tell me your experiences with the elliptical equaliser

[peter sikking]

hey everybody,

I am peter sikking and I design and build studio gear at 51dB audio.

I was researching (the modern use of) the elliptical equaliser over on GS
—in order to design new one from the ground up—and I got told two things:

1. very few vinyl cutters over there, try at Lathe Trolls;
2. an elliptical equaliser is something that vinyl cutters rather avoid using.

combining the two above, I want to ask you: what don’t you like about
using an elliptical equaliser?

thanks, —ps

[Phinster]

it solves more vertical problems than it creates...usual crossover points are 60 , 150, and 300 hz

[dmills]

The thing is a fix for the technical limitations at low frequency, but it does effect the sound obviously, so you only use it if you have to, bit like the acceleration and velocity limiters.

These things are necessary because both the lathes and the replay doings have some hard geometric and mechanical limits on what is possible, and you need to respect them if you intend to get a playable disk, this does not imply that having to use them is a good thing, and you could almost see a dynamic limit on low frequency S component as possibly being interesting, only limit the out of phase (vertical) low frequency modulation if it becomes problematic, bears thinking on.

Wayne Kirkwood designed a modern second order version that gets the required first order allpass response that you need to make things sum back together correctly right. There is a circuit on the Pro Audio Design Forum.

[peter sikking]

thanks to Phinster and dmills for jumping in.

I have got an understanding of the technical side of the EE—keeping the context of the RIAA curve
and the transducer nature of cutter heads in mind—and found another approach of doing EE with
much more control.

but that is actually quite meaningless without a good understanding of the daily practice
of the vinyl cutter engineer in the year 2021—where it comes to dealing with (too much)
vertical amplitude.

and this is where this thread can be a resource for me and other circuit designers,
to find out what really matters to you and your clients. and to come up with something
better than the good old elliptical equaliser.

so here are a couple of questions for you:

is modern (computer) music much more demanding in the bass region,
with more vertical component?

are your clients more demanding today where it comes to bass?

when thinking of the overall goal of ‘a really well-cut slab of vinyl’—for
you and your client—what are really the pain/regrets when you have
to use an elliptical equaliser?

thanks, —ps

[misjah]

is modern (computer) music much more demanding in the bass region,
with more vertical component?

>> not really. i've had plenty of jazz or classical that was way too wide to cut without EE

are your clients more demanding today where it comes to bass?

>> nobody likes their bass cut off. not then, not now.

when thinking of the overall goal of ‘a really well-cut slab of vinyl’—for
you and your client—what are really the pain/regrets when you have
to use an elliptical equaliser?

>> in ideal situation you don't need to use EE at all...any use of EE is a regret. it never adds anything IMO.

[dubcutter89]

an elliptical equaliser is something that vinyl cutters rather avoid using.

...and produce records that skip/do not track on playback?? I'm not sure about that...

what are really the pain/regrets when you have to use an elliptical equaliser?

it will alter the stereo image (because that's what it should do!) and only rarely improves on the source...
almost like: if you need it you'll hear it, and if you don't hear it you prbably don't need it (or it has no effect since there is almost no vertical bass...)


But I don't want to discourage you - maybe start and play around with some processing in DAW (since cheap + easy + everything possible there) and have a listen (or even better conform your results cutting a record!)

Lukas

[Phinster]

Cut a lot of rock and dance sides...over the years mixers have learned to center the bass track, so you don't need to use the
EE or use the lower frequency of 60 hz..as misjah says , you get music with a wide sound field, so you have to use it..factories don't like it when you send in a side with a lot of 6 mil vertical excursions..very few people seem to be aware
of the effect it has...maybe get two people a year who say don't use the EE and I have no regrets using it when the program
demands it!

[peter sikking]

thanks for the replies; we are getting closer to (me understanding) the heart of the matter.

it will alter the stereo image (because that's what it should do!) and only rarely improves on the source...
almost like: if you need it you'll hear it, and if you don't hear it you prbably don't need it (or it has no effect since there is almost no vertical bass...)

