EQ Hi-Quality = Bad. (for somethings)

Discuss music production with Ableton Live.
Angstrom
Posts: 14928
Joined: Mon Oct 04, 2004 2:22 pm
Contact:

Post by Angstrom » Sat Mar 22, 2008 3:49 am

we are witnessing the first stages of the redundancy of the human.
Robert is now replaced by a video of himself, which comments on threads for him.

Nogi
Posts: 480
Joined: Fri Dec 16, 2005 2:18 am
Location: C@L

Post by Nogi » Sat Mar 22, 2008 4:02 am

I think what we are intended to glean is the increase in Master level (pushing it into the red) as we switch to HQ mode even with all EQ bands off (but the device on). Some may interpret this to mean we are getting modified output with what may appear to be an inactive EQ. Of course, this isn't a particularly surprising result if you consider the nature of oversampling.

However, this is also dependent on your audio settings. If you are at 96kHz nothing much will happen.

Tone Deft
Posts: 24152
Joined: Mon Oct 02, 2006 5:19 pm

Post by Tone Deft » Sat Mar 22, 2008 4:11 am

right again Nogi, thanks I see it now just over a 5dB gain. freaky, my guess is that it's a bug.

8)
In my life
Why do I smile
At people who I'd much rather kick in the eye?
-Moz

rhythminmind
Posts: 333
Joined: Fri Aug 13, 2004 6:34 pm
Location: Los Angeles
Contact:

Post by rhythminmind » Sat Mar 22, 2008 6:27 am

Thanks Nogi. So the video clears up the fact that it's an upsampleing EQ. I have to say i don't like resampling algorithm thats going on with EQ8. Artifacts are unavoidable with resampling but this is more then expected. Also listen to what happens when you bring the highpass all the way to the max.
in normal mode it's a very tight slope. In HI-Q it lets allot more pass. Why that is i'm not sure. You would think it would be even tighter with the added freq range in HI-Q but its not the case.
http://rhythminmind.net | http://signaltonoize.com | http://popsound.com
Image
"Any experiment of interest in life will be carried out at your own expense."

rhythminmind
Posts: 333
Joined: Fri Aug 13, 2004 6:34 pm
Location: Los Angeles
Contact:

Post by rhythminmind » Sat Mar 22, 2008 6:57 am

I want to clear up that this is not a complaint. It's a why? I think EQ8 is a nice EQ & HI-Q mode with bell bands has it's place. But HI-Q mode with HP/LP setting is just odd. Also the quality of the oversampling is in question. But i have to say in "normal mode" it performs wonderfully. It's not my goto musical "colored" EQ. I use EQ8 as a utility. I use it for the HP/LP, making FX like my divider, very surgical things. This is why i have these quality/whats going on under the hood questions.
http://rhythminmind.net | http://signaltonoize.com | http://popsound.com
Image
"Any experiment of interest in life will be carried out at your own expense."

Nogi
Posts: 480
Joined: Fri Dec 16, 2005 2:18 am
Location: C@L

Post by Nogi » Sat Mar 22, 2008 7:52 am

rhythminmind wrote:Thanks Nogi. So the video clears up the fact that it's an upsampleing EQ. I have to say i don't like resampling algorithm thats going on with EQ8. Artifacts are unavoidable with resampling but this is more then expected. Also listen to what happens when you bring the highpass all the way to the max. in normal mode it's a very tight slope. In HI-Q it lets allot more pass. Why that is i'm not sure. You would think it would be even tighter with the added freq range in HI-Q but its not the case.
When you say 'artifacts', I presume you are referring to the difference between the oversampled audio running a set at 96kHz and the upsampling of HQ mode while running your set at a lower samplerate. Let's test that...

1 - I took a 16/44.1kHz white noise sample into a project at 32/44.1kHz. Added a HP filter in EQ8 and rendered at 32/44.1kHz.

2 - Then, took the same sample and rendered with EQ8 in HQ mode. The difference between 1 and 2 was the result of the fact that the HQ filter was preserved and the Normal mode filter got squished against the Nyquist frequency. What you are referring to as a 'tighter slope' is deformation in the Normal mode filter - in this case the HQ filter is the more clinically correct as it is able to preserve itself well past the limits of human hearing.

