Why is it better to reduce levels with the EQ rather than boost?

[Doodski]

@amirm always reduces levels and others have recommended the same. Why?

 

[gene_stl]

I second the question.?

 

[noobie1]

I always thought the nulls are due to waves being out of phase at those frequencies. If you boost your signal by X amount, the out of phase reflected signal will be reduced by X amount. Or something like that.

 

[warnerwh]

One problem you may run into boosting the nulls is over doing it. Boosting only 3db is now drawing double the power. Being as 3db is not much more than audible it is easy to go to say 6db and now you are putting 4 times the power to those frequencies. If you know what you are dealing with boosting nulls can be done safely.

 

[pozz]

Attenuating is generally safer.

In the signal domain, if you boost you run the risk of clipping or overload if you don't watch the gain staging.

Acoustically, there are room modes and there are standing waves, and EQ is effective only for the latter (room modes are related to the dimensions of the room, while standing waves are constructive/destructive interference patterns). The more energy you put into it those regions, stronger the interference becomes. In other words boosting can deepen the null you're trying to fix.

Say your loudspeaker/headphone has a set of drivers where the upper end of the response is basically tuned break-up, or the end result produces an unexpected null due to the way the driver enclosure/baffle is designed. The same acoustic rule applies there: attenuate or risk making the problem worse by exceeding power handling, your amp's power delivery, or increasing distortion.

@warnerwh Said it more plainly: boost if necessary, but don't overdo it.

 

[restorer-john]

@amirm always reduces levels and others have recommended the same. Why?

Always? I don't think that's quite the case. If you have a slightly rolled off top end or bottom end, you aren't going to reduce the entire spectrum to match. You'd introduce a bit of targeted boost.

A slightly recessed midrange on a two way? A bit of boost.

Trouble with gain is too much of it. Graphic equalizers of the past with +10dB on every slider...

Now with Parametric EQ in the digital domain you can be quite precise with your Q.

 

[Vasr]

@amirm always reduces levels and others have recommended the same. Why?

You can boost but with 3 caveats one needs to understand:

(This is for general EQ upstream before a DAC and amp as a separate processing step, not for any EQ you might do inside an active speaker. The latter may avoid some of these problems as they have full control over the amp and signal processing)

1. Boost requires more power from the amp. Every 3db boost doubles the power needed. At higher frequencies this is not usually an issue but at lower frequencies this can stress the amp and lead to clipping if playing at loud volumes and the amp does not have the headroom required for the boost.

2. It may not be effective depending on why you need to boost. If it is because of a deficiency of the speaker, you cannot always boost it to correct. The boosting by asking the amp to supply more power may hit the limits of cone excursion or whatever the limits of the speaker are. So, for example, one should never try to extend the bass for a speaker by boosting at low frequencies. If it is because of the room mode, it depends on whether it is a null (direct and reflected wave canceling out) which cannot be fixed by boosting the power at that frequency. It will cancel out with higher direct and reflected waves. So, you are just wasting amp power and risking hitting speaker limits.

3. It may lead to digital clipping or reduced dynamic range if this correction is being made in the digital domain as room corrections typically are. To understand this, when you specify a filter to boost, what you are doing is changing the digital volume level of the signal at the frequencies affected by the filter. So, if a signal came in with -3db volume at a frequency and you boost it at that frequency by 3db, the digital signal is 0db at that frequency when it reaches the downstream DAC. BUT, what if the signal came in at -2db volume (louder) at that frequency? Remember that recorded content can have signal peak volumes up to 0db and can only represent values of 0db or less. So your boost will lead to what is called digital clipping. This usually has very bad effects on the signal as you lose information in the signal and it may even introduce noise depending on how it is handled in your device.

So, typically room eq systems reduce the overall digital volume of the signal by a certain amount (the max boost they are willing to do). They can vary from 3db to 10db or more. This way it will never digitally clip but the overall volume is reduced across the spectrum. With this on, the DAC will produce lesser voltages in analog and the amps will output less voltage and the speakers will play at lower volume. Essentially, your max volume has been capped lower and so you have lost the full dynamic range possible without that adjustment.

So, the less boost you do, less of the above problems occurring in the digital domain.

Attenuation does not suffer from the above issues and generally safer to do although it may reduce the dynamic range a bit also depending on what is being done with the filters.

 

[RayDunzl]

Why is it better to reduce levels with the EQ rather than boost?

You need to drop the level of the entire signal before applying digital EQ.

Why?

Clipping.

It is non-intuitive.


My experiment:

Using REW, create -3dBFS pink noise.

Peak Sample received at the end of a digital chain is -0.58dBFS. See the little box, top right on the chart.

Later in the chain, apply a cut, not a boost, of 6dB at 500Hz with a Q of 1.

The signal is now clipping. Peak sample is 0.00dBFS

Why?

I defer to anyone who has the expertise in DSP to explain.

 

[Blumlein 88]

You need to drop the level of the entire signal before applying digital EQ.

Why?

Clipping.

It is non-intuitive.


My experiment:

Using REW, create -3dBFS pink noise.

