Some questions on gain staging

Have an urge to learn, or a calling to teach? Want to share some useful Youtube videos? Do it here!
jimmyklane
Posts: 740
Joined: 16 Apr 2018

14 May 2018

selig wrote:
14 May 2018
jimmyklane wrote:
14 May 2018
True, but I fail to see the alternative when the gain change is isolated to a single frequency band. Overall volume and gain adjustments will change the level of the entire frequency spectrum, but boosting or cutting one frequency band can improve its presence relative to the static level.


THIS is where EQ ***before*** compression really comes in handy, because you can EQ quite heavily and then even out the volume spikes while still retaining the “flavor” that you needed to get by EQing in the first place. Sometimes sending that track to another bus and EQing again after the compressor can work really well.

EQing in parallel also works very well in situations where you want a track to have “air” but not get harsh....parallel channel, HP filter a bit and boost 16k by 6dB and mix it in under the main track.
EQ into compression is fine, but note that it likely will change the amount of compression which may or may not be desirable. Also, not all compressor settings will "even out the volume spikes" - this statement assumes a fast attack and high ratio, more akin to limiting than compression, correct?

How is parallel EQ different from just using less EQ in the first place? Yes, multiple bands of parallel EQ (internally) will interact differently, but this is not the same thing at all as putting a serial EQ on a parallel channel…

Also, boosting at 16 kHz can have extremely different results depending on the type of EQ used. Some EQs will hardly do anything set to 16 kHz, while others may affect the spectrum many octaves below 16 kHz. This comes up again and again when folks set two EQs to the same frequency and notice they don't sound the same, and assume it's because of some different "magic" used by one or the other. In most cases it's because the resultant curves look TOTALLY different when set to the same parameters.
Yeah, I meant “even out the frequency spikes”....poor choice of words on my part...in effect, compressing after EQ even with program compression settings tends to flatten the spectrum back out a little.

In addition, there are several technical reasons that a 16kHz boost would sound different. Almost all of them have to do with the Q and resultant phase response. Shelving filters can also use different topologies (differing math in digital) that cause ripples in phase and frequency. Sampling frequency can have an impact if the EQ handles the upper octave poorly, but in my personal experience that’s only present on very DSP (like my digital console....EQ is awful for more than filters and gentle dips)...
DAW: Reason 12

SAMPLERS: Akai MPC 2000, E-mu SP1200, E-Mu e5000Ultra, Ensoniq EPS 16+, Akai S950, Maschine

SYNTHS: Mostly classic Polysynths and more modern Monosynths. All are mostly food for my samplers!

www.soundcloud.com/jimmyklane

jimmyklane
Posts: 740
Joined: 16 Apr 2018

14 May 2018

selig wrote:
14 May 2018
Parallel EQ (SSL EQ) comparison:

+ 6 dB @ 16 kHz on a parallel channel exactly equals +3.5 dB @ 16 kHz on a single channel.

So if you want "less EQ", why not just use less EQ?

NOTE: when using parallel "anything" you're adding 6 dB if both channels are equal, so you need to compensate accordingly so as not to be fooled by the additional gain.

OK, so now, what if we lower the parallel channel by say, 6 dB, what is that going to give us? Turns out it's the same as setting the signal channel EQ to a boost of around + 2 dB (SSL won't give me exact enough values to match it precisely, but it's awful close!).

In other words, I'm not seeing any case where an EQ on a parallel channel does anything other than gives you less boost/cut while adding more overall gain. Now if you're ALSO adding compression, saturation, distortion or similar, you'll get results you can't necessarily get otherwise but it's down to the non-linear processing rather than the EQ (and the results may be very similar to what you'd get using a single channel in many cases, depending on settings of course).
Having the exact matching curve is not my experience...but the caveat is that I’m applying analog lessons to a digital world. Plugins like the “Console N” from Plugin Alliance also follow this general pattern, but if I mult a track, compress and EQ the hell out of it and blend it in under the track it’s really different to applying the processes individually. Even EQ on a mult seems to sound different to my ears, but of course that could be saturation in an EQ unit driven hard.
DAW: Reason 12

