How should I understand the graph on the parametric EQ? (not a beginners question)

Discuss music production with Ableton Live.
Post Reply
romariozen
Posts: 63
Joined: Thu Jul 25, 2019 3:47 pm

How should I understand the graph on the parametric EQ? (not a beginners question)

Post by romariozen » Wed Dec 08, 2021 2:45 pm

My question is, let's say, we have a 20 Hz wave and 10 kHz wave. Both have a specific x coordinate that has to be a straight line. Why do I see a broad hill when I monitor the low frequency contrary to the high frequency hill that almost looks like a straight line? When I have two frequencies, 20 and 21 Hz simultanelusly, It would be hell of a task to identify that there are 2 tones actually.

The purpose of mine is to understand how I should read the rumble below 20 Hz,. Even when I implement steep low cut below 20 Hz, I can see that that low frequency rumble < 20 Hz doesn't disapper, it just has its slope to the lowest mix tone a little lower.
Is it just graph, or there is some physics for that?

jestermgee
Posts: 4500
Joined: Mon Apr 26, 2010 6:38 am

Re: How should I understand the graph on the parametric EQ? (not a beginners question)

Post by jestermgee » Wed Dec 08, 2021 9:30 pm

You have to keep in mind "resolution" when dealing with frequencies.

The lower the frequency the less information you have in it but the more power you have, it's a physics translation. That's why you can hear a thumping sub from a mile away but often hear no actual musical content.

If you had a 20 and 21Hz signal you would not be able to see these easily on a spectrograph as separate frequencies, only as a slight rolling ripple but almost nothing visually because there is simply not enough information. Higher frequencies have much more info to detect so can utilise narrower bands.

When you refer to the "hill" that would be the "Q" or "Bandwidth" of the frequency you are focusing on, the range either side of the frequency point. Because low frequencies have less info there is a much wider bandwidth given to fit something within. Higher frequencies have a narrower bandwidth because there is more info in there and so you can focus in on a finer frequency range. This is why you see frequency scales as "logarithmic" rather than linear because you need to have some visual to assist in the lower end otherwise you just wouldn't see enough width for low end analysis on a graph or the graph would be huge to cover everything evenly.

Image

This is why there is a drive to make higher and higher frequency phone towers, more info with narrower bandwidth channels = more data and more users. If you are following along and understand your physics tho, it becomes apparent that these higher frequencies would need more power to transmit the same distance tho so what happens is you need either much stronger towers and phones (risk of having phones transmit high power and fry your brain) or you need more towers in more locations with lower power. If you are from the days where older analog CDMA mobile phones were a thing you would recall seeing huge towers that covered areas up to 50KM (that's 31 miles) or something but we were sending basic voice signals at a low frequency, now a 5G signal I think has something like 1000ft of range so there is a real relationship between frequency/information and range.

When you create a low shelf and roll off the low end you will always still see some signal there because you can't just hard cut it completely but the signal ratio will be reduced a lot which will take it's power out of your signal and essentially make it mute. The goal is to not completely remove things, just reduce its influence enough so the info you have is able to clearly be heard.

romariozen
Posts: 63
Joined: Thu Jul 25, 2019 3:47 pm

Re: How should I understand the graph on the parametric EQ? (not a beginners question)

Post by romariozen » Thu Dec 09, 2021 1:50 pm

jestermgee wrote:
Wed Dec 08, 2021 9:30 pm
You have to keep in mind "resolution" when dealing with frequencies.

The lower the frequency the less information you have in it but the more power you have, it's a physics translation. That's why you can hear a thumping sub from a mile away but often hear no actual musical content.

If you had a 20 and 21Hz signal you would not be able to see these easily on a spectrograph as separate frequencies, only as a slight rolling ripple but almost nothing visually because there is simply not enough information. Higher frequencies have much more info to detect so can utilise narrower bands.

When you refer to the "hill" that would be the "Q" or "Bandwidth" of the frequency you are focusing on, the range either side of the frequency point. Because low frequencies have less info there is a much wider bandwidth given to fit something within. Higher frequencies have a narrower bandwidth because there is more info in there and so you can focus in on a finer frequency range. This is why you see frequency scales as "logarithmic" rather than linear because you need to have some visual to assist in the lower end otherwise you just wouldn't see enough width for low end analysis on a graph or the graph would be huge to cover everything evenly.

