R/E/P Community

Please login or register.

Login with username, password and session length
Advanced search  

Pages: 1 2 [3]  All   Go Down

Author Topic: High/low pass  (Read 17117 times)

Andrew Hamilton

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 573
Re: High/low pass
« Reply #30 on: December 19, 2010, 01:45:56 PM »

Geoff Emerick de Fake wrote on Sun, 19 December 2010 11:43

Andrew Hamilton wrote on Fri, 17 December 2010 07:14


If filters had no effect on the parts of the spectrum that lie outside their bandwidth (Q), I'd settle on a 20 Hz switch.  But when I invoke any (natural) filter down at 20-ish, there's usually too much taken away from the good bass up through the low mids.  
Then these filters have a serious design or implementation flaw. Having to shift the corner frequency nearly an octave away in order to prevent it gnawing at the useful range is extreme.



Now you can't be saying there's actually something wrong (in the wrong way) about the Massivo's Mastering filters, now, can you?   Rolling Eyes    As I mentioned last time, they go down to 12 Hz before turning off.  I just want to have more selections in that zone to match more mixes.   I suppose it's sometimes worth going to extremes to be subtle.  

Andr
Logged
www.serifsound.com
premastering for CD and DVD-A.  Featuring FTP load in and delivery as well as analog tape transfers.

Gold

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1453
Re: High/low pass
« Reply #31 on: December 20, 2010, 11:29:19 AM »

Filters are defined by the -3dB point. If what's above the filter knee affects the audio too much then a filter is the wrong tool. Use a shelf.

This whole discussion is amusing in the way that puffery over .5dB at 17K is. Soldier on bearers of extreme audio quality. Your fight doesn't go unnoticed. Wait, maybe it does.
Logged
Paul Gold
www.saltmastering.com

On the silk road, looking for uranium.

Geoff Emerick de Fake

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 348
Re: High/low pass
« Reply #32 on: December 20, 2010, 12:00:31 PM »

Andrew Hamilton wrote on Sun, 19 December 2010 12:45

Geoff Emerick de Fake wrote on Sun, 19 December 2010 11:43

Andrew Hamilton wrote on Fri, 17 December 2010 07:14


If filters had no effect on the parts of the spectrum that lie outside their bandwidth (Q), I'd settle on a 20 Hz switch.  But when I invoke any (natural) filter down at 20-ish, there's usually too much taken away from the good bass up through the low mids.  
Then these filters have a serious design or implementation flaw. Having to shift the corner frequency nearly an octave away in order to prevent it gnawing at the useful range is extreme.



Now you can't be saying there's actually something wrong (in the wrong way) about the Massivo's Mastering filters, now, can you?   Rolling Eyes  
honestly, I'm not in awe with Manley products, so that would not be out of the question. Now, I have checked the specs, they appear to be bona fide 3-order RLC filters with the correct frequency response plots. So, if these plots are genuine, cutting at 18 Hz should not take away "from the good bass up through the low mids". Unless their coils saturate in a manner that affects the upper range...
Abstract from the user's manual:
We mentioned the inductance value can change with applied
power. This also turns out to be a surprising advantage. For
example, in the low shelf, with heavy boosts and loud low
frequency signals, at some point, the inductor begins to
saturate and loses inductance. Sort of a cross between an EQ
and a low freq limiter. The trick is to design the inductor to
saturate at the right point and in the right way.
 
Which seems to me a questionable justification for using inferior components.
Quote:

 As I mentioned last time, they go down to 12 Hz before turning off.  I just want to have more selections in that zone to match more mixes.   I suppose it's sometimes worth going to extremes to be subtle.  

Andr
Logged

Geoff Emerick de Fake

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 348
Re: High/low pass
« Reply #33 on: December 20, 2010, 12:06:57 PM »

Gold wrote on Mon, 20 December 2010 10:29

Filters are defined by the -3dB point. If what's above the filter knee affects the audio too much then a filter is the wrong tool. Use a shelf.
A shelf has a much gentler slope than any HPF, so for a specific attenuation, it will affect the upper frequencies more. My answer is: If what's above the filter knee affects the audio too much then the order of the filter is insufficient - increase the slope.

