Cut or Boost? How Much

[mpompey]

I've been looking at some of the graphs folks have been posting up here. How do you get your sub responses so flat by cutting freqs alone. Do you ever give any a boost? If so, by how much. Is there a dB limit for boosting a freq range? I'm not talking about a room null. I understand that you can't electronically even that out with an eq alone.

[brucek]

Is there a dB limit for boosting a freq range?

Zero..........

How do you get your sub responses so flat by cutting freqs alone.

Turn up the sub amp and cut....

There is really no upside to boosting in the BFD.... it decreases dynamic range, reduces headroom in the DSP and reduces signal to noise ratio. Any boost at any frequency before the crossover frequency requires reduction in the input level of the BFD to make room for the boost.

brucek

[mpompey]

brucek wrote:

Zero..........


Turn up the sub amp and cut....


brucek

Of course, I never thought to do that. I've left the sub amp at the same level I used prior to the BFD and REW. It never occurred to me to turn it up; thereby raising the response of the low points, and then cutting everything else.

I feel so stupid!

[cyberbri]

On the other hand, from personal experience, if you cut too much to get the majority of frequencies down to the level of a dip, then you have to turn the gain on the sub way up. When you turn the gain up, you are amplifying "noise" from the BFD to be louder.

For example, in my case my HK receiver puts out a lot of voltage for the sub signal. So I had to greatly reduce the sub level in the receiver to make sure the signal doesn't clip going into the BFD. But that means I had to raise the gain on the sub a lot. With one pass at the BFD, I cut a ton of frequencies, and had to raise the gain on my sub past 1/2 way. Getting close to the sub I could hear noise/static coming through. Lowering the gain reduced this greatly, so I re-did my BFD filters with less cuts, and was able to get the same final level without the audibly loud noise coming through the sub.

Here's the thread where I talk about it:
http://www.hometheatershack.com/foru...igh-freqs.html

Here's my results, current filters:
Freq BW Cut/Gain
74Hz 6 -13
45Hz 8 -7
66Hz 2 -4
54Hz 8 +4
37Hz 4 +4
53Hz 3 +2

Response w/out BFD, just bass traps in the room:


After above filters:




Before that, I was using these filters:
24Hz 8 -3
30 10 -4
48.5 3 -5
40 6 -7
44.5 6 -12
53.25 3 +6
72.35 10 -16
83 6 -5
51 2 +4
89 2 +6
37.88 3 +3
100 4 -7

And this was my response:

(Granted, my speakers are now in a different position from when this one was taken, reflected in the above graphs)

[mpompey]

I think I got it. I raised the gain knob on my sub from a quarter turn to about a third and remeasured with REW. Here are the graphs of the initial (unfiltered) response and then the final one with the filter. (This is with a 5dB house curve from 30 to 100Hz)

However, after I applied the filter and checked the level of the fronts versus the sub with Avia DVD, the sub was about 4 dB lower. So I raised the gain knob on the sub to just shy of halfway and the levels matched. Did I do that correctly? I didn't think I needed to touch the signal from the receiver, as to that would mean I would need to REW the whole thing over again.

Here are the graphs:

[mpompey]

Thanks for the replies and the links to your personal history, but I noticed that you have a few boosts in your filter settings. I thought boosts were a No-no. How do you make sure you aren't taxing your sub/amp?

[cyberbri]

mpompey wrote:

I think I got it. I raised the gain knob on my sub from a quarter turn to about a third and remeasured with REW. Here are the graphs of the initial (unfiltered) response and then the final one with the filter. (This is with a 5dB house curve from 30 to 100Hz)

However, after I applied the filter and checked the level of the fronts versus the sub with Avia DVD, the sub was about 4 dB lower. So I raised the gain knob on the sub to just shy of halfway and the levels matched. Did I do that correctly? I didn't think I needed to touch the signal from the receiver, as to that would mean I would need to REW the whole thing over again.

Here are the graphs:

Yes, you did that correctly. After eqing the sub, you need to re-calibrate the sub's level.

