How do you judge an EQ?

[Christian Budde]

Christian, I would be pleased to make some tests with my hardware analog EQs if you need me to do so.

You can download my tool for capturing impulse responses here: http://www.savioursofsoul.de/Christian/SDC.zip
It uses ASIO for I/O. The deconvolver is the same I used in the stand alone version. Unfortunately there seems to be a problem with using the "gear" to deconvolve DA->cable->AD on the fly. Try it out by recording the same signal with "gear" (first the reference and then use this reference as gear). If you don't get a single spike, then something is wrong. Also the repetitions sometimes makes problems in case of cheap build in soundcards (if you use it with ASIO4ALL and such)
The current version is limited in the way, that either you may have lots of repititions (to reduce the each repetition (theoretically) by 6dB) or use a long sweep. Both improve your SNR, but they are not allowed at the same time.

It would be nice, if you send me the recorded impulse responses (and maybe publish them here).

Kind regards,

Christian

P.S.: If someone can provide me a set of IRs, than I'll donate an unregistered version.

[poonna]

Personally I think this discussion is better off in the DSP Developers forum instead of here in the Effects section. I think Poonna (the original starter of this thread) has lost interest after page 2.

Well, I'm just chiming in to let you know that, although I don't quite get all the deep technical stuff here, I do enjoy having it discussed and I can actually gulp some. So please go on as it is. It's really a very good educational thread and I've learned quite a lot.

Poonna

[jah1234]

Christian, I would be pleased to make some tests with my hardware analog EQs if you need me to do so.

It would be nice, if you send me the recorded impulse responses (and maybe publish them here).

Yeah, publish them here!

[BertKoor]

I think Poonna (the original starter of this thread) has lost interest after page 2.

Well, I'm just chiming in to let you know that, although I don't quite get all the deep technical stuff here, I do enjoy having it discussed and I can actually gulp some. So please go on as it is. It's really a very good educational thread and I've learned quite a lot.

In that case I'll shut up
Keep it going, guys!

[Barbarossa]

If we assume, that both are phase-minimum EQs and the filters are similarily adjusted, what kind of effect is this, that smoothens/smears the transients?

The assumpition is wrong. The mystery EQ (I know what it is but can't reveal), is a design similar to manley massive passive, far from minimum phase. It's the kind of design where the phase error was in full knowledge *left in* and even emphasised. This way we can (even algorithmically) preserve transients as well as technically possible. The trade off - well audible in the mystery example - is an odd smootheness of the EQ sound, in form of emphasised phase warp.

Hm, i don't understand, how that fits to the sound examples: the WavelabQ IMO somehow smears the transients, while the other EQ leaves them more intact.
If you say now, that the other EQ is not a phase-minimum system, then the preservation of the transients would not depend on the amount of phase shift. But that must be wrong.

Yes filter phase errors can sound *good*.

That's what i say all the time, but what doesn't fit now, is that the WavelabQ shows more transient smearing, and you say that the MysteryEQ is not phase minimum. So either the WavelabQ has even more phase-shifting and therefore isn't a phase-minimum EQ, too, or the phase-shift isn't responsible for the smearing.
Confused.

[Zaphod (giancarlo)]

phase-shift isn't responsible for the smearing exactly as magnitudo imperfections are not responsible for the smearing exactly as short irs are not responsible for the smearing exactly as (put here other possible things).... Wrong calculations and roundings are responsible for the smearing. This is my theory.
Now hit me.

[martian]

I'm always just a hands-on/"I want results" kind of guy and I'd rather get quick results than try to fight with some plugin that COULD get the same results after understanding completely how it works and what one should do exactly to get those results. Sometimes it might be even impossible. I'm sure some of the magic of old analogue EQs is that they have been designed for specific tasks and frequency ranges that the human ear/brain likes. This is why I'm less interested in orange vs orange comparison but rather "EQ1 that offers bla bla bla bla" versus "EQ2 that offers bla bla bla". I like comparing equalisers that bring something to the mastering table as it's a specific task (some are for instance excellent at boosting the high frequencies without being harsh and some are excellent at cutting away offending frequencies at a narrow band).

this i can certainly emphathise with, but ive been dissapointed in the digital world, each time to realise its the same physics i was just being fooled/guided. i feel once i understand why something seemed better i can apply that knowledge with more control. so actually being fooled so much has helped me understand what i like a lot. but i dont like not knowing or understanding.

Here is another little a/b comparison. This time I tried to make it more apples vs apples by closely approximating Wavelab Q eq with the mystery eq. I used whitenoise and voxengo span + my ears to get as close as I could on the music clip (without spending many hours on it). It got pretty close and I think in this clip it's rather obvious that Wavelab Q kills the transients in a bad way (this could be due to aliasing, I don't know).

Wavelab Q

"Mystery" eq

Cheers!
bManic

thanks very much for the examples!!

had a quick listen and sure Q sounds poorer to me, the bass seems undefined and muddy, hi hats smeared.

but my problem lies in your comparison, for general eq comparison its fine. but you treat Q like its an example of shit digital eq, fair enough maybe it is. but this is what i want envidence of, since compared to other digital biquad i cant find it myself.

i'll do some better examples.

[Christian Budde]

had a quick listen and sure Q sounds poorer to me, the bass seems undefined and muddy, hi hats smeared.

You can clearly hear that "hi hats smeared" and that "the bass seems undefined and muddy" on the other hand - if you compare the frequency response of both EQs against each other - the mystery EQ still gives about 1.5dB more bass and 1.5dB more highs (while having a dip around 200Hz). Ok, this is not apple/apple, but it would be interested to hear if the mudding/smearing is still there if the frequency response is corrected to be exactly the same. Anyway, I have other things to do right now and I won't find lots of time to visit the forum again.

