Music listening/mastering EQ

[k brown]

In the mid '80s (at the peak of my life as an audiophile) Mark Levinson started a new company called Cello. It's flagship product was a high-end and extremely expensive equalizer, of all things - something shunned by the entire audiophile community. But designed by someone who knows what they're doing and using the very best circuit components and construction methods, even the tweakiest critics had to admit that it made a huge percentage of commercial recordings much more pleasant to listen to. The Cello Audio Palette was not for room or system correction; it assumed a well-balanced system to start with - it was strictly intended for correcting poorly-balanced source material. It even became used by many professional mastering engineers as the final 'sweetening', or correcting older material for re-issue.
http://www.celloseattle.com/ctdocs/prod ... lette.html

We can, of course approximate such a device with FS, and I was very surprised to see that there is essentially no plug-in based on it! (EDIT: I finally found one - https://www.danielhertz.com/master-class-software - Mac only) One can set up a software 6-band parametric to do the job and save the settings as a preset, but I wanted a plug-and-play dedicated Cello-like EQ. Plus, the toolbox parametric is very difficult to adjust without shifting the frequency points. I'm sure someone could figure out how to adapt the graphic EQ that Spogg has used lately, but I sure wasn't able to. But even my clumsy attempt using the Martin-corrected SVF, the results were very effective. I was quite surprised that even listening with HiRes headphones, I could couldn't hear any difference with the circuit switched in and out, with all the controls flat.

Listening to a number of my own live (mostly classical) recordings, I've yet to come across one that wasn't considerably improved by this. An interesting aspect of the Audio Palette's design was that the top and bottom bands are not shelves, but are peaking EQs set to above and below human hearing (25kHz and 15 Hz). So one is hearing only the 'shoulders' of these bands as they 'slope' into hearing range. The 'Blend' position of the Equalization switch blends the stereo channels together below 40 Hz. Can often 'tighten' the bass on certain recordings.

Viola.fsm

(941.95 KiB) Downloaded 1237 times

Screeshot copy.png (113.34 KiB) Viewed 21560 times

Not being terribly imaginative, and unable to replicate the Cello's knobs and panel, I did this up in my 'Pultec' style.

I'm wondering if the ZDF multimode would be a better choice for this. I imagine Martin could probably do the whole thing up in a single chunk of code!

[k brown]

For those wanting to correct or improve my take on it, here are the specs for the Audio Palette.
One forum post lists the Q's as such:

Cello 'Q's copy.png (13.22 KiB) Viewed 21556 times

Another claims they are all 0.7. Unfortunately, none of the published Cello documents specifies what they were.

As you can see the boost/cut at the center two fr are only 6dB, and all the Qs are quite mild. As you can also see, the first specs I show are incorrect as regards the top fr - it's 25kHz, not 20. This is verified by the panel legending, which states '25 kHz'.

CELllFIG6.jpg (38.15 KiB) Viewed 21556 times

Someone with experience reading such measurements could probably deduce the Qs from this chart.

All the controls were stepped - they were actually ladders of fixed resistors, not pots. The fr controls each had 24 steps.

1.-Cello-Audio-Palette-EQ-Front copy.jpg (22.21 KiB) Viewed 21556 times

The 'Absolute Phase' control is an interesting one. It allows inverting either or both stereo channels. I actually have a couple of early stereo LPs that have one channel out of phase! Inverting both channels can be quite audible with headphones or very HiRes speakers; especially on minimaly-mic'd acoustic recordings - lifts a proverbial veil. Usually not perceptible on multi-track pan-potted pop/rock recordings. Though I once had an interesting experience with absolute phase on such a recording, back when I had a very HiRes system. The Peter Gabriel track 'Red Rain' exhibited quite different mix balances with phase reversed on both channels (i.e.: both channels inverted or non-inverted).

[Spogg]

Very interesting Kevin, even though my knowledge of filters is pretty much limited to just using what Martin makes!

The thing that came to mind was more of a philosophical point. Who’s to say what is the best mix or balance? It must depend on personal preference of either the mix or mastering engineer or the listener. And the listener’s hardware and environment must surely play a part. Finally our brains tend to kinda “normalise” sound levels and spectra as we listen.

When I’m making presets it often happens that I’m happy with one but when I go back to check it out, maybe the next day, it sounds less agreeable than before and I have to tweak it to give back the impression I had when I was making it.

There seems to be so many factors at play it’s surprising that anything is ever finally released and I can understand the audiophile approach of no EQ or everything set flat. Ideally I think a listener wants to get close to what the mastering engineer and the artist heard and approved of.

Sorry for the waffle!

[k brown]

Personally I prefer pancakes.

The original intent of the Audio Palette was for home listening; to make somewhat less well-balanced recordings more pleasant to the listener - no claim of what's right or wrong, only set it to please yourself. Through many years of 'active listening' I was seldom satisfied with either the purist 'no tone controls' or the limited abilities of the usual bass and treble controls. This was exactly how Levinson felt and why he adapted Burwen's research to a product that became the Cello Audio Palette. It was only later that mastering engineers began to use it.

Proof is in the listening - play some recordings you're familiar with, and have a twiddle; bet you anything that more often than not you will prefer an adjusted sound to the flat. You listen a while, make a small adjustment; listen some more make a few more - after a bit you get curious and switch off the EQ, and go "yuck"!

[martinvicanek]

It is quite straightforward to extract the Q values from the curves. If f0 is the center frequency, then determine the frequencies f1 and f2 to the left and right of f0 where the transfer curve attains half the peak (or dip) height in dB. Then Q = f0/(f2 - f1).

Example: for the trebble boost curve f0 is 20 kHz and the peak height is 20 dB. The curve goes through 10 dB at f1 = 7.5 kHz and f2 = 50 kHz. Hence Q = 20/(50 - 7.5) = 0.47.

A quick estimate indicates that all Qs in the graph
are around 0.5.

If you build this thing in digital make sure to use some sort of matched peaking filters (not plain RBJ) at least for the trebble control.

[k brown]

Thanks for weighing in Martin - would that be the same as the ZDFs? I was thinking of trying this with the Multimode stereo ZDFs. The way it's all connected up, it doesn't utilize the BP outs, but uses LP and HP in series for the BPs.

[lalalandsynth]

You can buy one for cheap on Reverb
https://reverb.com/item/7592795-cello-audio-palette-eq

[martinvicanek]

Here is an emulation of the EQ part, according to the transfer curves in the first post. Kevin, please check thre transfer function of your Viola, I believe there is an issue with negative dB settings (cuts).

[k brown]

I'm surprised that's the only issue!; thanks for having a look.

Can't wait to try 'Cellino'. Sounds like a tasty sugu.

I had a hunch those ZDFs were the right components for this - I just didn't know how to implement them; thanks TONs for doing this. Now I can really do this up right!

P.S. - the Cello's top fr was actually 25kHz - would there be any problem in pushing it up that high?

[martinvicanek]

From the transfer curves, the top frequency would seem to be 20 kHz for boost, however 25 kHz for cut. The peaking filter matching scheme only works for center frequencies within the Nyquist range, which is < 22.05 kHz for a sampling rate 44.1 k. Therefore, II would leave it a 20 kHz.

BTW you can do the matching for any filter topology (direct forms, etc.). I recommend ZDF for its superior performance truncation error wise at low frequencies (where matching is actually less imposrtant).