OK, this is the unavoidable, this-is-how-it-works, primary effect. to reduce the vertical amplitude by a certain amount
at a given frequency is to reduce the stereo width at that frequency by the same amount.

but then there seem to be side effects, which could be avoidable, or reduced.

nobody likes their bass cut off. not then, not now.

are you saying that using an EE has a side effect of (subjectively) reducing the bass impact?


a second side effect that gets mentioned is the phase shift that is introduced in the vertical/side component.
seems to be especially noticeable at the frequencies where the EE filter starts to work.

can some of you tell me what this side effect sounds like, to the ears of seasoned cutting engineers?
any more side effects to using an EE?


another thing:

Cut a lot of rock and dance sides...over the years mixers have learned to center the bass track, so you don't need to use the
EE or use the lower frequency of 60 hz

I read that as: “use the EE with the lower frequency of 60 Hz.”

this setting gives protection for the vertical component against the doubling below 50Hz of lateral & vertical amplitude
(caused by the RIAA curve levelling off at 50Hz), right?

[dmills]

are you saying that using an EE has a side effect of (subjectively) reducing the bass impact?

If the Bass is out of phase of course it does, that IS rather the point after all.
However the EE is classically a first order network so it tends to have an effect way above whatever you have dialled in as its 3dB point, and IMHO it is usually the fact that an EE at 100Hz is still screwing with the width at 250Hz that is the main issue with the things.

this setting gives protection for the vertical component against the doubling below 50Hz of lateral & vertical amplitude (caused by the RIAA curve levelling off at 50Hz), right?

The 3180us time constant in the RIAA curve has little to do with it, 50Hz is just low enough that most of the EE networks effect has vanished by the time you get to audio that provides serious directional cues to humans (few hundred Hz or so).

I suppose that today it IS possible to build a linear phase EE, but at the low frequencies in play I am not sure that the pre ringing is less objectionable.

I wonder about doing a dynamic EE where the thing is really a frequency dependent limit on the low frequency S component, might be tricky to make it not pump the width however!

For the most part, given sane masters done by someone who understands the medium, the EE is the least of the issues that Vinyl has, and if you get handed some super wide thing mastered for CD you are probably stuffed anyway you cut it.

[Aussie0zborn]

How about focusing on a vertical amplitude limiter?

[dmills]

Trick is to avoid it pumping, because limiting S will cause the width to vary, but to some extent whatever you do is going to impact the audio, question is what hurts least.

The problem I see with dynamic EQ things down in the LF region is that as the gain on the low passed bit of S changes, so does the phase shift when you put the whole thing back together, and phase shifts that change are drastically more audible then static ones.

I think if I was going for something new in vinyl electronics I would probably go after tracing error, there was a lot of work done on this back in the day but the tech was inconvenient to implement, today it is a modest DSP problem. See the AES compendiums on disk recording (Well worth the price of admission) for some fascinating papers around this stuff.

[Discomo]

I never use any elliptical eq or limiter.
Just go through your wav files visually, adjust volumes and eq manually and you'll get a much more natural result.
Basically the basis for good cutting lies in a good master, good ears and patience. Instead of shortcuts like elliptical eq's or limiters.

[Phinster]

CBS Labs invented a device back in the 60's called an ASRA followed by a QUASRA which sorted out of phase bass etc.
I t had 3 limiters and mystery circuitry which rolled down a curve...it assumed that LF direction was difficult to determine
CBS engineers avoided using it too!

[peter sikking]

the new replies are really helping to get a grip on the whole issue,
but there is one piece of information that is missing:

there must be an overall hpf on the stereo signal.

I certainly cannot fathom that the cutter electronics are DC-coupled
from the input to the cutter head.

so there must be a default hpf in the cutting system, the lowest
you can cut. what are the numbers on that today (Hz, filter order)?

this is an important puzzle piece, so please fill me in on this.

thanks, —ps

[dmills]

Within the lathe itself it is usually just a combination of the various inter stage coupling caps and their load resistors, maybe a few Hz or so, pretty standard for audio doings.