And sometimes (quite often in fact), choosing the settings of an HPF is a compromise between not taking enough of the noise and taking too much of the signal.
Logged

Patrik T

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 833
Re: High/low pass
« Reply #34 on: December 20, 2010, 12:14:47 PM »

There is a reason I decided to put 11 pos ELMAS instead of 24 pos ELMAS for frequency choices on the 5033 EQ.

I will continue to use my additional 50 Hz HPF (only), sitting in another Neve-ish design, until I find something with maximun 3 frequency choices between 0-100 Hz.

Thanks everyone so far, especially you Paul.


Best Regards
Patrik

Logged

Andrew Hamilton

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 573
Re: High/low pass
« Reply #35 on: December 22, 2010, 08:19:00 AM »

Geoff Emerick de Fake wrote on Mon, 20 December 2010 12:00

honestly, I'm not in awe with Manley products, so that would not be out of the question. Now, I have checked the specs, they appear to be bona fide 3-order RLC filters with the correct frequency response plots. So, if these plots are genuine, cutting at 18 Hz should not take away "from the good bass up through the low mids". Unless their coils saturate in a manner that affects the upper range...


Actually, they don't offer a corner at 18 Hz, on either version, though they do, at 16 Hz (on the mastering version)...

Also, the frequency plots don't show the group-phase plots.  That's where I'm doing half of my work.   Massaging the balance as the signal goes through a daisy chain of impedences.

The extra icing of eq action (beyond the obvious benefit provided by the attenuation or equalization of target signals with only boosts or cuts in selective frequencies) is the un-twisting of phase that can help to un-mangle a bad mix of frequencies, not unlike the twin action of de-emphasis networks in phono pre-amps and tape machines.   It's a pitch and time phenomenon.   Not only how loud was 300 Hz in relation to 30 Hz, but _when_ did the 300 Hz component of the transient hit you, as compared to the 30 Hz component?  

If your loudspeakers have high order cross-overs, you might not be as plagued by impulse response anomalies in mixes than I.  (duck)



Geoff Emerick de Fake wrote on Mon, 20 December 2010 12:00


Abstract from the user's manual:
We mentioned the inductance value can change with applied
power. This also turns out to be a surprising advantage. For
example, in the low shelf, with heavy boosts and loud low
frequency signals, at some point, the inductor begins to
saturate and loses inductance. Sort of a cross between an EQ
and a low freq limiter. The trick is to design the inductor to
saturate at the right point and in the right way.
 
Which seems to me a questionable justification for using inferior components.


I don't hear the unwanted attenuation as a dynamic deficiency, but rather an eq one, so the saturation, itself, however limiter-like it may be (in the wrong hands?) isn't causing the sonic aenemia I'm trying to describe.  

Also, Manley's design goal for the Massive Passive has never been transparency.  The marketing slogan is "A Pultec on Steroids."  Color is desired, here.   The presence of saturating transformers and inductors is the reason for choosing Manley.   If you can get digital recordings sounding plausibly nostalgic-sounding by using only surgical tools, I salute you.  

Geoff Emerick de Fake wrote on Mon, 20 December 2010 12:00

It's your absolute right to wish for more, but I think you'll find that there is no audible benefit there, as much as I understand the desire of more resolution between 20 and 40 Hz.



Wasn't there an Aussie ME who couldn't take the brightening effect of the Lavry Engineering AD-122 MK III's digital high pass filter for blocking DC offsets?  He had to turn to a vintage Pacific Microsonics to get satisfaction (and no unwanted attenuations).  Granted, the Lavry digital de-DC offset filter is minimum phase, but it's extremely low in center frequency - and high in quality.

My wish is for my analog high pass to have all the garbage disposing power I want, but at selectively subtle bite points, so as not to lower the body or weightiness of vocals and other instruments in the least.   If you can't hear any damage that a 22 Hz High Pass filter does to an electric guitar, then either the mix needed some slimming down across the board, after all, or you moved your head!   Shocked  





Andrew
Logged
www.serifsound.com
premastering for CD and DVD-A.  Featuring FTP load in and delivery as well as analog tape transfers.