But looking at your "after" graph, it looks like you could still use some more tweaking. Did you use REW's "find peaks" and the "suggest filters" and "optimize filters" options? Ie., some if your cuts are in the right spot, but could be widened (bigger bandwidth) for a larger bandwidth.

I think I've configured my BFD filters 6 times or so before I settled on my current filters.

For example, it looks like you put a cut at 58Hz, but it's pretty narrow so it leaves some humps on either side. Same with the cut at 20hz, although you probably don't even need that. Depending on the final level/graph, you could widen the filter, raise the frequency it hits a little, not cut it quite so much, and smooth that out. You might want more of a house curve, so there's more boom down low compared to the 40~80hz range.

Keep experimenting and you'll hit a combination of filters that really make it smooth.

[cyberbri]

mpompey wrote:

Thanks for the replies and the links to your personal history, but I noticed that you have a few boosts in your filter settings. I thought boosts were a No-no. How do you make sure you aren't taxing your sub/amp?

I cut some other frequencies quite a bit, so I'm not too worried about a few small boosts. I didn't want to go over 3~4dB boost, though.

And I checked with some very taxing scenes - The Haunting and WotW at -15 ~ -10, no chuffing, no audible distress. Those are about as hard-hitting as it will get, and it seemed okay.

You can cut and cut and cut until you get to the level of the lowest dips. But will the dips stay at the same level as the other frequencies? And since I had to reduce the level of the sub in the receiver so much (HKs put out a very strong LFE/sub signal) so I don't send a clipped signal to the BFD, I'm already pretty low. The more I cut, the higher I had to turn the gain on the sub itself, which caused audible noise. So I re-did it, used fewer filters, a big cut or two, a few small boosts, and I'm good to go.


BTW, I made a flickr account.
Here's my room with speakers, sub, and treatments:
http://www.flickr.com/photos/6850963...7594507788330/

And here's my latest graphs:
http://www.flickr.com/photos/6850963...7594507798589/

[cyberbri]

NM - just testing.

[lovingdvd]

brucek wrote:

There is really no upside to boosting in the BFD.... it decreases dynamic range, reduces headroom in the DSP and reduces signal to noise ratio. Any boost at any frequency before the crossover frequency requires reduction in the input level of the BFD to make room for the boost.
brucek

Can you please explain this in some more detail. In a couple of places I used a +2 or +3dB gain for a small bandwidth maybe 2-4 (IIRC on about 25hz and another about 60hz.

I'd like to better understand how this may be hurting things. If this one of those "in principle" things or am I definitely getting lesser dynamic range/sound from the sub as a result? Or is it something I'm really not going to notice sound or power-wise with small amounts?

In my case the sub has plenty of power. I have a nearly perfect house curve at +8dB 20 to +0dB at 80hz. The response follows the curve almost perfectly. The sub know is turned up only 1/4 of the way. When I listen to music or movies with heavy bass action scenes the sub sounds nice and strong with plenty seemingly left over if I wanted to go even louder (which I don't).

Is there a way to measure definitively whether any gains I've made no matter how small or large are hampering the subs ability to perform. Basically I turned the sub up considerably to begin with to get things at the right "starting point" and then made cuts throughout - but to balance it out I needed a couple gains here and there...

Thanks!

[Ivaols]

brucek wrote:

Zero..........


Any boost at any frequency before the crossover frequency requires reduction in the input level of the BFD to make room for the boost.

brucek

I have read that boosting is not that bad, or "less worse", if you do it on a higher frequency. Is this true?

[brucek]

I have read that boosting is not that bad, or "less worse", if you do it on a higher frequency. Is this true?

It's fine in varying amounts on any frequency that will enter the BFD at a maximum voltage that is less than the BFD's maximum input.

This would be those frequencies around and above the crossover.

With an 80Hz crossover, as an example, a 120Hz signal will always enter the BFD ~24dB lower than a 60Hz signal. Problem is, equalization is less effective as you move along along the target curve. Any eq at 120Hz would be ineffective.

