Channel Strip EQ Woes

[DParris]

A couple of complaints about the EQ on Reason's SSL-style channel strips, which I otherwise love using. I'm sure other folks have run into the same issues. Just hoping that by making enough noise about it, we might get an improvement with 11.

First, the discrete jumps between gain amounts and frequencies are huge. I understand the difficult of trying to map a large range of numbers onto 127 values, but just do away with the adherence to midi values. I'd love to be able to double-click onto an EQ knob and type in a value--especially useful for being able to use the EQ for corrective purposes rather than just character or coloring. For folks using external midi controllers to operate the EQ, the knobs could still map to the nearest value as the controller sends midi data.

Second, the narrowest available Q is relatively wide. Authenticity be damned. There is no downside to leaving the smallest Q values available and increasing the range of values up past 2.5. Yes, I can drop an M-Class into the track, and I frequently do that. Yes, I have a crapload of other plugins that have narrower Qs. But why not make it available right in the channel strip?

Implementing either or both of these features would not take away from what the EQ already does. Broad strokes rather than surgical precision would still be there.

Just a thought.

[reggie1979]

Nope, I agree. It's a totally awesome useful feature that needs updating. For me, it's too small and doesn't use a mousewheel.

There is a thread about someone trying to make a hardware controller which would be ace, but I don't think it's happening anytime soon.

The "graphic" part of the eq is fine, but I'd rather personally "dial in" my frequency, and that is hard to do. Too much jumping, not enough fine control. As I struggle to see even the text I'm typing, it's nearly impossible to read the built-in's. I think this should be addressed, but it's just me.

[DParris]

Nope, I agree. It's a totally awesome useful feature that needs updating. For me, it's too small and doesn't use a mousewheel.

There is a thread about someone trying to make a hardware controller which would be ace, but I don't think it's happening anytime soon.

The "graphic" part of the eq is fine, but I'd rather personally "dial in" my frequency, and that is hard to do. Too much jumping, not enough fine control. As I struggle to see even the text I'm typing, it's nearly impossible to read the built-in's. I think this should be addressed, but it's just me.

Yes, I've been following Allison's hardware controller project with great interest! Still a ways out, as you say, but she's clearly capable and committed. I think it will happen.

A nice addition to the spectrum window might be to choose a gain scale when clicking and dragging nodes. For instance, having the window extend to just 12db above and below zero instead of 36db. How much freaking easier that would make everything!

[Loque]

Sometimes i cannot follow decisions to implement all disadvantages of old hardware in new software...

For the EQ itself i don't care anymore since i have ProQ. Ofc i think this should be build in a an EQ, but who would pay the extra 180 bucks?

[selig]

Just to clarify a few things:
There are 1000 frequency steps, not 127.
The gain resolution is about 0.3 dB, which is a greater degree of resolution than I've ever needed for a channel EQ (speaking only for myself here, of course).
Same for Q, plenty of resolution there.

With 1000 steps across about 3-4 octaves for each band, that's around 300 steps/octave, or 25 steps per semitone on average. That's overkill for an EQ that doesn't have super surgical Q bands, and would still be plenty of resolution even for a fairly surgical EQ.

For scientific work, I can see the need to precisely enter a very specific frequency and expect precision results. For music production that's not typically the case - musical pitches are a more likely level of precision that would be useful rather than being able to type in "103.014' Hz on an EQ - the Q would have to be incredibly narrow to ever appreciate that amount of precision! And what would be the use case for that amount of precision? Surely it wouldn't need to happen on every mixer channel, right?

For compatibility, I would suggest a "Hi Q" button on the EQ (if only there was room, maybe on the back of the rack?), I would think that would solve this issue for most users.

And yes, I too would like to type in values in Reason just for speed, but EQ frequency is probably the last place that would need to happen in my work because I would rarely if ever actually know the frequency I'm needing to EQ without listening first. But I would still support having this option Reason-wide!