3 - Took the same noise file into a set at 32/96kHz added EQ8 and rendered at 32/44.1kHz. Difference between the output of 2 and 3? Nothing. Perfect phase cancellation.

The artifacts as I've defined in my premise in this specific case either don't exist or cancel exactly between the 44.1->96 upsample and the 96->44.1 render.
rhythminmind wrote:...But HI-Q mode with HP/LP setting is just odd. Also the quality of the oversampling is in question. But i have to say in "normal mode" it performs wonderfully. It's not my goto musical "colored" EQ. I use EQ8 as a utility. I use it for the HP/LP, making FX like my divider, very surgical things. This is why i have these quality/whats going on under the hood questions.
The strange thing here is that what's going on under the hood is moot when you are using the EQ8's filters to split frequency bands. Any coloration or artifact, real or imagined, will exactly cancel by phase inversion and is thus inaudible (the only evidence of its existence being the meters peaking higher than you'd expect). As I said in your other thread, and believe me I've been down this road a LONG way already, if what you are really after is a steep filter for surgical editing in a multiband device you should use EQ3 with the 48dB setting. It's coloration cancels out and it becomes a transparent brick wall.

Appreciate your curiosity. Rock on.

rhythminmind
Posts: 333
Joined: Fri Aug 13, 2004 6:34 pm
Location: Los Angeles
Contact:

Post by rhythminmind » Sat Mar 22, 2008 8:44 am

I'm not on crack.:) It's more then just level issues from resampling
Nogi try this... Run @ 44.1 play your noise sample with a high pass as high as it will allow. Under Normal mode it's at Nyquist 22kish not audible. Put it in HI-Q it's now audible. The slope also falls back.
Thats what i don't understand
http://rhythminmind.net | http://signaltonoize.com | http://popsound.com
Image
"Any experiment of interest in life will be carried out at your own expense."

Nogi
Posts: 480
Joined: Fri Dec 16, 2005 2:18 am
Location: C@L

Post by Nogi » Sat Mar 22, 2008 9:16 am

See '2' above.
Last edited by Nogi on Sat Mar 22, 2008 9:38 am, edited 2 times in total.

rhythminmind
Posts: 333
Joined: Fri Aug 13, 2004 6:34 pm
Location: Los Angeles
Contact:

Post by rhythminmind » Sat Mar 22, 2008 9:25 am

Yes the frequency range is extended at 96k. But that doesn't explain to me why the slope starts lower. If anything it should be higher. The HP slope starts @ around 10k when maxing it out in HI-Q mode or 96k. In normal @ 44.1 it's so high it's beyond hearing.
http://rhythminmind.net | http://signaltonoize.com | http://popsound.com
Image
"Any experiment of interest in life will be carried out at your own expense."

Nogi
Posts: 480
Joined: Fri Dec 16, 2005 2:18 am
Location: C@L

Post by Nogi » Sat Mar 22, 2008 9:51 am

rhythminmind wrote:Yes the frequency range is extended at 96k. But that doesn't explain to me why the slope starts lower. If anything it should be higher. The HP slope starts @ around 10k when maxing it out in HI-Q mode or 96k. In normal @ 44.1 it's so high it's beyond hearing.
Study the LP filters. Set them as low as they will go. Now toggle between HQ and Normal. Now turn the frequency up. The Normal mode filter changes slope drastically as it approaches Nyquist thus not preserving the original filtering. The HQ just slides over as it is nowhere near its limit.

I think that might be as clear as I can make it.

rhythminmind
Posts: 333
Joined: Fri Aug 13, 2004 6:34 pm
Location: Los Angeles
Contact:

Post by rhythminmind » Sat Mar 22, 2008 10:24 am

Ok things are clear to me. Thanks.
Theres now a functionality question for the ableton guys.
Lets talk usage now.
@ 44.1 normal mode We can do a highpass sweep way past 10k. This is a very audible and common effect.
If we wanted to do this @ 96k we can't. Not without stacking bands.
Yes i know there are 1000 ways to recreate this with other plugins.
But as it is now If we open a mix started at 44.1 with EQ8's and open or switch to 96k. it will sound different. & i'm not talking about fidelity.
But it's good this is now a known situation. It's the unknown that causes the problems.
At least now i wont have the "why does my mix sound off @ 96k question"
http://rhythminmind.net | http://signaltonoize.com | http://popsound.com
Image
"Any experiment of interest in life will be carried out at your own expense."