SAMPLERS: Akai MPC 2000, E-mu SP1200, E-Mu e5000Ultra, Ensoniq EPS 16+, Akai S950, Maschine

SYNTHS: Mostly classic Polysynths and more modern Monosynths. All are mostly food for my samplers!

www.soundcloud.com/jimmyklane

Nielsen
Posts: 100
Joined: 05 Nov 2017
Location: Denmark

14 May 2018

selig wrote:
14 May 2018
You’re not way off, but it shows that adding EQ doesn’t always give an overall improvement once you remove the added gain. No problem adding 2 dB gain IMO, I don’t get THAT anal about levels especially towards the end of a mix, so I’m defining not saying you need to compensate for EVERY single decibel!
Ok, that clears a great deal of my doubts.
selig wrote:
14 May 2018
BTW: You are not lowering the entire spectrum by 2 dB because you added 4 dB EQ which resulted in a 2 dB hotter level overall.
I meant after compensating back down to -12 dBFS with the gain plugin post EQ. That would be lowering the entire frequency spectrum of the channel by 2 dB, right? From -10 dBFS to -12 dBFS, in other words.
selig wrote:
14 May 2018
What if you instead cut the lower frequencies by 4 dB (instead of boosting the upper frequencies by 4 dB), resulting in the same response curve in the end? You would likely need to boost the level to get back to where you were before, but the results would likely be the same. But maybe you would “feel” better about it because you were adding gain instead of subtracting it?
I just tried doing this and the post EQ gain compensation actually seemed to work out better within the context of the mix. It's something I'll keep in mind. I can see situations where it won't work, but this could demonstrate why it's often a good idea to try cutting before boosting. (when possible). Thanks for the tip.
selig wrote:
14 May 2018
Boosting one band is simply changing the level of something less than the entire frequency spectrum. It’s still gain, just affecting fewer frequencies!
That was my point.
selig wrote:
14 May 2018
To be clear, I boost bands to improve clarity all the time because at a certain point in the mix process it’s more efficient to focus the energy at the most effective point. I’m sure what you’re doing is absolutely fine - only change your workflow if you’re not satisfied with results (not because someone such as myself does it differently). Many paths to the top of the mountain, yada yada, and Yoda too!
I'm first and foremost trying to have the concept of gain structure illuminated, even if it had more relevance back in the analogue days. Without doubt a bonus if recommended approaches work out better than my own. If not, great too because additional knowledge doesn't hurt the ability to make informed decisions in different situations. I really appreciate the help and advice shared in this thread!
selig wrote:
14 May 2018
OK, think about this “fader precision” bit above. Exactly how much fader precious do you need? In other words, what is the smallest amount of gain adjustment you can hear or you typically use when mixing?

Because when I ask that question, most folks say somewhere between 0.1 dB (few can hear this small of an adjustment) to 1.0 dB. Speaking for myself, I don’t often make finer adjustments than 0.5 dB or so, while on occasion I may feel I need 0.25 dB resolution.

Now look at the faders in Reason, and tell me how far down you can move them and STILL have the desired about of “resolution”. Let’s say you need 0.25 dB maximum fader resolution, which is really quite “fine” by most standards. How far can you move the fader before you cannot achieve this level of resolution? How about 0.1 dB resolution?

Turns out you can get 0.1 dB resolution down to -29 dB on Reason’s faders. Doesn’t sound like much? It’s over 3/4 of the way to the bottom. In fact, you get 0.25 dB resolution all the way down to below -55 dB, which is where the fader is over around 90% to the bottom. To get 0.5 dB resolution (which is what I typically use) you can go all the way to -70 dB, where the fader almost touches the bottom and covers the “screw” in the panel graphic. Something to think about when you’re worried about not having enough fader “resolution”…
;)
It's not uncommon that I make fader adjustments resulting in a difference around 0.30 dB. Do I need more precision? Not sure, but I've simply made a habit out of not adjusting the channels faders too much until the very end. I'll keep it in mind.
jimmyklane wrote:
14 May 2018
THIS is where EQ ***before*** compression really comes in handy, because you can EQ quite heavily and then even out the volume spikes while still retaining the “flavor” that you needed to get by EQing in the first place. Sometimes sending that track to another bus and EQing again after the compressor can work really well.