Image

This is why there is a drive to make higher and higher frequency phone towers, more info with narrower bandwidth channels = more data and more users. If you are following along and understand your physics tho, it becomes apparent that these higher frequencies would need more power to transmit the same distance tho so what happens is you need either much stronger towers and phones (risk of having phones transmit high power and fry your brain) or you need more towers in more locations with lower power. If you are from the days where older analog CDMA mobile phones were a thing you would recall seeing huge towers that covered areas up to 50KM (that's 31 miles) or something but we were sending basic voice signals at a low frequency, now a 5G signal I think has something like 1000ft of range so there is a real relationship between frequency/information and range.

When you create a low shelf and roll off the low end you will always still see some signal there because you can't just hard cut it completely but the signal ratio will be reduced a lot which will take it's power out of your signal and essentially make it mute. The goal is to not completely remove things, just reduce its influence enough so the info you have is able to clearly be heard.
So, if I get it right, an equaliser has to do some work to understand how fast the wave oscillates that's why having 20 Hz oscilation is approximate, because it's harder fto identify ideal equal cycles within some time period to make the equaliser surely admit it's ideal 20 Hz. And we can also say that an equalizer works less precise in the lows (not in terms of the scale but in terms of the influence on the sound) right? By the way, thanks so much for your response.

jestermgee
Posts: 4500
Joined: Mon Apr 26, 2010 6:38 am

Re: How should I understand the graph on the parametric EQ? (not a beginners question)

Post by jestermgee » Thu Dec 09, 2021 9:53 pm

So, if I get it right, an equaliser has to do some work to understand how fast the wave oscillates that's why having 20 Hz oscilation is approximate, because it's harder to identify ideal equal cycles within some time period to make the equaliser surely admit it's ideal 20 Hz.
Not quite. Does not matter the freq, an EQ doesn't have to "work" to find out/know what it is dealing with. It's the wrong way to think of it. An EQ will very accurately apply to a single band, it's how much affect it has to a range based on the bandwidth.

Probably a better way to think is how speakers produce sound. Thinking in terms of a 3 way speaker you have low, mid and high. A 3 band EQ affects the approximate sound of each of these drivers so if we look at the range of each:

Low: 20-200Hz (Bandwidth = 180Hz)
Mid: 200-2200Hz (Bandwidth = 2000Hz)
Hi: 2200-12000Hz+ (Bandwidth = 9800Hz)

Quite imply to refer to your original question, we have only 180Hz of "resolution" at the low end to work with compared to approx. 10,000Hz at the high end. This is why the log scale shows larger adjustments at low end (larger display of the bandwidth) because there is just less info to work with. Adjusting a boost of 40Hz and then shifting that up to 80Hz will affect the sound greatly and it's only a difference of 40Hz. However, adjusting 2200Hz up to 2240 will almost not affect the sound at all so basically we don't need to see such fine detail in the high end to make large adjustments. It is, however, much easier to make finer and more precise operations on high frequencies to tame them than low.

May be an idea to just look into how things like the EQ8 work and why. Here's a quick article that may be handy:
https://www.productionmusiclive.com/blo ... -to-use-it

I personally only know the basics really, understand the physics of what is happening but there is so much to dive into you could get lost for days learning.

romariozen
Posts: 63
Joined: Thu Jul 25, 2019 3:47 pm

Re: How should I understand the graph on the parametric EQ? (not a beginners question)

Post by romariozen » Sun Dec 12, 2021 12:36 pm

jestermgee wrote:
Thu Dec 09, 2021 9:53 pm
So, if I get it right, an equaliser has to do some work to understand how fast the wave oscillates that's why having 20 Hz oscilation is approximate, because it's harder to identify ideal equal cycles within some time period to make the equaliser surely admit it's ideal 20 Hz.
Adjusting a boost of 40Hz and then shifting that up to 80Hz will affect the sound greatly and it's only a difference of 40Hz
I get this logarythmic concept, I just get confused on why these hills representing the waves on the EQ have any form (that looks like some hill that you can never climb up) and not just a strict vertical line, I just can't get that. If we generate a sine wave 10 Hz, its hill's incline starts at 65 Hz(!) and even more it's peak doesn't even look like a pure peak, iit definately has 2 hills like a potato or something. This thing with EQ I just can't get why it's that. Going further, I checked taht wave with 2 different parametric EQs, Ableton's one shows start of the incline at 65 Hz, Pro Q 3' monitor shows start of the incline at 100 Hz, how should I interpret that.