Geoff Emerick de Fake

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 348
Re: High/low pass
« Reply #36 on: December 22, 2010, 09:25:51 AM »

Andrew Hamilton wrote on Wed, 22 December 2010 07:19

Geoff Emerick de Fake wrote on Mon, 20 December 2010 12:00

honestly, I'm not in awe with Manley products, so that would not be out of the question. Now, I have checked the specs, they appear to be bona fide 3-order RLC filters with the correct frequency response plots. So, if these plots are genuine, cutting at 18 Hz should not take away "from the good bass up through the low mids". Unless their coils saturate in a manner that affects the upper range...


Actually, they don't offer a corner at 18 Hz, on either version, though they do, at 16 Hz (on the mastering version)...

I don't own a Massive Passive, so that's what I read on one of the front panel line drawings.
Quote:

 Also, the frequency plots don't show the group-phase plots.
Oh yes they do! Even manley products have to obey the Bode relationship between phase and frequency response.
Quote:

 That's where I'm doing half of my work.   Massaging the balance as the signal goes through a daisy chain of impedences.
I don't see any direct relationship between time (or phase) response and impedance...
Quote:

 The extra icing of eq action (beyond the obvious benefit provided by the attenuation or equalization of target signals with only boosts or cuts in selective frequencies)

I don't get it...Isn't EQ about boosting or cutting selective frequencies?
Quote:

 is the un-twisting of phase that can help to un-mangle a bad mix of frequencies, not unlike the twin action of de-emphasis networks in phono pre-amps and tape machines.   It's a pitch and time phenomenon.   Not only how loud was 300 Hz in relation to 30 Hz, but _when_ did the 300 Hz component of the transient hit you, as compared to the 30 Hz component?
We all know that there is a very clear relationship between frequency response and time response. But I never had any proof that moving different components of the sound on the time-axis had any significant effect on the perceived sound, provided the signal path (including air and ears) is linear (at least in the limited amount a HPF could move them).
Quote:

 If your loudspeakers have high order cross-overs, you might not be as plagued by impulse response anomalies in mixes than I.  (duck)
In doubt, I check on headphones.
Quote:

 
Geoff Emerick de Fake wrote on Mon, 20 December 2010 12:00


Abstract from the user's manual:
We mentioned the inductance value can change with applied
power. This also turns out to be a surprising advantage. For
example, in the low shelf, with heavy boosts and loud low
frequency signals, at some point, the inductor begins to
saturate and loses inductance. Sort of a cross between an EQ
and a low freq limiter. The trick is to design the inductor to
saturate at the right point and in the right way.
 
Which seems to me a questionable justification for using inferior components.


I don't hear the unwanted attenuation as a dynamic deficiency, but rather an eq one, so the saturation, itself, however limiter-like it may be (in the wrong hands?) isn't causing the sonic aenemia I'm trying to describe.  

Also, Manley's design goal for the Massive Passive has never been transparency.  The marketing slogan is "A Pultec on Steroids."  Color is desired, here.   The presence of saturating transformers and inductors is the reason for choosing Manley.   If you can get digital recordings sounding plausibly nostalgic-sounding by using only surgical tools, I salute you.
OK, so that doesn't explain why you hear something at 200Hz when you HPF at 20. Something else must be wrong.
Quote:

 
Geoff Emerick de Fake wrote on Mon, 20 December 2010 12:00

It's your absolute right to wish for more, but I think you'll find that there is no audible benefit there, as much as I understand the desire of more resolution between 20 and 40 Hz.