EQing in parallel also works very well in situations where you want a track to have “air” but not get harsh....parallel channel, HP filter a bit and boost 16k by 6dB and mix it in under the main track.
I'll keep that in mind too. Thanks.

User avatar
selig
RE Developer
Posts: 11681
Joined: 15 Jan 2015
Location: The NorthWoods, CT, USA

15 May 2018

jimmyklane wrote:
selig wrote:
14 May 2018
EQ into compression is fine, but note that it likely will change the amount of compression which may or may not be desirable. Also, not all compressor settings will "even out the volume spikes" - this statement assumes a fast attack and high ratio, more akin to limiting than compression, correct?

How is parallel EQ different from just using less EQ in the first place? Yes, multiple bands of parallel EQ (internally) will interact differently, but this is not the same thing at all as putting a serial EQ on a parallel channel…

Also, boosting at 16 kHz can have extremely different results depending on the type of EQ used. Some EQs will hardly do anything set to 16 kHz, while others may affect the spectrum many octaves below 16 kHz. This comes up again and again when folks set two EQs to the same frequency and notice they don't sound the same, and assume it's because of some different "magic" used by one or the other. In most cases it's because the resultant curves look TOTALLY different when set to the same parameters.
Yeah, I meant “even out the frequency spikes”....poor choice of words on my part...in effect, compressing after EQ even with program compression settings tends to flatten the spectrum back out a little.

In addition, there are several technical reasons that a 16kHz boost would sound different. Almost all of them have to do with the Q and resultant phase response. Shelving filters can also use different topologies (differing math in digital) that cause ripples in phase and frequency. Sampling frequency can have an impact if the EQ handles the upper octave poorly, but in my personal experience that’s only present on very DSP (like my digital console....EQ is awful for more than filters and gentle dips)...
Phase response with EQ = curve. If two curves match, then the phase response matches (unless we’re talking about linear phase EQ).

So yes, when comparing two EQs if there’s a different phase response, there’s a different EQ curve and thus the two curves will be (and sound) different.



Sent from some crappy device using Tapatalk
Selig Audio, LLC

User avatar
selig
RE Developer
Posts: 11681
Joined: 15 Jan 2015
Location: The NorthWoods, CT, USA

15 May 2018

jimmyklane wrote:
selig wrote:
14 May 2018
Parallel EQ (SSL EQ) comparison:

+ 6 dB @ 16 kHz on a parallel channel exactly equals +3.5 dB @ 16 kHz on a single channel.

So if you want "less EQ", why not just use less EQ?

NOTE: when using parallel "anything" you're adding 6 dB if both channels are equal, so you need to compensate accordingly so as not to be fooled by the additional gain.

OK, so now, what if we lower the parallel channel by say, 6 dB, what is that going to give us? Turns out it's the same as setting the signal channel EQ to a boost of around + 2 dB (SSL won't give me exact enough values to match it precisely, but it's awful close!).

In other words, I'm not seeing any case where an EQ on a parallel channel does anything other than gives you less boost/cut while adding more overall gain. Now if you're ALSO adding compression, saturation, distortion or similar, you'll get results you can't necessarily get otherwise but it's down to the non-linear processing rather than the EQ (and the results may be very similar to what you'd get using a single channel in many cases, depending on settings of course).
Having the exact matching curve is not my experience...but the caveat is that I’m applying analog lessons to a digital world. Plugins like the “Console N” from Plugin Alliance also follow this general pattern, but if I mult a track, compress and EQ the hell out of it and blend it in under the track it’s really different to applying the processes individually. Even EQ on a mult seems to sound different to my ears, but of course that could be saturation in an EQ unit driven hard.
This echos my experiences (and comments above) about parallel processing. It’s typically when you’re doing multiple stages of processing on the parallel channel that you get the most valuable results. :)

I would also add that for us folks coming from the analog world, we do need to be clear when making comments on a forum about an all digital medium! Most of the users here have little to no experience in the analog world, let alone experience working on a big format console in a well tuned room.

In the digital world you DO get exact matching curves with EQ on a parallel channel (not to be confused with parallel EQ). With any non-linear system (or analog) this won’t necessarily be true, but unless you’re saturating the hell out of something the curve will be relatively similar to the ear (besides the saturation bit!), which I think is what you’re talking about if I’m following you correctly.