Thanks for the discussion. I've learned a lot! Especially regarding usage.

See you in some month,

Christian

[bmanic]

Actually, I made sure that Q was a bit lower with the bass and highs as it sounds even worse when I boost them more! I didn't want to end up with a situation where Q boosts more, hence more shrill. Trust me on this one. Q can't beat this mystery eq.

Anyways, I agree with martian very much that it might be interesting to compare true biquad versus biquad. I'll try to do the same curve with Nyquist and posihfopit. My guess is that both fare a bit better than Q simply because of higher internal resolution/precision.

Cheers!
bManic

[Zaphod (giancarlo)]

right christian; but 1.5 dB doesn't justify what is happening here.

Other point: on some hardware gear, when you change 200Hz (it's an example), sometimes also 20 Hz and 20000 Hz are affected a bit. This is a musical behaviour, for little changements (and for someone) this could be great. So an hardware equalizer could never be compared with software, because there is ALWAYS something wrong with its response. But you could remedy. )))))
Add 1.5 dB with your electric-q to the wavelab example. Tell me if they are even closer....

[Zaphod (giancarlo)]

Actually, I made sure that Q was a bit lower with the bass and highs as it sounds even worse when I boost them more! I didn't want to end up with a situation where Q boosts more, hence more shrill. Trust me on this one. Q can't beat this mystery eq.

Anyways, I agree with martian very much that it might be interesting to compare true biquad versus biquad. I'll try to do the same curve with Nyquist and posihfopit. My guess is that both fare a bit better than Q simply because of higher internal resolution/precision.

Cheers!
bManic

Ok. I'll try to make an example for tomorrow.
We'll match IRs before, so it will be orange with orange and we could focus better on this 'transients' thing.
It's good even a filter? A not labeled example and you judge? On highpass filters the difference is very easy to understand.

[martian]

edit: bettering test files.

[Kingston]

If we assume, that both are phase-minimum EQs and the filters are similarily adjusted, what kind of effect is this, that smoothens/smears the transients?

The assumpition is wrong. The mystery EQ (I know what it is but can't reveal), is a design similar to manley massive passive, far from minimum phase. It's the kind of design where the phase error was in full knowledge *left in* and even emphasised. This way we can (even algorithmically) preserve transients as well as technically possible. The trade off - well audible in the mystery example - is an odd smootheness of the EQ sound, in form of emphasised phase warp.

Hm, i don't understand, how that fits to the sound examples: the WavelabQ IMO somehow smears the transients, while the other EQ leaves them more intact.
If you say now, that the other EQ is not a phase-minimum system, then the preservation of the transients would not depend on the amount of phase shift. But that must be wrong.

Let me try to re-iterate.

1. wavelab Q is a nearly stock standard biquad (or very similar anyway). This can be considered a phase warping filter (although math-wise it's still minimum phase). Consider this your average digital EQ.

2. mystery EQ is not a biquad, and its design has even *more* phase warping.

A kind of a trade off is possible with EQs: the more phase warp you are willing to live with (within limits) the sharper the transients stay. This one of the reasons (but not the whole story) why mystery EQ preserves transients better. You can test this for example with Voxengo HarmoniEQ. The difference here is somewhat similar to the softP and phaseP modes.

The mystery EQ is also of extremely high quality, and even though it technically has far too much phase distortion, wavelab Q still sounds worse (ie worse accuracy and transient killing filter type).

Yes filter phase errors can sound *good*.

That's what i say all the time, but what doesn't fit now, is that the WavelabQ shows more transient smearing, and you say that the MysteryEQ is not phase minimum. So either the WavelabQ has even more phase-shifting and therefore isn't a phase-minimum EQ, too, or the phase-shift isn't responsible for the smearing.
Confused.

To cut some of the confusion. First of all, forget all that talk of minimum phase filters. That was just Christian Buddes (IIRC) offhand comment on algorithm behaviour.

To clear things up, minimum phase EQ generally refers to something between a linear phase and phase warping filters. They are a compromise on the two aspects. Both of these EQs can be considered phase warping filters *with no impulse response generated delay*.

I would venture to guess in these examples wavelab Q transient smear is actually due to accuracy problems. That's the problem with apples to oranges comparisons. The nyquist EQ, for example, behaves much nicer in this respect, with very little transient smear.

[Rangtangtang]

Can you use a flat response from whatever device as the basis of a EQ plug, by shaping that flat response whith a volume curve or envelope?

....just curious.....

[Christian Budde]

Ok. I'll try to make an example for tomorrow.
We'll match IRs before, so it will be orange with orange and we could focus better on this 'transients' thing.
It's good even a filter? A not labeled example and you judge? On highpass filters the difference is very easy to understand.

Out of curiousity, I already did though (otherwise I fear that I maybe can't get to sleep this night). Here are three examples. At least one is the mystery and at least one is the corrected Q. The third example is either one of both. You can clearly find the false sheep here:

http://www.savioursofsoul.de/Christian/ ... high-H.wav
http://www.savioursofsoul.de/Christian/ ... high-R.wav
http://www.savioursofsoul.de/Christian/ ... high-S.wav

Christian

P.S.: I've changed the length (cut start/end) to make it hard to be influenced by other properties.
P.S.S.: Please don't post your thoughts too early, only a "I got it" or it can be clearly determined is enough. Then everyone can see, if he hears the difference as well.


*** Edit ***
Just to add some comment here:
- I think that it's still not only the phase difference here. Also the volume and non-linearities may cause the difference (I don't know much about the mystery EQ, but at least the volume of those examples differ slightly).
- I only tried to fit the frequency response and nothing else, anyway the phase changes as well of course...