amp. Maybe someone can elaborate more on this design? I don't really get the idea behind that approach...
It's a VLN opamp. Grafting VLN bipolar to a "standard" opamp results in higher OLG and lower noise voltage. Typically with this approach, the noise voltage density is about 1nV/sqrtHz, which is ca. 10 dB better than a good 5534. Noise current density is higher but it does not matter much since the source impedance presented by the bus is very low (about 200r).
To sum things up, is open loop gain the major performance parameter for a virtual ground summing amp?
Phase margin, GBW and noise are also very important. You can have two opamps with identical OLG, but one having a GBW product 10x higher. The latter would probably be preferrable, unless there are stability issues.
Would there be further improvement to replace the LME49710 with a trans amp for summing?
Probably, on the condition that all sources of error are minimized. That means longitudinal bus impedance, longitudinal ground noise, power rails decoupling, ground circulation.
 

Schnitzel

Member
Joined
Jul 10, 2017
Messages
14
It's a VLN opamp. Grafting VLN bipolar to a "standard" opamp results in higher OLG and lower noise voltage. Typically with this approach, the noise voltage density is about 1nV/sqrtHz, which is ca. 10 dB better than a good 5534. Noise current density is higher but it does not matter much since the source impedance presented by the bus is very low (about 200r).
Thank you for the explanation!
Phase margin, GBW and noise are also very important. You can have two opamps with identical OLG, but one having a GBW product 10x higher. The latter would probably be preferrable, unless there are stability issues.
For example the LME49710 datasheet only tells us the phase margin is "excellent". Whatever that means... GBW product is specified as 55 MHz (shouldn't this be a graph, gain plotted against frequency?). Is GBW even an issue with modern op amps? 55 MHz seems more than enough to me. I haven't run into any real noise issues before. With 2.5 nV/√Hz it should outperform a 5532 anways.
Probably, on the condition that all sources of error are minimized. That means longitudinal bus impedance, longitudinal ground noise, power rails decoupling, ground circulation.
What does this mean in practical terms? Keeping longitudinal bus impedance as low as possible (good connections etc.?), keeping ground noise as low as possible (good shielding / grounding, etc.?), power rails decoupling in the sense of not being stingy with 100 nF capacitors? Also what does ground circulation mean in this context?
 

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
13,456
Location
Marcelland
GBW product is specified as 55 MHz (shouldn't this be a graph, gain plotted against frequency?). Is GBW even an issue with modern op amps? 55 MHz seems more than enough to me
For a 5534, it's only 10MHz. When you have 50+ stems, it translates as a -3dB point at about 200kHz, which leaves about only 20dB gain margin 20kHz. Gain margin is important as it is the dominant factor in NFB distortion reduction.
I haven't run into any real noise issues before. With 2.5 nV/√Hz it should outperform a 5532 anways.
Yes it does, by a factor of about 6dB. A good trans amp could bring 12dB improvement over a 5534.
What does this mean in practical terms? Keeping longitudinal bus impedance as low as possible (good connections etc.?)
Yes.
, keeping ground noise as low as possible (good shielding / grounding, etc.?)
Yes
, power rails decoupling in the sense of not being stingy with 100 nF capacitors?
More like where to put them and managing the return currents
Also what does ground circulation mean in this context?
How the currents delivered to the various loads return to ground.
Not something that can be discussed in a few words.
Consider that ground is constituted of a superposition of resistive elements, through which currents develop volyages.
 
Top