[DParris]

Just to clarify a few things:
There are 1000 frequency steps, not 127.
The gain resolution is about 0.3 dB, which is a greater degree of resolution than I've ever needed for a channel EQ (speaking only for myself here, of course).
Same for Q, plenty of resolution there.

With 1000 steps across about 3-4 octaves for each band, that's around 300 steps/octave, or 25 steps per semitone on average. That's overkill for an EQ that doesn't have super surgical Q bands, and would still be plenty of resolution even for a fairly surgical EQ.

For scientific work, I can see the need to precisely enter a very specific frequency and expect precision results. For music production that's not typically the case - musical pitches are a more likely level of precision that would be useful rather than being able to type in "103.014' Hz on an EQ - the Q would have to be incredibly narrow to ever appreciate that amount of precision! And what would be the use case for that amount of precision? Surely it wouldn't need to happen on every mixer channel, right?

For compatibility, I would suggest a "Hi Q" button on the EQ (if only there was room, maybe on the back of the rack?), I would think that would solve this issue for most users.

And yes, I too would like to type in values in Reason just for speed, but EQ frequency is probably the last place that would need to happen in my work because I would rarely if ever actually know the frequency I'm needing to EQ without listening first. But I would still support having this option Reason-wide!

Thanks for clarifying Selig. I assumed that there were 127 steps, because that's what's available in the automation window, and mousing the knobs feels so clumsy and imprecise. But if you say there are a thousand, I believe it.

I'm confused about your math, though. Just checked now with the low-mid band, and in the vicinity of 440hz I could only manage around 7 steps per semitone, even with the shift key held down. I checked higher and lower octaves, and the same held true.

Don't get me wrong--that's plenty enough precision for me. My complaint about those steps being relatively difficult to access remains, however.

Edit: there are 127 steps available in gain, and you were referring to frequency. I read too quickly.

[jam-s]

Thanks for clarifying Selig. I assumed that there were 127 steps, because that's what's available in the automation window, and mousing the knobs feels so clumsy and imprecise. But if you say there are a thousand, I believe it.

Pressing 'shift' while using the mouse to dial in a value will give you much finer resolution.

[selig]

IMO, the important question becomes "how much resolution do you need?".

Correcting my original math for the frequency resolution, here's what I was referring to…
SSL EQ Bands/ranges:
Low: 40-600 Hz = E1 - Eb5 (4 octaves)
Low Mid: 200-2000 Hz = Ab3-C7 (4.3 octaves
High Mid: 600-7000 Hz = Eb5-A8 (3.5 octaves)
High: 1500-22,000 Hz = G6-F10 (4 octaves)

So 1000 steps divided by an average of 4 octaves = 250 steps per octaves.
250 steps divided by 12 semitones should give around 20 steps per semitone.

For example, on the low mid band, a value of 343 gives you 440.6 Hz
If 440 = A, then Ab (one semitone higher) = 466.1638 Hz.

For the SSL EQ low-mid band,
343 = 440.6 Hz
344 = 441.6 Hz
345 = 442.6 Hz
(skipping ahead)
368 = 466.7 Hz

368 - 343 = 25 steps of resolution per semitone.


With a bandwidth of one octave = a Q of 1.4
The SSL EQ has Q values from 0.7 to 2.5, which equals a bandwidth of 2 octaves to around 1/2 octave.
When your bandwidth is 1/2 octave, you would barely (if at all) be able to discern a frequency change equal to one semitone, and Reason has 25x that resolution. Even with a Q of 32 (MClass EQ), which is about 0.04 octaves, or around a 1/4 tone) you would still have enough resolution IMO.

For comparison, the MClass EQ mid bands cover from 40-20,000 Hz, around 9 octaves each, and also have a resolution of 1000 steps. This works out to around 10 steps per semitone and no one has complained about the lack of frequency resolution.