By the way, That line I quoted here is crucial to me, sometimes you dont' really pay attention by moving the band to the right and to the left. Thanks for that knowledge.

Tarekith
Posts: 19072
Joined: Fri Jan 07, 2005 11:46 pm
Location: Ableton Forum Administrator
Contact:

Re: How should I understand the graph on the parametric EQ? (not a beginners question)

Post by Tarekith » Sun Dec 12, 2021 3:42 pm

Tarekith
Ableton Forum Administrator
https://tarekith.com

romariozen
Posts: 63
Joined: Thu Jul 25, 2019 3:47 pm

Re: How should I understand the graph on the parametric EQ? (not a beginners question)

Post by romariozen » Sun Dec 12, 2021 4:48 pm

Tarekith wrote:
Sun Dec 12, 2021 3:42 pm
https://ethanwiner.com/EQPhase.html
Hi, Tarekith, thanks for the topic, I've read this post now, but it's not about my quewstion. Maybe you know why when we play a pure sine 20 Hz we see the broad hill starting at 50 Hz on the EQ monitor instead of just showing the strict line?

Look how wide the hill of a pure sine is. I can even assume that it is not just "to look nice", we can clearly see how more wide the hill is to the left, which means it refers to the logarythmic frequency response and not just a nuance of the design

Image

https://1drv.ms/u/s!AnjxgdsMyUgAojIceGB ... 3?e=gyuwVP

I can increase the fidelity and the hill gets thinner, but it is still a hill with the left side widened. I start to realise that EQ after all makes some work to identify the frequency at some small time period after analizing its cycles.

yur2die4
Posts: 7154
Joined: Sat Oct 03, 2009 3:02 am
Location: Menasha, Wisconsin
Contact:

Re: How should I understand the graph on the parametric EQ? (not a beginners question)

Post by yur2die4 » Sun Dec 12, 2021 7:13 pm

A little thing I’d like to point out about your specific mention of 20 and 21 hz.

When you have simultaneous tones, they sum together. The tones themselves exist via cycles of amplitude change… so 21 hz isn’t just a ‘tone’, I mean it is, but it’s a tone as the result of a repeated amplitude change, ‘smoothly (sine)’ at 21 hz.

If you simultaneously play a 20 hz tone, the activity of Both consistent amplitude changes are summed together to the ‘output’ which could be a speaker. A speaker is not doing two independent ‘lines’ of audio, it is just replicating the merging of these two tones. The two tones are now a single
More complex cycle in some ways… even if you can also ‘Conceptually’ imagine them as perfect individual tones.

The result being summed, this ends up giving you a ‘beating’ tone (I suggest looking up beating) which is the frequency of the two averaged, approximately, so let’s say 20.5 hz, it is that tone, except it gets louder and quieter at a rate of the Difference of those two tones, so let’s say it goes wub wub wub at a rate of 1 hz.

When you have that kind of activity with tones so close to one another, it gets really messy. And that is on top of noise, headroom, and the tools we use to ‘analyze’ these things. There are other types of analysis other than peaks (spectral), it’s just that this kind of graphic superimposes most closely to the EQ layout.

As mentioned previously, when you have the higher frequencies, they are drastically further apart and their differences are more clearly defined, a 1000 hz and a 2000 hz sine simultaneously will have a much more different interaction when summed. Summed to 1500 hz with a 1000 hz beating? (My numbers and calculations might be off, been a while since I looked at the actual specific equations)(and to make matters worse, I used octaves, oops, maybe 1000 and 1100 )) . BUT of you try 1001 hz with 1000 hz it’s a different story.