Sent from some crappy device using Tapatalk
Selig Audio, LLC

jimmyklane
Posts: 740
Joined: 16 Apr 2018

15 May 2018

selig wrote:
15 May 2018
jimmyklane wrote:
Having the exact matching curve is not my experience...but the caveat is that I’m applying analog lessons to a digital world. Plugins like the “Console N” from Plugin Alliance also follow this general pattern, but if I mult a track, compress and EQ the hell out of it and blend it in under the track it’s really different to applying the processes individually. Even EQ on a mult seems to sound different to my ears, but of course that could be saturation in an EQ unit driven hard.
This echos my experiences (and comments above) about parallel processing. It’s typically when you’re doing multiple stages of processing on the parallel channel that you get the most valuable results. :)

I would also add that for us folks coming from the analog world, we do need to be clear when making comments on a forum about an all digital medium! Most of the users here have little to no experience in the analog world, let alone experience working on a big format console in a well tuned room.

In the digital world you DO get exact matching curves with EQ on a parallel channel (not to be confused with parallel EQ). With any non-linear system (or analog) this won’t necessarily be true, but unless you’re saturating the hell out of something the curve will be relatively similar to the ear (besides the saturation bit!), which I think is what you’re talking about if I’m following you correctly.


Sent from some crappy device using Tapatalk
You know, I see you’re totally correct....using a mult on my trusty O2Rv2 gives me the exact same spectrum once adjusted for amplitude. It’s when this is done in analog (for example the Kush clairphonic rack unit) that it seems to have a difference.

I’m still an advocate of setting a firm reference level, like +4dBu=-18dBFS for example, and monitoring at something very close to 85dB SPL....anyone that ever uses any outboard rack equipment...even if it’s just a preamp/EQ/compressor combo for input will find that they’re always driving the correct levels into their equipment as well as the audio interface. That being said, the most fun in the analog world comes with things like saturating your preamp’s transformers and driving your other equipment “into the red” to get a certain tone out of it....I do that all the time with a synth setup I’ve come up with (Demeter Tube DI into Neve Portico 5012 into Warm Audio WA2A with the compressor providing the output attenuation) in order to get a certain grit and character to cleaner synths like the Bass Station II....

In the box....strictly within Reason....I think there is something like 1500dB of internal dynamic range before the mix bus, where it goes up to 64 bits and a truly (in effect) limitless, right up to the point that it hits your D/A....which you know but others may not. It’s interesting that in the digital algorithms that parallel and serial EQ is essentially the same. When I do that inside Reason, I tend to create a type of “exciter” effect where I’ll HP quite severely and boost extreme high end, compress and sneak the track in till it just slightly brightens the original track....seems like your results show that I could simply apply a small amount of EQ and compression on the original channel and get the same exact result. Interesting!
DAW: Reason 12

SAMPLERS: Akai MPC 2000, E-mu SP1200, E-Mu e5000Ultra, Ensoniq EPS 16+, Akai S950, Maschine

SYNTHS: Mostly classic Polysynths and more modern Monosynths. All are mostly food for my samplers!

www.soundcloud.com/jimmyklane

User avatar
selig
RE Developer
Posts: 11681
Joined: 15 Jan 2015
Location: The NorthWoods, CT, USA

15 May 2018

jimmyklane wrote: You know, I see you’re totally correct....using a mult on my trusty O2Rv2 gives me the exact same spectrum once adjusted for amplitude. It’s when this is done in analog (for example the Kush clairphonic rack unit) that it seems to have a difference.
OK, let me stop you right there and repeat something I said earlier.

Using a serial EQ on a parallel channel is NOT the same thing as using a parallel EQ (like Kush).

Most EQ is serial from band to band, but some are parallel like the Kush (and Combinators I built for my ColoringEQ). What this means is that if you have two bands, the input feeds both bands rather than going into band 1 then band 2, and the output is compensated for the doubling of voltage (dropped by 6 dB). What happens is a different interaction between the two bands that you don’t get with serial EQ.