Now the mastering console, that will likely have a switchable HPF, and maybe a switchable lowpass as well), there is little point in trying to cut down below 20Hz or so, but as ever you want the corner frequency a decade lower to avoid phase shift in the audible band if you are not deliberately filtering.

The VG74 went from a "Lower Boundary Frequency" of 7Hz to less then 2Hz in the VG74 B version, with the phase shift at 20Hz from about 70 degrees to less then 20 degrees. See Neumann announcement and conversion information 82 192 dated February 1982.
Looking at the schematics, the time constants are mostly around half a second, give or take.

Regards, Dan.

[dubcutter89]

so there must be a default hpf in the cutting system, the lowest you can cut. what are the numbers on that today (Hz, filter order)?

The default hpf in cutting systems may be different from system to system. Some have it maybe at 20Hz or so, others may be DC coupled...
But what has not changed in the last 100 years is that this super low frequencies will cause tracking issues! This is due the fact that the playback tonearm+cartridge is a spring-mass system and typically has a resonance frequency in the range around 10Hz (below audio band and above turntable revolution to be able to follow the groove on a more or less warped disk).

If you need a hpf or not depends on the individual system + source material to be cut...no default, no lowest you can, all depending on various factors

Lukas

[peter sikking]

thanks mills and dubcutter89 for giving me a picture of the low end limitations.

I have got a new appreciation for how all kind of factors determine the low-end spectrum
that is actually cut on the record. also a new appreciation for the decisions, both practical
and musical, that the cutting engineer must make.

luckily there is something fixed to rely on in this complex cluster: the ‘around 10Hz’ resonance
of the playback tonearm+cartridge that cannot be exited. thus the low-end spectrum must be
sufficiently rolled off by then. hence ‘there is little point in trying to cut down below 20Hz or so.’

and I will keep in mind that in many cases, for practical and musical reasons, the actual
low-end roll-off is (quite) a bit higher than 20Hz. I am thinking all the time of the
deep bass vs. loud vs. playing length triangular trade-off. this must have a big
finger in this pie.

with this new insight I am now ready to talk about how I think elliptical equalisation
can be done with the least amount of damage. that will be my next post.

[peter sikking]

OK, so an elliptical equaliser is something one only wants to use when necessary.
but if it is necessary, let’s have a look at how to do the least amount of damage
to the music.

just to be clear: all the things I describe below are only applied to the S/vertical
component, unless noted.

here is my list of ‘rules’ for doing the least damage:

1. use just enough dBs of attenuation to get the job done.
   mindlessly rolling off causes stereo narrowing and phase shift that are completely
   unnecessary, within and above the filter band.
2. use the lowest filter frequency that gets the job done.
   put that ‘just enough attenuation’ at the frequency where it is needed,
   so that the side effects above the filter band are pushed down to lower frequencies.
3. going from 500 to 20Hz, the amount of attenuation is either constant or increasing.
   I was taking rule 1&2 to their logical extend (i.e. graphical EQ) and then thought:
   maybe it is undesirable that going from 500 to 20Hz the EE narrows stereo field,
   then narrows it less, then narrows it more, less again, then more again.
   this is pure intuition from my side, so let me know.
4. below 50Hz the cutting characteristic turns from constant amplitude to constant
   velocity; thus maybe some extra attenuation is needed below 50Hz.
   I say maybe because at the same time the low-end roll-off of the program material
   (as adjusted by the cutting engineer) come into play. maybe it is at 20Hz, maybe
   above 50 (no bass instruments on the record?), or somewhere in between.
5. no attenuation or filter frequencies that are variable with time (i.e. dynamic EQ,
   or multiband compression/limiting).
   avoid pumping/slowly morphing of the stereo width, or phasing effects modulating the
   program material.
   I can see the EE only being adjusted for every song, or for every (musically
   vastly contrasting) section of a song.

and that’s it. what do you think? after taking your feedback into account,
we can move on to an EE design that implements these rules.

[dmills]

Add to that a high enough order that the effect is mostly gone by the time we get up where stereo direction cues start to matter.

