FabFilter

Not for nothing - but automation in DAW's is tracking them by BAND NUMBER. so wouldn't re-ordering create chaos with existing automation? Or am I missing something.

Yes you're right. Hence why I said about having a button, so that the user has the choice. You probably would choose not to reorder if you already have automation set, but in most cases I think that's a non issue, as most people's EQ work is done well before considering automation. TDR Nova has had that feature for a long time, it works great.

Actually scrap that idea...

Upon reflection, I think the intentionality of when a processor effects TONE is MORE important.

Here's something I realised with fresh ears... is that it's 'generally' best to apply a processor that makes the signal darker FIRST, before adding a processor that makes the signal brighter (regardless of how masking works on a technical front). You're essentially choosing to place processing in order of what is more heard, first (I know, this seems obvious lol, stay with me...).

I'll explain, using the same example as above...

Let's say we have 2 separate instances of a DeEsser plugin (Both with bell filters, and always dialled in accordingly).
One is set to 4k, the other set to 250Hz.

In most cases you'll hear the 4k as the most dominating frequency BEFORE you hear any problems with 250Hz (due to Equal Loudness Curve etc.). So you'd place 4k instance before the 250 Hz instance.

In the name of CLARITY (what I was prioritising above) this would seem incorrect, but in the name of TONE it makes perfect sense. The 4k DeEsser makes the signal darker first, then the 250 Hz cleans up the signal, without taking away too much or making the overall signal sound too thin or unnatural.

On the other hand if you place the 250 Hz instance before the 4k, the 250 DeEsser makes the signal brighter before you even address the 4k. If you A/B with the dry signal, this configuration tends to be further from the original signal, and is thinner and less natural.

In this case, TONE is prioritised over CLARITY. It better preserves the very nature of the source.

From what I've realised, tone should always be considered FIRST, before clarity. I've had much more preferable results using the "darker, then brighter" approach.

-----

So with all that said, what perhaps would be nice to see in Pro-Q is the ability to choose which band comes before others, when you need it most. I.e. when you've placed a multiple bands in the same instance, but would prefer if one came before the other - as currently it seems that band CREATION is what dictates the order of how bands are processed in series. Sure, you could create another instance, but that's unnecessary latency (especially as Pro-Q uses the same latency per instance, regardless of band count). It would be great if it was controllable within the single instance.

Maybe a photoshop "send to back" kind of thing when you right click on a node. It could be called 'Position', and have the following options:
• Forward
• Backward
• To Front
• To Back

***EDIT***
see below for correction, under "*CORRECTED*".
-----

Actually scrap that idea...

Upon reflection, I think the intentionality of when a processor effects TONE is MORE important.

Here's something I realised with fresh ears... is that it's 'generally' best to apply a processor that makes the signal darker FIRST, before adding a processor that makes the signal brighter (regardless of how masking works on a technical front). You're essentially choosing to place processing in order of what is more heard, first (I know, this seems obvious lol, stay with me...).

I'll explain, using the same example as above...

Let's say we have 2 separate instances of a DeEsser plugin (Both with bell filters, and always dialled in accordingly).
One is set to 4k, the other set to 250Hz.

In most cases you'll hear the 4k as the most dominating frequency BEFORE you hear any problems with 250Hz (due to Equal Loudness Curve etc.). So you'd place 4k instance before the 250 Hz instance.

In the name of CLARITY (what I was prioritising above) this would seem incorrect, but in the name of TONE it makes perfect sense. The 4k DeEsser makes the signal darker first, then the 250 Hz cleans up the signal, without taking away too much or making the overall signal sound too thin or unnatural.

On the other hand if you place the 250 Hz instance before the 4k, the 250 DeEsser makes the signal brighter before you even address the 4k. If you A/B with the dry signal, this configuration tends to be further from the original signal, and is thinner and less natural.

*CORRECTED*
In this case, CLARITY is prioritised instead of TONE, which makes the signal thinner, and it doesn't preserve the nature of the source nearly as well.

From what I've realised, tone should always be considered FIRST, before clarity. I've had much more preferable results using the "darker, then brighter" approach.

-----

So with all that said, what perhaps would be nice to see in Pro-Q is the ability to choose which band comes before others, when you need it most. I.e. when you've placed a multiple bands in the same instance, but would prefer if one came before the other - as currently it seems that band CREATION is what dictates the order of how bands are processed in series. Sure, you could create another instance, but that's unnecessary latency (especially as Pro-Q uses the same latency per instance, regardless of band count). It would be great if it was controllable within the single instance.

Maybe a photoshop "send to back" kind of thing when you right click on a node. It could be called 'Position', and have the following options:
• Forward
• Backward
• To Front
• To Back

Hi Ryan,

The filters in Pro-Q are completely linear (unless you're using Character mode in Pro-Q 4). It does not matter in what order you create or process the bands, the result will be the same. Therefore the order of the bands does not have a sonic difference.

I just checked the TDR manual, and they only mention this function as a convenience:

"Use the convenient “ORDER BY FREQ” function (p.14) to reorder the bands in the familiar ascending sequence."

In terms of the de-esser, this would only make a difference if the de-esser is reacting to the complete signal and not to a filtered side-chain. Most de-essers work with a filtered part of the audio signal triggering the side-chain, so your 4K band will not be triggered by any frequency at 250Hz, regardless of its level.

I was about to slam Ryan yesterday and did a null test with two DYNAMIC bands that overlapped a little - and then swapped band order, and the end result didn’t null out for more than 30dB, which implies a difference.

I’m not sure how dynamic bands work, but it makes sense- first band alters waveform which will alter how second band detects it, if topology is series.