Think of using gain in the BFD this way. A 20Hz signal that enters the BFD at its maximum input level will drive the yellow LED on. That's fine. At this point, how much gain can you add to a 20Hz filter to clip that 20hz signal and turn on the red LED? Zero........ gonna have to turn that input level down. Now the softest 20Hz signal just went into the noise...... So, the DSP may have a little less to work with, the difference between the maximum and minimum signal that can pass through the BFD is less, so you have less dynamic range.......................... can you hear the difference? Maybe, maybe not.

brucek

[mpompey]

Brucek, thanks for taking the time for that message. I think I'm beginning to understand. Do you have any links that might explain the whole dynamic headroom vs range thing a little more.

My next question, when setting the level between my subs and fronts, should I use the flat or the house curve filter?

[lovingdvd]

mpompey wrote:

Brucek, thanks for taking the time for that message. I think I'm beginning to understand. Do you have any links that might explain the whole dynamic headroom vs range thing a little more.

My next question, when setting the level between my subs and fronts, should I use the flat or the house curve filter?

House curve issues is addressed in the house curve sticky. The consensus seems to be that a flat response in the sub range is rather dull and I certainly found that to be the case. So at a minimum you'll want to experiment with a few various house curves and see how it sounds. Of course just for your own curiosity you may want to hear how things sound flat but be prepared to do this only as an experiment not as a "keeper" setting (most likely).

[Wayne A. Pflughaupt]

On the other hand, from personal experience, if you cut too much to get the majority of frequencies down to the level of a dip, then you have to turn the gain on the sub way up. When you turn the gain up, you are amplifying "noise" from the BFD to be louder.

Exactly. You have to look at the whole picture; the BFD is only one component in the signal chain. Assuming you have your BFD’s input level set just below clipping, if your output level is –10 or -15 dB (or even less) after equalization, that’s 10 or 15 dB (or more) you reduce the sub’s S/N when you turn it up to compensate. Well, a high gain/low signal scenario is a know recipe for audible noise.

Does that matter much at the end of the day? As cyberbri and PeteD have discovered, sometimes it matters a lot.

Typically a powered subwoofer is going to be the “weak link” in the signal chain. We know what the BFD’s maximum S/N ratio is, but most sub manufacturers don’t even publish them. Either they feel it doesn’t matter much, or it’s a figure too embarrassing to publish.

When I can find a noise spec for a powered subwoofer, it’s usually 90 dB or less, with no qualifier such as “unweighted between 20 Hz- 20 kHz” (to quote from the BFD’s spec). Without the qualifier, their 90-dB rating isn’t terribly meaningful.

But for the sake of this discussion, let’s pretend that it is. If our signal from the BFD is 10-15 dB lower than what came into it, we’ve sucked our sub’s (questionable) 90 dB S/N down to a (questionable) 75 or 80 dB when we increase its gain.

At that point it isn’t terribly relevant that you managed to maintain the BFD’s excellent >94 dB noise rating. Your overall S/N can never be higher than the weakest link in the signal chain anyway.

So, the best approach IMO is to try to get the BFD’s output close to its input. Most response curves have both peaks and valleys, so the most efficient equalizing (i.e., using the fewest filters) means a combination of boost and cut filters applied to them. After EQ you’ll often end up with the BFD’s output close to its input, so you don’t burden your sub with a low-level signal.

Regards,
Wayne

[brucek]

So, the best approach IMO is to try to get the BFD’s output close to its input. Most response curves have both peaks and valleys, so the most efficient equalizing (i.e., using the fewest filters) means a combination of boost and cut filters applied to them. After EQ you’ll often end up with the BFD’s output close to its input, so you don’t burden your sub with a low-level signal.

Is "using the fewest filters" our ultimate goal?