This is because there are a very few cases where frequency precision matters to a high degree. In fact, the ONLY case I've ever encountered was when needing to remove a feedback squeal (which is basically a sine wave) in a live recording. And even in that case, you couldn't type in a value since the value is unknown.

Also important is that the frequency controls on an EQ are not linear, so there will be larger steps as you go higher, and more resolution per Hz as you go lower (to maintain a constant number of steps per semitone/octave). So while the steps for the SSL EQ are 1 Hz apart at 440, they would be spread much wider at higher frequencies (half as many steps per Hz for each higher octave).

A real world example: the Subtractor filter (and some others in Reason) only has 127 steps, but I doubt many folks ever noticed the "lack" or resolution. Even my own ColoringEQ only has 120 steps (semitones) for every band, though it ALSO has a fine tune in cents allowing 12,000 steps of resolution in total (100 steps of resolution per semitone).

The thing is, folks don't EQ by Hz because it's linear - we hear frequency non-linearly as pitch, in octaves and further divided in semitones. So while there are only 20 Hz in the lowest audible octave, there are 10,000 Hz in the highest. If the frequency knob on an EQ covered the entire audible band linearly, the middle position would be around 10,000 Hz - and this would be pretty useless since most of the fine tuning with EQ is done in the lower octaves. In contrast, the middle of the audible spectrum in octaves is around 640 Hz.

So how much resolution DO you need with an EQ? As always, it depends. It depends on the Q/bandwidth and it depends on the job at hand. I would argue that for 99% of your EQ work, semitone resolution is fine - and even that is overkill unless you're using extremely narrow cuts/boosts.

[DParris]

IMO, the important question becomes "how much resolution do you need?".

Correcting my original math for the frequency resolution, here's what I was referring to…
SSL EQ Bands/ranges:
Low: 40-600 Hz = E1 - Eb5 (4 octaves)
Low Mid: 200-2000 Hz = Ab3-C7 (4.3 octaves
High Mid: 600-7000 Hz = Eb5-A8 (3.5 octaves)
High: 1500-22,000 Hz = G6-F10 (4 octaves)

So 1000 steps divided by an average of 4 octaves = 250 steps per octaves.
250 steps divided by 12 semitones should give around 20 steps per semitone.

For example, on the low mid band, a value of 343 gives you 440.6 Hz
If 440 = A, then Ab (one semitone higher) = 466.1638 Hz.

For the SSL EQ low-mid band,
343 = 440.6 Hz
344 = 441.6 Hz
345 = 442.6 Hz
(skipping ahead)
368 = 466.7 Hz

368 - 343 = 25 steps of resolution per semitone.


With a bandwidth of one octave = a Q of 1.4
The SSL EQ has Q values from 0.7 to 2.5, which equals a bandwidth of 2 octaves to around 1/2 octave.
When your bandwidth is 1/2 octave, you would barely (if at all) be able to discern a frequency change equal to one semitone, and Reason has 25x that resolution. Even with a Q of 32 (MClass EQ), which is about 0.04 octaves, or around a 1/4 tone) you would still have enough resolution IMO.

For comparison, the MClass EQ mid bands cover from 40-20,000 Hz, around 9 octaves each, and also have a resolution of 1000 steps. This works out to around 10 steps per semitone and no one has complained about the lack of frequency resolution.

This is because there are a very few cases where frequency precision matters to a high degree. In fact, the ONLY case I've ever encountered was when needing to remove a feedback squeal (which is basically a sine wave) in a live recording. And even in that case, you couldn't type in a value since the value is unknown.

Also important is that the frequency controls on an EQ are not linear, so there will be larger steps as you go higher, and more resolution per Hz as you go lower (to maintain a constant number of steps per semitone/octave). So while the steps for the SSL EQ are 1 Hz apart at 440, they would be spread much wider at higher frequencies (half as many steps per Hz for each higher octave).