Robert Henke
Posts: 1193
Joined: Fri Jun 04, 2004 10:38 am
Location: Berlin

Post by Robert Henke » Sat Mar 22, 2008 11:28 am

Hello.
I am not sure if I got this thread in all detail. So, if my comments are totally wrong, correct me...

This is what the High Qaulity mode does: It doubles the samplerate before the EQ is calculated, then, after the EQin, it downsamples by a factor of 2.

Pros:
-it improves the behaviour when it comes to EQing high frequencies.
This can be seen directly in the filter graph. The filter graph is calculated using the same math as the actual filter. What you see there, is what you get.

Cons:
- it needs more CPU ( twice as much for the EQ, a bit for the upsamling, a bit more for the downsampling )
- it add some tiny little bit of distortion due to the downsampling.

So, as usual in digital audio and in real life, there is no black versus white here.
It all depends on what you want to do. Here is my personal opinion:
a) in most cases you will not need the High Quality mode, unless you EQ things above say 5 kHz. If it does not make a visible difference in the graph, it will not make much of a difference in the signal. In doubt: trust your ears, what sounds good IS good, regardless of any theory.
b) in the master i would use it
c) the distortion of the downsampling is theoretical. It is a magnitude below any dithering. ( did not test, but would assume far less than -120dB in worse case )
d) "High Qauliity" is a tricky term. But it is more understandable for most users than "Oversampling" and this is why we use it here.

Cheers, Robert

Nogi
Posts: 480
Joined: Fri Dec 16, 2005 2:18 am
Location: C@L

Post by Nogi » Sat Mar 22, 2008 7:29 pm

rhythminmind wrote:Ok things are clear to me. Thanks. Theres now a functionality question for the ableton guys. Lets talk usage now. @ 44.1 normal mode We can do a highpass sweep way past 10k. This is a very audible and common effect. ... But as it is now If we open a mix started at 44.1 with EQ8's and open or switch to 96k. it will sound different. & i'm not talking about fidelity."
True, they sound different. Normal mode becomes a very aggressive HP filter at high freqs and would be a better choice for dramatic sweeps or DJ style transitions than HQ. Similarly, in your case it is a better choice if you need a filter to isolate frequency bands by phase inversion as HQ mode won't cut enough at really high frequencies as it retains its 6db/8ve behavior across the spectrum.

Also, I'll grant you the second part as well. Namely, that it is not well known just how different these modes are at high freqs. This is something users should bear in mind if they work at different samplerates often. Even the manual seems to be in error regarding HQ mode.
The Manual wrote:Enabling Hi-Quality mode via the (PC) / Ctrl (Mac) context menu can improve the sound quality, particularly with low frequency signals or when working at high sample rates.
In fact the higher your working sample rate the less impact HQ mode will have. (The manual could also do with more detail in general. An innovative product like Live will continue to attract more propellerheads -no pun- and 90% of these inquiries could then be headed off with a RTFM. Perhaps an enginerd version of the manual?)

I think I understand, you were expecting a HQ mode with higher fidelity whose filter behaviors are the same as in Normal mode across the spectrum. Or, at least a frequency knob that went to 48kHz or so in HQ mode to allow you to complete the 'HP sweep to silence' type move with a single 6dB filter.

Perhaps the thread title should be: "EQ Hi-Quality = Different. (for somethings)"

SubFunk
Posts: 7853
Joined: Sat Jan 28, 2006 3:41 pm
Location: A Big Toilet Called Berlin
Contact:

Post by SubFunk » Sat Mar 22, 2008 7:40 pm

Robert Henke wrote:
trust your ears, what sounds good IS good, regardless of any theory.

spot on,

this should be by far more often mentioned in all to often useless tech discussion threads.

(not saying that all tech threads are useless!)
*** Image GAFM ***

Nogi
Posts: 480
Joined: Fri Dec 16, 2005 2:18 am
Location: C@L

Post by Nogi » Sat Mar 22, 2008 7:51 pm

In this discussion in fact, the OP's ears are partly what lead him to pursue a line of investigation. In other words it is also valid to say, if it doesn't sound subjectively "good" or as expected, why not? Much to the chagrin of many, I'm sure.

Post Reply