Whether you apply gain filters to response dips (and turn the BFD's input level down to allow it), or you apply cut filters to areas around response dips (and setup the input level to maximum), it makes no difference to the output signal level of the BFD. The former example reduces dynamic range and degrades signal to noise ratio, the latter example does not.

Which do you choose?

Setting a best input level maximizes the number of bits used in the BFD. The signal to noise ratio in a digital device is quite sensitive to input level. The BFD is a 24bit device. The ADC has a fixed theoretical noise floor of 147 dB that is a function of bit resolution (6.125x24=~147db). The spec for the BFD is ~94dB, so we know already that not only is the LSB hidden in the noise, but quite a few others. The spec of 94dB equates to between 15 and 16 bits of actual resolution. We have about 8 bits lost in the noise.

As you lower the input signal to the BFD, the noise internally rises exponentially. We know this, since each digital bit we lose drops about 6dB from our noise figure. As we lower the input level, the softest signal we can resolve (and therefore more low order bits) are lost in the noise. The BFD system becomes a 15bit, 14bit, 13bit device. An exponential drop in dynamic range occurs.

Since we know that we can obtain the same output level whether we use cut or gain, why choose gain when it creates a problem as I've described.

brucek

[Ayreonaut]

A bit of boost is not a problem if it is in combination with cuts so that the sum of all the filters results in no net gain.

[cyberbri]

Ayreonaut wrote:

A bit of boost is not a problem if it is in combination with cuts so that the sum of all the filters results in no net gain.

That's what I assumed and hoped was correct - that my large cuts and several small boosts would average out or cancel each other out.

brucek wrote:

It's fine in varying amounts on any frequency that will enter the BFD at a maximum voltage that is less than the BFD's maximum input.

This would be those frequencies around and above the crossover.brucek

brucek, perhaps you could share your thoughts on whether the following example (well below the x-over) would constitute a boost?

For our example let’s say we have a peak at 35hz that’s 15db above the target line. When we cut it down to the target we now find the filter has created a new 6db dip at 28hz.

So here lies the question: Is creating a filter at 28hz with a positive value (gain) to counter the effects of the neighboring cut filter considered a boost simply because the filter has been set with a gain?

Obviously, if you add enough gain to push that filter above the “flat filter” target line that REW also displays, then I can see it clearly being a boost. But I’m not talking about that.

Rather, let’s say our 28hz filter has +5db gain, yet the net result still remains below the flat filter line (due to the effect of the neighboring cut filter), how would this compromise BFD headroom?

Thanks

[brucek]

that my large cuts and several small boosts would average out or cancel each other out.

Unfortunately, it doesn't work that way..............

brucek

[cyberbri]

brucek wrote:

Unfortunately, it doesn't work that way..............

brucek

Technically, no, because a sound effect or musical note could hit just the boosted frequency and not the cut frequency. So the benefit of the cut frequency is not there, with only the drawback of the boosted frequency. But overall, without pushing music to max volume, I would assume that the "torture test" scenes in movies would the WoTWs and Hauntings and LOTRS, where bass is hitting at many, many frequencies over the range.

[Ayreonaut]

I should have added "at any frequency."

Keep an eye on the "filters" graph in REW and make sure that it doesn't go over the target level at any point.

You can use positive filters to help shape the EQ curve so that the wide cuts don't further attenuate your measured dips, but you can't use positive filters to boost the level at the dips.

For example here's one with two filters. Note that there is no net gain at 53.3 Hz despite the large boost, because the sum of the two filters is always less than (or equal to) zero. Thus, we do not further attenuate our original dip, but we don't boost out of it either.

41.5 Hz, -14 dB, 1/1 octave wide
53.3 Hz, +12 dB, 1/6 ocatave wide

[brucek]

I should have added "at any frequency."

Yeah, that's correct.

perhaps you could share your thoughts on whether the following example (well below the x-over) would constitute a boost?

There's a couple points to make. Indeed, as ayreonaut implies, if I have a +10dB filter at 50Hz and a -10dB filter at 50Hz, the resultant effect on the output signal is 0dB. In that regard, although I gave an extreme example, I can use gain to shape a negative filter without providing any gain to the input signal and as such, the output signal will not clip (requiring a reduction in the input level to compensate).