This IMHO was always the weakness of the first order version of the thing, if it is narrowing things to mostly mono below 80Hz, it is still messing with the stereo width up at 300Hz! Trouble is, higher order filters equate to more significant changes in group delay, so it is a bit of horse trading in design.

I would note that the 'no dynamic processing' MIGHT be over stating it because most cutting chains have frequency selective limiting up above 2.2kHz or so anyway, to deal the HF limitations (IMHO more of a problem then the bass most of of the time), it would take some experiments to see what is least objectionable in an EE design, first/second/third order, static or limiting operation. It is also not clear to me that the EE really belongs as part of the lathe so much as part of the mastering console.

On the shelf Vs lowpass thing, remember that the way this works is a little odd when it comes to summing in something like an MS matrix:

Consider if at some frequency we have low shelfed the bass in S down by say 10dB, (Ignoring the phase shift to save on trig, but it makes the point), and BASS is full out of phase so M = 0, S = 1, then our 10dB reduction in S (reducing the amplitude by a factor of 3), has in this case only reduced the amplitude in either speaker by 10dB, so far so good, but what happens of this bass hit was just in the left?

M=L+R = 1, S = L-R = 1....
S/3, so S = 0.3 near enough.

L= (M+S)/2 = (1.3)/2 = 0.65, R = (M-S)/2 = 0.33, so overall bass level has dropped ~6dB but separation has also dropped to about 6dB.
The EE is brutal on out of phase content (which is what we very much want), but is unfortunately ALSO brutal on uncorrelated signals.... Might be a hint there about how to get clever.

I can maybe see the virtue in an EE with a level control, to set the amount of shelf, but I think that probably as you drop the shelf you will quickly reach the point below which it may as well be a simple high pass.

It should be possible to have some software do the calcs to figure out exactly how much EE you need at each of the possible frequencies, still going to take a human to decide which combination makes sense however.

For me, tracing errors are actually a much more interesting target to go after, there is extensive discussion in the AES anthologies on disk cutting, and today we have the technology to trivially do what CBS and Decca tried and mostly failed to do back in the day.

[peter sikking]

before I jump in: thanks for the thoughts!
I will not in this post go into my design too much, but more
give my thoughts on the other things brought up.

Add to that a high enough order that the effect is mostly gone by the time we get up where stereo direction cues start to matter.

This IMHO was always the weakness of the first order version of the thing, if it is narrowing things to mostly mono below 80Hz, it is still messing with the stereo width up at 300Hz! Trouble is, higher order filters equate to more significant changes in group delay, so it is a bit of horse trading in design.

what I am trying to make use of in the design is that when applying a shelf/roll-off (they are closely related)
at less than full strength (half, or a third), then the unwanted narrowing in the midrange is also half/a third and
the phase shift everywhere is also half/a third.

actually second-order filtering is part of the design—when needed.
when we get there, there will be graphs to check it out.

I would note that the 'no dynamic processing' MIGHT be over stating it because most cutting chains have frequency selective limiting up above 2.2kHz or so anyway, to deal the HF limitations

gotcha. I should have said explicitly that I am only considering EE below 500Hz.
that is the scope of this ‘project.’

It is also not clear to me that the EE really belongs as part of the lathe so much as part of the mastering console.

I am agnostic in that regard (unless there is a practical implication; let me know).
it does not change the fact that an EE needs to be double-stereo (4 channels)
with ganged controls, right?

On the shelf Vs lowpass thing, remember that the way this works is a little odd when it comes to summing in something like an MS matrix:

I have been doing some simulations in the weeks before starting this discussion.
one of the things I did was go in 5 steps from fully mono to signal-in-one-channel-only,
and from there in 5 steps to fully out of phase. interesting stuff. when we get further with
the design I will show those graphs here and we can share what we think.

It should be possible to have some software do the calcs to figure out exactly how much EE you need at each of the possible frequencies, still going to take a human to decide which combination makes sense however.

fully agree.

For me, tracing errors are actually a much more interesting target to go after, there is extensive discussion in the AES anthologies on disk cutting, and today we have the technology to trivially do what CBS and Decca tried and mostly failed to do back in the day.

I guess that technology is digital?