So… this may not explain a Lot, but it is something to consider with regard to the example of 20 vs 21. But also how crazy generated tones can be… you’re kind of feeling around in the dark to a degree. The graphic display doesn’t know in advance what you’re trying to throw at it. So it gives an interpretation to the best of its ability, in the ideally most useful way, without going overboard on cpu or making too many assumptions about how it’s supposed to display the info.

romariozen
Posts: 63
Joined: Thu Jul 25, 2019 3:47 pm

Re: How should I understand the graph on the parametric EQ? (not a beginners question)

Post by romariozen » Sun Dec 12, 2021 9:42 pm

yur2die4 wrote:
Sun Dec 12, 2021 7:13 pm

. The graphic display doesn’t know in advance what you’re trying to throw at it. So it gives an interpretation to the best of its ability, in the ideally most useful way, without going overboard on cpu or making too many assumptions about how it’s supposed to display the info.
Yes, I think so too. All things considered, I should not worry about some info below 20 hz if I can't get rid of it.

jestermgee
Posts: 4500
Joined: Mon Apr 26, 2010 6:38 am

Re: How should I understand the graph on the parametric EQ? (not a beginners question)

Post by jestermgee » Sun Dec 12, 2021 10:05 pm

No you shouldn't, I think I mentioned already (too much to read now) you will never simply just cut everything below (without making some audio issues) what you do is roll off the information so it is basically no more apparent than just background noise. All you want to do is reduce it which is what the "x4 slope) setting is for in EQ8. Setting this you will still see signal and because of everything explained it will still "look" significant but there will be such low signal it is basically non existent.

I'd really suggest hitting youtube and watching some videos on the subject, reading up on it then just accept what is happening and don't worry too much about what isn't understood, it will become clearer as time goes on.

dsu
Posts: 141
Joined: Thu Sep 02, 2010 4:22 pm
Contact:

Re: How should I understand the graph on the parametric EQ? (not a beginners question)

Post by dsu » Sun Dec 12, 2021 10:48 pm

romariozen wrote:
Wed Dec 08, 2021 2:45 pm
My question is, let's say, we have a 20 Hz wave and 10 kHz wave. Both have a specific x coordinate that has to be a straight line. Why do I see a broad hill when I monitor the low frequency contrary to the high frequency hill that almost looks like a straight line? When I have two frequencies, 20 and 21 Hz simultanelusly, It would be hell of a task to identify that there are 2 tones actually.
I've been following this thread with some interest and there are many useful observations. In a subsequent post you observed that when you increased resolution the hill got smaller but still not a single line which is what you would expect for a sine wave.

Between 20 hz and 21 hz there are an infinite number of frequencies. The question is how much resolution and accuracy does a instrument need. Isotope RX has a much higher degree of resolution than the spectral analyzer in Live but it doesn't do the analysis in real time and it costs as much as Live does.

In the old days (1970s) the standard for audio analysis was a a 1/3 octave which was "good enough" precision to address most situations. Now a days there is more power and precision but creating a hill centered around a frequency highlight seems like a reasonable trade off between cost and utility since isolated sinewaves rarely occur in the wild unless one does a lot of experimental music.

romariozen
Posts: 63
Joined: Thu Jul 25, 2019 3:47 pm

Re: How should I understand the graph on the parametric EQ? (not a beginners question)

Post by romariozen » Mon Dec 13, 2021 9:44 pm

dsu wrote:
Sun Dec 12, 2021 10:48 pm
romariozen wrote:
Wed Dec 08, 2021 2:45 pm
My question is, let's say, we have a 20 Hz wave and 10 kHz wave. Both have a specific x coordinate that has to be a straight line. Why do I see a broad hill when I monitor the low frequency contrary to the high frequency hill that almost looks like a straight line? When I have two frequencies, 20 and 21 Hz simultanelusly, It would be hell of a task to identify that there are 2 tones actually.
I've been following this thread with some interest and there are many useful observations. In a subsequent post you observed that when you increased resolution the hill got smaller but still not a single line which is what you would expect for a sine wave.

Between 20 hz and 21 hz there are an infinite number of frequencies. The question is how much resolution and accuracy does a instrument need. Isotope RX has a much higher degree of resolution than the spectral analyzer in Live but it doesn't do the analysis in real time and it costs as much as Live does.

In the old days (1970s) the standard for audio analysis was a a 1/3 octave which was "good enough" precision to address most situations. Now a days there is more power and precision but creating a hill centered around a frequency highlight seems like a reasonable trade off between cost and utility since isolated sinewaves rarely occur in the wild unless one does a lot of experimental music.
That makes sense

Post Reply