My point in my upthread posts is that if you provide any resultant gain in the input signal, you will have to reduce your input level to the BFD (with your processor subwoofer output trim). This lowered input level to the BFD causes negative effects as I discussed upthread. I am simply trying to get the maximum signal to the A/D converter. You can't do that if you apply gain because the output will clip.

A +10dB filter at 50Hz and a -10dB filter at 50Hz doesn't constitute gain in the output. I might point out though, excessive filter gain or cut value are better served divided into smaller bites. There will be less internal DSP distortion to two -5dB cuts at 20Hz than a single -10dB cut at 20Hz. Divide your big filters in half.

brucek

[cyberbri]

Good advice about splitting large cuts into several filters. I may try that with my -13 cut at 74Hz. I'll also look into adding a bit of boost to any areas around the dip/null around 85~95Hz that were lowered because of the cuts and see what happens.

With regards to input level into the BFD, I used ch5 of WotW DTS to set the input level by checking for the peak. I think I turned the sub off, but had the volume at the max level I would probably play that scene at (-12 to -10 or so, IIRC) and watched the input level meter on the BFD.

I ended up raising the speaker levels and lowering the sub levels in the receiver to get proper calibration with a low enough sub output into the BFD. Before the sub was 3~4 notches below the level of the speakers in the receiver, with the sub's gain at about 1/2 way between 0 and the first tick on the HSU gain knob. But since my HK puts out a very strong sub signal, I raised the speaker levels to about +2~+3 all around, and have the sub level at -9 for DD/DTS. This is re-calibrating the receiver so that -5 on the dial is reference level (-10 = 80dB with Avia, so 5 under) rather than 0 equalling RL 0.


This is all great advice. I'll re-evaluate my settings, although I think my small boosts are fine as the sub hasn't shown any stress on even The Haunting DTS or WotW DTS at high volume (max volume I would watch at).

Yes! It seems we have a complete answer to the original question that started this thread.

mpompey wrote:

I've been looking at some of the graphs folks have been posting up here. How do you get your sub responses so flat by cutting freqs alone. Do you ever give any a boost? If so, by how much. Is there a dB limit for boosting a freq range? I'm not talking about a room null. I understand that you can't electronically even that out with an eq alone.

And the following reply:

Is there a dB limit for boosting a freq range?

Zero..........

and

How do you get your sub responses so flat by cutting freqs alone.

Turn up the sub amp and cut....
There is really no upside to boosting in the BFD.... it decreases dynamic range, reduces headroom in the DSP and reduces signal to noise ratio. Any boost at any frequency before the crossover frequency requires reduction in the input level of the BFD to make room for the boost.

You can see how someone reading this thread from the top could get the wrong impression that any filter with “any amount” of gain is to be avoided…unless it were close to or above the x-over.

That’s why it seemed further clarification of what defines “a boost” was in need.
Many could assume that any filter with any amount of gain is a boost, when that is not necessarily the case.

Utilizing the filter graph in REW as a means to evaluate filters that employ a bit of gain (but no actual boost) to determine if you are within the safe zone is very useful technique in smoothing out your response after those larger cuts needed to tame peaks down to the target line have created some modest dips in the process.

brucek wrote:

A +10dB filter at 50Hz and a -10dB filter at 50Hz doesn't constitute gain in the output. I might point out though, excessive filter gain or cut value are better served divided into smaller bites. There will be less internal DSP distortion to two -5dB cuts at 20Hz than a single -10dB cut at 20Hz. Divide your big filters in half.

brucek

brucek, In dealing with that last question you brought up the technique of splitting larger cuts into two. Do you have any basic guidelines as to when a cut should be split?

I’ve found myself doing this of sorts by using two fairly close spaced cut filters when dealing with a large odd shaped peak. However, I never knew if that was good practice or not?