Wasn't there an Aussie ME who couldn't take the brightening effect of the Lavry Engineering AD-122 MK III's digital high pass filter for blocking DC offsets?  He had to turn to a vintage Pacific Microsonics to get satisfaction (and no unwanted attenuations).  Granted, the Lavry digital de-DC offset filter is minimum phase, but it's extremely low in center frequency - and high in quality.
You can't compare the action of a FIR filter with an analog one, particularly at VLF, where FIR introduces audible amounts of pre-ringing.
Quote:

 My wish is for my analog high pass to have all the garbage disposing power I want, but at selectively subtle bite points, so as not to lower the body or weightiness of vocals and other instruments in the least.
I understand that very well; my view is that there's more useful signal between 20 and 40Hz than below 20, so I want my resolution where I need to be selective.
Quote:

 If you can't hear any damage that a 22 Hz High Pass filter does to an electric guitar, then either the mix needed some slimming down across the board, after all, or you moved your head!
I can hear it eliminating pick thump, which I find nice... The guitar has no musical content below 80Hz (6-string); the rest is noise. I don't want to respect the integrity of the guitar sound, I want to present it in the most interesting musical way. Just like on spanish guitar, I would minimize soundboard thump and finger noises, and on piano I would try to get rid of pedal noise. A purist would try to keep all these extraneous noises for the sake of "respecting the natural sound of the instrument"; I'm a music maker.
Logged

Andrew Hamilton

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 573
Re: High/low pass
« Reply #37 on: December 22, 2010, 01:07:09 PM »

Geoff Emerick de Fake wrote on Wed, 22 December 2010 09:25

Andrew Hamilton wrote on Wed, 22 December 2010 07:19

 Also, the frequency plots don't show the group-phase plots.
Oh yes they do! Even manley products have to obey the Bode relationship between phase and frequency response.


Oh, I didn't wish to imply that Manley break the laws of Physick.

Do they actually _show_ the phase plots?  (I only found the frequency plots.)  

Geoff Emerick de Fake wrote on Wed, 22 December 2010 09:25

Andrew Hamilton wrote on Wed, 22 December 2010 07:19

That's where I'm doing half of my work.   Massaging the balance as the signal goes through a daisy chain of impedances.
I don't see any direct relationship between time (or phase) response and impedance...


You might try altering the load that your chain "looks into" and listen to whether there's any change in sound.  I have noticed a dramatic change in sound depending on the handshake applied between gear.  Others do, too.  

Geoff Emerick de Fake wrote on Wed, 22 December 2010 09:25

Andrew Hamilton wrote on Wed, 22 December 2010 07:19

The extra icing of eq action (beyond the obvious benefit provided by the attenuation or equalization of target signals with only boosts or cuts in selective frequencies)

I don't get it...Isn't EQ about boosting or cutting selective frequencies?


That, and, especially in mastering, where you are eq'ing mixes of tracks, rather than individual parts, the art of manipulating the balloon animal of sound by varying the delay of assorted groups.   Impulse response is more important than frequency response in achieving realism.

Quote:

 ...I never had any proof that moving different components of the sound on the time-axis had any significant effect on the perceived sound...


That's alright.  Your secret's safe with me.  (;

Quote:


In doubt, I check on headphones.


Then you will have to sum the phantom center image in your mind - or use the Lavry DA11?


Quote:

OK, so that doesn't explain why you hear something at 200Hz when you HPF at 20. Something else must be wrong.


I'm sorry you aren't already aware of what I'm talking about.  Man's hearing is highly sensitive to the timing of near-simultaneous events.   Whereas man's frequency response in hearing is a crazy undulation.  

Why does eq change the sound in a way that can exceed the apparent action presented by the gain change, alone?  (It may be useful to recall that mastering engineers are concentrating their cuts and boosts more so in the vicinity of 1/2 dB changes (or less!), rather than 6+ dB changes.)

When you alter the phase of only a group of frequencies within a multi-complex signal (such as a mix of instruments would present) you introduce the possibility of masking or unmasking subharmonic, as well as harmonic, partials that change the balance, just like the amplitude adjustments do.  It's not possible to separate the twin-action without a delay for reverse filtering (adding half the desired boost, or cut, each time) (to undo the phase-shift).  