But more importantly, a parallel EQ is different from a serial EQ on a parallel channel because it is STILL a serial EQ. So you don’t get the same interaction between bands as you do with a parallel EQ device.
jimmyklane wrote: I’m still an advocate of setting a firm reference level, like +4dBu=-18dBFS for example, and monitoring at something very close to 85dB SPL....anyone that ever uses any outboard rack equipment...even if it’s just a preamp/EQ/compressor combo for input will find that they’re always driving the correct levels into their equipment as well as the audio interface.
No argument there, it’s just that most folks don’t have interfaces that allow you to adjust the reference level such as you do on high-end devices. Since the typical interface cannot be adjusted, one can say they use a “firm” reference level! ;)

The reason for doing this is so that your meters in the analog world relate to a known standard in the digital world. This means if you use alignment tones to set your 0 VU levels on your analog gear, you know what the digital equivalent will be. OTOH, if your analog gear has no VU meters, this is a moot point, as even if you DO run alignment tones you won’t be able to see the levels.

For example, on my UA gear, the VU meters are a part of the compressor, and since I cannot adjust my converter’s reference level the only thing I need to be concerned about is understanding the gain staging of the pre amp vs compressor in each device, and hit a target of peaks around -12 dBFS in the digital domain.

jimmyklane wrote: That being said, the most fun in the analog world comes with things like saturating your preamp’s transformers and driving your other equipment “into the red” to get a certain tone out of it....I do that all the time with a synth setup I’ve come up with (Demeter Tube DI into Neve Portico 5012 into Warm Audio WA2A with the compressor providing the output attenuation) in order to get a certain grit and character to cleaner synths like the Bass Station II....

In the box....strictly within Reason....I think there is something like 1500dB of internal dynamic range before the mix bus, where it goes up to 64 bits and a truly (in effect) limitless, right up to the point that it hits your D/A....which you know but others may not. It’s interesting that in the digital algorithms that parallel and serial EQ is essentially the same.
Again, it’s same for analog if you mean putting a serial EQ on a parallel channel. The basic rules for EQ and levels don’t change between analog and digital, at least not in this regard and not in my experience - will be happy to stand corrected if this is not your experience!
jimmyklane wrote: When I do that inside Reason, I tend to create a type of “exciter” effect where I’ll HP quite severely and boost extreme high end, compress and sneak the track in till it just slightly brightens the original track....seems like your results show that I could simply apply a small amount of EQ and compression on the original channel and get the same exact result. Interesting!
Again referring you to my previous comments, when you add a non-linear process to a parallel channel you CAN get different results compared to the same process on a serial insert (so I don’t think you’ll get the same results unless you’re talking about EQ alone).

This is especially true with any process that adds harmonics because you can distort the crap out of a signal and mix in a slight amount - with a serial insert you cannot simultaneously distort the crap out of something AND use only a small amount. That’s been my drum bus processing (limiting and tape compression on a parallel channel) for years now. ;)

This is also the basic theory behind my ColoringEQ: saturate a specific frequency or frequency range, and mix in small amounts with the original or EQ’d signal.

Fun topic, btw, enjoying shooting the crap with you!


Sent from some crappy device using Tapatalk
Selig Audio, LLC

jimmyklane
Posts: 740
Joined: 16 Apr 2018

15 May 2018

selig wrote:
15 May 2018
jimmyklane wrote: You know, I see you’re totally correct....using a mult on my trusty O2Rv2 gives me the exact same spectrum once adjusted for amplitude. It’s when this is done in analog (for example the Kush clairphonic rack unit) that it seems to have a difference.
OK, let me stop you right there and repeat something I said earlier.

Using a serial EQ on a parallel channel is NOT the same thing as using a parallel EQ (like Kush).

Most EQ is serial from band to band, but some are parallel like the Kush (and Combinators I built for my ColoringEQ). What this means is that if you have two bands, the input feeds both bands rather than going into band 1 then band 2, and the output is compensated for the doubling of voltage (dropped by 6 dB). What happens is a different interaction between the two bands that you don’t get with serial EQ.