Havent tried the same with proMB dynamic phase yet.
Crossovered modes should be parallel anyway.

It’s not something im gonna lose sleep over tho

Yes, dynamic bands are a different story, as they are a non-linear process. All bands, including dynamic bands, are serial. There is a small difference with spectral bands though. As spectral bands are a linear process by default, when using Pro-Q 4 in zero latency or natural phase mode all spectral bands are grouped together and processed after the static and dynamic bands. This is to prevent multiple conversations to and from linear phase if multiple spectral bands are used.

Hi guys, I should've emphasised that all my tests were with DYNAMIC filters.

I did the same test as @Ploki within a single instance, as well as using 2 separate instances of Pro-Q 3, with a band in each - of course expecting the same result as both would occur in series. I said 'DeEsser' in the original post as an example of practical application, but I did in fact use Pro-Q.

@Ploki, although I haven't tested it, I would not expect Pro-MB in 'Dynamic Phase' to do the same, as that mode literally changes processing to a Parallel Topology.

I agree that if we used Pro MB in Dynamic Phase for the 250/4k example, that we would hear no difference as processing is summed together before the output, so the 250 Hz & 4k bands would effectively be processed at the same time.

I was not referring to overlapping bands, but rather how you said "first band alters waveform which will alter how second band detects it". In fact, that's exactly what I'm asking about, and I'm still very interested to hear you're opinions this!

Just so you know, I do understand the practical benefits/drawbacks of using serial vs parallel topologies for certain applications:

**SERIAL TOPOLOGY** (a.k.a. Series, or Sequential) - Splits the signal into bands (filters) and sums it back together in series.

DYNAMICS in Serial:
When the compression threshold goes back to 0 dB, the filter is still present, so the phase shift remains for that band.

LINEAR RESPONSE in Serial:
• 2 boosts in close proximity ‘combine’ to make one larger boost (e.g. 2x 6dB bells at the same frequency, make 12dB).
• 2 adjacent cuts combine to form one larger cut.

APPLICATIONS of Serial:
✅ Subtractive work (both Static & Dynamic) - Filters predictably sum together if close together or overlapping.

(vs)

**PARALLEL TOPOLOGY** - Does NOT break the signal apart and reconstruct it. Filters only present when compression is engaged.

DYNAMICS in Parallel:
When the compression threshold goes back to 0 dB, the filter effectively 'disappears’ so there is no phase shift being caused by that band.

Also, Artefacts from dynamic processing add up (via blending) in parallel, instead of multiplying (via summing) in series, which can potentially give you a much cleaner overall result. Source

LINEAR RESPONSE in Parallel:
• 2 boosts in close proximity DO NOT ‘combine’ to make one larger boost (e.g. 2x 6dB bells at the same frequency, make 7.9 dB).
• 2 adjacent cuts remain 2 separate cuts, and cut less overall.

APPLICATIONS of Parallel:
✅ Broad dynamic moves - Compression (incl. DeEssing) or expansion will sound more natural/musical, with less audible compression artifacts.
🔴 Tight adjacent subtractive EQ - Somewhat unpredictable IF filters are close together or overlapping. May require additional Q adjustments.

---

With all that said, it would seem that the cleanest results would be with Pro-MB in 'Dynamic Phase' mode with it's Parallel Topology, instead of using Pro-Q for this as bands are not overlapping in any way, however I would still really like to hear your opinions on the whole 250 Hz masking thing, as seen in the short videos above.

Same setup as before, 2 separate instances of Pro-MB, one with a bell at 250 Hz, and the other with a bell at 4k - only now with less compression artefacts, and filters only present when comp occurs... but which instance IN THEORY should come first?

P.S. I know this might seem like a 1% improvement game, but heck why not!

Ralph (FabFilter), you said:
"In terms of the de-esser, this would only make a difference if the de-esser is reacting to the complete signal and not to a filtered side-chain. Most de-essers work with a filtered part of the audio signal triggering the side-chain, so your 4K band will not be triggered by any frequency at 250Hz, regardless of its level."

I understand what you mean, but it's not about triggering. It's about how the second DeEsser reacts to a signal that now has 250 Hz under control (which is said to mask frequencies above it).

There is an audible difference when using 2 instances of RDeEsser in series. Both using split band, bell sidechain to trigger, one at 250 Hz, the other at 4k. If you swap the instances, there you can hear it.

@Ralph (FabFilter)
Here's a summary to conclude this, after more extensive real-world testing.

DEESSING:
When it comes to DeEssing, it's a dynamics first thing; it makes sense to Compress/DeEss before Static EQ as to ensure that what you're trying to turn down with EQ, isn't brought up again with compression.
Doing things in this order (Compression > Static EQ) makes the EQ "more effective".

BELLS, SHELVES, TILTS:
Of course, bells, shelves, and tilts CAN be dynamic...
Again, I would argue that dynamics should always come before static EQ but I'm aware that would be a nightmare to code, what with automation n' all.

FILTERS:
As filters aren't dynamic, they should always be processed after dynamics, and it would be nice to see them automatically allocated to the last filters in the sequence, e.g. 23 and 24 — so that if a dynamic bell or shelf is in the same instance, it always get's processed before the static filters. This means you'd always have "more effective" filters. Hopefully this would be a littler easier to implement too, as they're maybe more predictable in terms of the user uses them.

Following dynamics, if I were to pick an order for static filters, it'd be: Notch > BandPass > HP > LP → Output.

For now, I've saved my default preset with a HP & LP filter, using bands 23 and 24.

Thanks for reading!