The realism of the sound of the master is dialed in by addressing not the gain of each frequency, but more specifically, by correcting the "train" of frequencies.  (Of course, the listening auditorium's acoustics and monitors will determine to large degree the ability of the mastering engineer to dial in a correction (if needed) that will translate globally to other hi-fis.)


Geoff Emerick de Fake wrote on Wed, 22 December 2010 09:25


...You can't compare the action of a FIR filter with an analog one, particularly at VLF, where FIR introduces audible amounts of pre-ringing.


Dan's digital de-DC offset filter is minimum phase.  So there's no pre-ringing.   You're talking about linear phase.  Nevertheless, this expensive HP filter in the digital domain is too brutal for our friend in Brisbane.  I have heard it, too, frankly, and have learnt to live with it/work against it.


Geoff Emerick de Fake wrote on Wed, 22 December 2010 09:25

Andrew Hamilton wrote on Wed, 22 December 2010 07:19

 If you can't hear any damage that a 22 Hz High Pass filter does to an electric guitar, then either the mix needed some slimming down across the board, after all, or you moved your head!
I can hear it eliminating pick thump, which I find nice... The guitar has no musical content below 80Hz (6-string); the rest is noise.


If you can high pass at 80 Hz and not have the guitar come out too bright, it was too darned dark to begin with! (:  Such is the case with mixing.  

Geoff Emerick de Fake wrote on Wed, 22 December 2010 09:25


I don't want to respect the integrity of the guitar sound, I want to present it in the most interesting musical way...


To me, there is no difference.   To betray the integrity of the sound is to venture away from music and into psychedlia - which can be fun, but then there are no clear objectives, are there?

Geoff Emerick de Fake wrote on Wed, 22 December 2010 09:25


Just like on spanish guitar, I would minimize soundboard thump and finger noises, and on piano I would try to get rid of pedal noise. A purist would try to keep all these extraneous noises for the sake of "respecting the natural sound of the instrument"; I'm a music maker.


I see.   And I am a mastering engineer who _does_ want the thumps, noises, squeaks, and pedal sounds to stay in - just not the digital clicks and crackle   ):


Furthermore, I have yet to invoke a min-phase HP filter that doesn't eat away at the body of guitars and vocals.  When they need it, or can tolerate it, it's great.  But when they don't want cutting, it's quite difficult to find a low cut that isn't deleterious in some small way or other.   We are in mastering trying to put as much toothpaste back into the tube as possible, of course.   Do _no_ harm.   (Not that Thou wouldst.)



Cheers,
    Andrew
Logged
www.serifsound.com
premastering for CD and DVD-A.  Featuring FTP load in and delivery as well as analog tape transfers.

Allen Corneau

  • Full Member
  • ***
  • Offline Offline
  • Posts: 224
Re: High/low pass
« Reply #38 on: December 23, 2010, 09:03:50 AM »

I don't know if this helps, hurts, or is just plain incorrect, but I'll toss i in there anyway.

Back in '05 I was doing some testing with the high-pass filters available to me one of which is the MP. One of the things I was trying to understand was the relationship between the frequency cut and the phase change.

I ran tones through the MP while changing the HP filer on one channel (leaving the other out) and looking at the phase meter on DigiCheck. Here is what I got...

index.php/fa/16074/0/

If this is just flat-out incorrect analysis please let me know so I'm not spreading bad info.

Logged
Allen
---
Allen Corneau
Mastering Engineer
Essential Sound Mastering
www.esmastering.com

Geoff Emerick de Fake

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 348
Re: High/low pass
« Reply #39 on: December 23, 2010, 11:27:20 AM »

As I said earlier, these filters behave like any normal 3rd-order maximally-flat (Butterworth).
You can see in particular that the phase-shift is 135
Logged

Laarsø

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 88
Re: High/low pass
« Reply #40 on: December 24, 2010, 08:50:38 AM »

Monsieur de Fake points out well that 0˚ can represent the full twist, in addition to the untwisted phase.  Anyone who has set azimuth on tape machine may have seen this phenomenon when 1 kHz looks like lasso, even though 10 k looks like thin line of correlation (only).  