A real world example: the Subtractor filter (and some others in Reason) only has 127 steps, but I doubt many folks ever noticed the "lack" or resolution. Even my own ColoringEQ only has 120 steps (semitones) for every band, though it ALSO has a fine tune in cents allowing 12,000 steps of resolution in total (100 steps of resolution per semitone).

The thing is, folks don't EQ by Hz because it's linear - we hear frequency non-linearly as pitch, in octaves and further divided in semitones. So while there are only 20 Hz in the lowest audible octave, there are 10,000 Hz in the highest. If the frequency knob on an EQ covered the entire audible band linearly, the middle position would be around 10,000 Hz - and this would be pretty useless since most of the fine tuning with EQ is done in the lower octaves. In contrast, the middle of the audible spectrum in octaves is around 640 Hz.

So how much resolution DO you need with an EQ? As always, it depends. It depends on the Q/bandwidth and it depends on the job at hand. I would argue that for 99% of your EQ work, semitone resolution is fine - and even that is overkill unless you're using extremely narrow cuts/boosts.

Cool, I'm definitely with you on the EQ being way more precise than I originally gave it credit for. I think my frustration with operating the UI compared to some of my favorite plugins was definitely more to blame than the parameters of the EQ itself.

BUT: I'm still not getting the results that your math says should be there. I just did a check of all four bands on the EQ. Now granted, this is purely academic, because the number of steps between semitones is still plenty for any corrective purposes. I'm just trying to reconcile what I'm actually able to accomplish with these knobs vs. what the math says should be possible.

I started by roughly adjusting the frequency of each band to match an A that was comfortably in the center of that band's range. Then, WITH THE SHIFT KEY HELD DOWN, I dragged the knob as slowly and carefully as I could from that pitch to the Bb one semitone above, or as near as I could get it.

I wrote down every step along the way that the text readout was giving me, and here's what I got. The actually frequencies for A and Bb in each octave are labelled in parentheses above and below the steps I recorded.

HF
(3.52khz)
3.53
3.56
3.59
3.62
3.65
3.68
3.72
(3.729)


HMF
(1.76khz)
1.77
1.79
1.80
1.82
1.83
1.85
1.87
(1.864)

LMF
(440hz)
441.6
444.7
447.7
450.8
454.0
457.1
460.3
463.5
466.7
(466.16)

LF
(110hz)
110.7
111.6
112.5
113.5
114.4
115.6
116.6
(116.54)

Again, I'm not complaining about this level of precision--it's more than enough, and I'll just have to work more carefully with the controls to get the results I'm after.

What I'm wondering is, are there fewer steps available by operating the mouse than by automating? Something here doesn't quite add up.

I'm enjoying this discussion!

[selig]

Automating a parameter with a wide range of values gives you access to each and every parameter, while holding shift sometimes will not. Also note that some fine values will change even though the tool tip says the value is the same (for non-stepped controls) - this caused me great confusion in the past, where two knobs showed the same value but were clearly NOT the same exact value!

An issue with the SSL and automation is that the actual parameters values are not given for the automation (hence the conversion I had to do above), for example EQ frequency is not in Hz, gain is not in dB, etc.

But with automation, you can type values in directly (and get to those "in-between" values as well). I have sometimes automated a parameter strictly for the ability to type in a value directly!

[DParris]

Automating a parameter with a wide range of values gives you access to each and every parameter, while holding shift sometimes will not. Also note that some fine values will change even though the tool tip says the value is the same (for non-stepped controls) - this caused me great confusion in the past, where two knobs showed the same value but were clearly NOT the same exact value!

An issue with the SSL and automation is that the actual parameters values are not given for the automation (hence the conversion I had to do above), for example EQ frequency is not in Hz, gain is not in dB, etc.

But with automation, you can type values in directly (and get to those "in-between" values as well). I have sometimes automated a parameter strictly for the ability to type in a value directly!

Got it. Thanks!