Peak Sample received at the end of a digital chain is -0.58dBFS. See the little box, top right on the chart.

View attachment 94821



Later in the chain, apply a cut, not a boost, of 6dB at 500Hz with a Q of 1.

The signal is now clipping. Peak sample is 0.00dBFS


View attachment 94822


Why?

I defer to anyone who has the expertise in DSP to explain.

No expert, my guess, and it isn't just digital. Some random bits of your noise were partly canceling other parts being out of phase. You reduced those with your filter, and the result is other frequencies no long partly canceled are stronger and therefore strong enough to clip. Wouldn't take too much. In pure analog you wouldn't exceed o dbFS as there is no such, but I bet the peak level goes up just the same.

 

[wwenze]

That's the "why is the amplitude of a square wave smaller than the the amplitude of the component frequencies" behavior

But does the explanation work with random noise, where we should assume the noise adds to and subtracts from each sample with the same probability. With oscilloscope the high pass filter always results in a smaller p-p. But then in this case the source signal itself is noise...

This might be more mathematic (probability theory) than electronics... imma go revise

With DSP filters tho a lazy explanation I can think of is the ringing in the response caused by the filter. My EqualizerAPO says it doesn't, but I'm going to test it.


Experiment: Using Audacity-generated 0dB pink noise file (so the exact noise is played back in both cases), WMP playback and Audacity record, with EqualizerAPO for the -6dB peak filter, peak is -0.213. Without the filter, peak is -0.028.

I think the cancellation hypothesis makes sense, because I look at the waveform and seems like only a few peaks even come close to hitting clipping. So we just need one of them to be... er... de-canceled by the peak filter.

 

[solderdude]

@amirm always reduces levels and others have recommended the same. Why?

There is a difference that needs to be established between analog and digital EQ.
Amir uses digital EQ so either he needs to 'boost' and lower the digital pre-amp to accommodate for the needed bits which also lowers the overall loudness of the original signal or you lower everything that does not need a boost and (almost) ensure you will not 'clip' the digital signal.

The digital described signal is an addition of signals of all instruments/voices in the recording. The peaks in that signal usually are low frequencies with higher frequencies riding on top of it (and each other). You cannot go above 0dBFS in the digital domain which is where most recordings already are extremely close to or are hitting.
Well.. with upsampling or filtering you can go over the 0dBFS (inter-sample overs) when calculating the changed waveform but you should not go there.
You should not simply because not all DACs can actually handle that.

So to prevent that you either only lower that what needs to be lowered or you do not boost and lower that what does not need to be boosted.
Some software EQ automatically lowers the overall level when a boost is used, others do not and you have to manually dial that in.
Something can be said for both methods and depends on which frequency (band) one boosts. Boosts in higher frequencies may need less attenuation than lower frequencies (because the amplitude in general is lower for higher frequencies)

The results is (almost) the same which is overall level coming out of the DAC is lower in all cases.

This brings us to the often heard complaint that amplifiers are not powerful enough. Well that's not the case in the vast majority of cases. The problem is that people apply digital EQ, lower the gain (or rest of the signal that does not need a boost) of the average signal by the boost needed.
When the needed boost is 6dB you loose 6dB (at least, some would like some room for inter-sample overs or to stay away from 0dB FS) and thus to play equally loud you need to either turn up the volume or increase gain by 6dB (or slightly more).
When you are using an inefficient headphone then you run out of gain and can't go loud enough anymore with your setup.

With analog EQ the same happens but in that case there is no loss of average signal so no need for a higher gain or more power.
The only thing you could run into is amp clipping at very loud levels because the boosted part may reach amp clipping levels.

 

[thewas]

Genelec has a good short answer on the topic

Why does GLM not fix dips on frequency response?

Often dips in the response are caused by reflections. Direct sound vs reflected sound from ceiling, floor etc. are out phase at certain frequency. If you boost this frequency, also magnitude of the reflected wave increases. Thus efficiency of boosting dips is not good.

Using parametric filters with positive gain to compensate for dips can have following negative side effects:
-Loss of headroom, as there is more gain at boost frequency. +3 dB means double power.
-Dip location and shape depends on listening location. When listening location changes, center frequency of the dip can change. At worst case, there is a peak at same frequency where the dip used to be. Which is amplified even more by the boosting parametric filter.
-Positive notch causes ringing in time domain. This is more likely be audible above low frequencies and high Q-values.
-Unsymmetrical notches will change phase response of the speaker, compromising stereo imaging. This is more likely to happen above low frequencies.

Source: https://support.genelec.com/hc/en-u...-does-GLM-not-fix-dips-on-frequency-response-

From that I would underline for this question Positive notch causes ringing in time domain. This is more likely be audible above low frequencies and high Q-values.

 

[detlev24]

Here's another great explanation incl. measurements, by Ethan Winer: Assessing Room EQ - Part 3

There's a CONCLUSION, with a recent UPDATE #2, if you would like to skip the experimentation part.

====
Another great, general, explanation about room EQ can be found in the REW help files: Why Can't I Fix All my Acoustic Problems with EQ?