Whatever Mr. Hamilton thinks he's hearing, it's obvious that there must be some audible difference between invoking a natural filter of a given order and a linear phase filter of comparable slope (outside of complaints of audible "pre-echo").  This is why it has been shouted online erroneously that transferring vinyl to digital, flat (no RIAA de-emphasis), can't result in good sound when put on CD "...since digital eq doesn't introduce the requisite phase shifts to de-emphasize the RIAA minimally-phased pre-emphasis, like analog eq does."  [sic]

Of course that is incorrect, since the type of digital filter which doesn't introduce any phase shifts wouldn't be the type used in (properly implemented) digital de-RIAA filter, yes?  


So, when we find linear phase plug-in not to be of service - _and_ the reason it isn't work is not because of coldness caused by pre-echo - then it must be because it dasn't untangle the phase in anticipated way.  To this extent, we are not hearing the effect (but are wishing we were!) of the time axis, while eq'ing, whether we know it as such, or not (i.e., intuitively, logical choice of gain change is not helping, since it's hurting something else, but also making the guitar sound like it was painted by Salvador Dali (you know, the sound illusion starts "looking" like a melting guitar - like the pocket watch in that one oil painting...), even though the levels of frequencies, themselves, are, ok, plausible.)


Also, when it comes to impedance, there is certainly a need to watch out for the effects of loading of networks - especially unbuffered passive ones.  (:  

For example, the Lavry M•DA-824 sounds great with only a T-pad between it and the amp.  It can drive a load as low as 600 Ohm without releasing magic smoke.   But the ideal load for the DAC to see, in this type of passive pre-amplification is about 2k Ohm.  I have tried both loads, as well as some that are in between, and think Dan was right when he recommended 2k.  Sounds the best.

(If you use a 2.5k Ohm T-pad, and your amp's single-ended input Z is, say, 11k Ohm, you end up pretty close to the target load (at 2.04k Ohm)).  



Speaking of transferring vinyl to digital flat, if you use a mic pre for amplification (instead of a traditional phono preamp that already has a de-emphasis network), you will need to use a  de-RIAA plug-in in your computer, afterwards, to make CD that is worth listening to...   You will first, however, need to determine the input Z of preamp and then calculate the appropriate impedance drain to place at input so that the phono cartridge "looks into" target load (which, in my case, was 47k Ohm.  Guess which type of cartridge design?)   Thanks to Glenn Meadows for this insight.  It is widely known, but it was his description that I first encountered.  (Also, Richard Hess and Burgess Macneal were both of great help in sorting out my confusion over the high frequency peak in the sweep on the Cardas burn-in record, mastered by Stan Ricker.   The peak is intentional - or, at least, is not a problem with your system.  The RCA record has no such peak with the same cartridge loading and preamp.)

When playing back the RCA New Orthophonic test record sweep, you can see the response curve flatten out as you progressively arrive at the target load.  


Clearly, telegraphed loading issues are largely squelched by use of voltage matching and buffer amps.  But each input still presents a load, for that distance of interconnect, just the same.  And, unless output Z is wery close to 60 Ohm and the destination Z is wery close to 20k Ohm, the reactance peaking will not necessarily perfectly cancel out the low pass effect of the output Z (and cable Z) and the cable capacitance.  Sometimes it seems there's too much reactance in a given handshake!  

The ability of driving amps to signal a given bandwidth flat, with a given transient performance (slew rate?)  to receiving stages is not uniformly constant across all designs.  This is compounded when incorporating equipment, the design of which is based on power matching signaling, with equipment designed for modern voltage matching signaling (or power bridging).   For example, the Manley vari-mu input Z is only 600 Ohm.   And the ATR-100's overbridge also is 600 Ohm (in/out), unless you remember to use the termination switch (where indicated), or use the unbal out from card cage.


Just some more holiday grist for the mastering mill.



Cheers
Logged
Space Camp Mastering

laarsoglethorpe@gmail.com


"Lack of DSD post filtering fried a lot of expansive twitters." (D. L.)
Pages: 1 2 [3]  All   Go Up