But more importantly, a parallel EQ is different from a serial EQ on a parallel channel because it is STILL a serial EQ. So you don’t get the same interaction between bands as you do with a parallel EQ device.
jimmyklane wrote: I’m still an advocate of setting a firm reference level, like +4dBu=-18dBFS for example, and monitoring at something very close to 85dB SPL....anyone that ever uses any outboard rack equipment...even if it’s just a preamp/EQ/compressor combo for input will find that they’re always driving the correct levels into their equipment as well as the audio interface.
No argument there, it’s just that most folks don’t have interfaces that allow you to adjust the reference level such as you do on high-end devices. Since the typical interface cannot be adjusted, one can say they use a “firm” reference level! ;)

The reason for doing this is so that your meters in the analog world relate to a known standard in the digital world. This means if you use alignment tones to set your 0 VU levels on your analog gear, you know what the digital equivalent will be. OTOH, if your analog gear has no VU meters, this is a moot point, as even if you DO run alignment tones you won’t be able to see the levels.

For example, on my UA gear, the VU meters are a part of the compressor, and since I cannot adjust my converter’s reference level the only thing I need to be concerned about is understanding the gain staging of the pre amp vs compressor in each device, and hit a target of peaks around -12 dBFS in the digital domain.

jimmyklane wrote: That being said, the most fun in the analog world comes with things like saturating your preamp’s transformers and driving your other equipment “into the red” to get a certain tone out of it....I do that all the time with a synth setup I’ve come up with (Demeter Tube DI into Neve Portico 5012 into Warm Audio WA2A with the compressor providing the output attenuation) in order to get a certain grit and character to cleaner synths like the Bass Station II....

In the box....strictly within Reason....I think there is something like 1500dB of internal dynamic range before the mix bus, where it goes up to 64 bits and a truly (in effect) limitless, right up to the point that it hits your D/A....which you know but others may not. It’s interesting that in the digital algorithms that parallel and serial EQ is essentially the same.
Again, it’s same for analog if you mean putting a serial EQ on a parallel channel. The basic rules for EQ and levels don’t change between analog and digital, at least not in this regard and not in my experience - will be happy to stand corrected if this is not your experience!
jimmyklane wrote: When I do that inside Reason, I tend to create a type of “exciter” effect where I’ll HP quite severely and boost extreme high end, compress and sneak the track in till it just slightly brightens the original track....seems like your results show that I could simply apply a small amount of EQ and compression on the original channel and get the same exact result. Interesting!
Again referring you to my previous comments, when you add a non-linear process to a parallel channel you CAN get different results compared to the same process on a serial insert (so I don’t think you’ll get the same results unless you’re talking about EQ alone).

This is especially true with any process that adds harmonics because you can distort the crap out of a signal and mix in a slight amount - with a serial insert you cannot simultaneously distort the crap out of something AND use only a small amount. That’s been my drum bus processing (limiting and tape compression on a parallel channel) for years now. ;)

This is also the basic theory behind my ColoringEQ: saturate a specific frequency or frequency range, and mix in small amounts with the original or EQ’d signal.

Fun topic, btw, enjoying shooting the crap with you!


Sent from some crappy device using Tapatalk
Right, the difference in a serial vs parallel process is that one band either feeds into the next and is influencing its response, or that it is a function that is isolated. To the best of my knowledge, all EQ is technically parallel in its structure. You’ve built an EQ, so you’re probably aware of bi-quad DSP algorithms. In the analog domain, we’re adding a phase shifted version of the signal to itself to obtain an amplitude shift at a particular corner frequency.

Of course, I realize that you mean that the “sections” of an EQ are in series and therefore you could theoretically boost and cut at the same frequency and Q in either a serial or parallel channel and null.

I’m not sure I’ve ever actually used a mult to simply EQ a signal....except for mute automation where I’m bringing in (for example) chorus vocals over a verse vocal.... I’m pretty much always adding a non-linear process in addition to the EQ.
DAW: Reason 12

SAMPLERS: Akai MPC 2000, E-mu SP1200, E-Mu e5000Ultra, Ensoniq EPS 16+, Akai S950, Maschine

SYNTHS: Mostly classic Polysynths and more modern Monosynths. All are mostly food for my samplers!

www.soundcloud.com/jimmyklane

Post Reply
  • Information
  • Who is online

    Users browsing this forum: No registered users and 3 guests