Why is high level to cure dips a bad thing?

[atledreier]

Hi!

I would think that having an uncorrected curve that was mostly above the target and using EQ to pull it down would be a good thing. I've been adviced that it's not, but haven't really got a good explanation why. Anyone care to explain this to me?

[brucek]

I think whether you consider it a bad thing or not is a matter of degree.

Forget about targets for a moment because that's simply an REW method of determining filters. It has nothing to do with the subwoofer outside of a measurement situation.

Remember that the voltage waveform that enters the BFD equalizer is perfect. There's no dips or peaks, and if you use no filters, that perfect waveform is fed to the subwoofer amplifier. The dips and peaks are a result of the room getting its hands on the signal.

Consider a situation where you had a perfect subwoofer response, except for a single dip at 50Hz. You decide to introduce a BFD filter to cut all the subwoofer signal down to the bottom of the dip to get a flat room response. Unfortunately the dips bottom was so deep, you had to cut almost all your signal to reach it. You won the war but lost the battle, since there's no voltage coming out of the BFD now except for the 50Hz signal left over that you didn't cut. You cut all the rest of the frequencies away. I can't really turn up my subwoofer amplifer to compensate for this situation or I would clip the amp with the 50Hz signal.....

See the problem.............. raising the entire waveform in REW and cutting everything to get a flat response is the same as I describe above.......

brucek

[Blaser]

Propably raising the dip (if not too wide) won't do much harm

[atledreier]

I see.. So in effect you're doing exactly the same as boosting the dip... Interesting, and well explained. Thanks!

[OvalNut]

Hey bruce,

Is that actually a true way to put it, that cutting peaks way down is electronically/functionally the same as adding alot of boost to a dip? Assuming that it's truly a deep dip and not a null in this case.

Tim

[brucek]

Is that actually a true way to put it, that cutting peaks way down is electronically/functionally the same as adding alot of boost to a dip?

No, because you have to consider the input level too.

If I feed a proper maximum input level to a BFD, I can't add any gain filter or the output will clip, so if I wanted to boost any frequency I would have to lower the input level to do so. Not the case if I simply wanted to cut. The input level would remain the same. My point to atledreier was that you can go too far and cut too much until you have no signal left.....

brucek

[OvalNut]

And the more you cut a given frequency, the less bits you then have available for that frequency?


Tim

[brucek]

And the more you cut a given frequency, the less bits you then have available for that frequency?

No. That concern is with respect to providing enough input analog signal level into the ADC (analog to digital converter). We would like the input signal to be as high as possible to make use of as many bits as possible.

The concern about too much cut is with respect to not supplying enough voltage to the subwoofer amplifier. If we cut too much, there simply isn't any signal left for the subwoofer to amplify.

brucek

[OvalNut]

OK, so that's where the FBQ2496 displays an advantage over the DSP1124P due it's wider dynamic range, higher S/N ratio? Even after any cuts there's more output dynamic headroom left for the signal with the FBQ2496?

I really appreciate the lessons.


Tim

[Wayne A. Pflughaupt]

Ovalnut,

Maybe you’re confusing the two? S/N is a noise floor spec – i.e., residual noise with the unit at idle (i.e., no input signal). Dynamic range is a “maximum signal level before clipping” spec.

In any event, brucek was talking about cutting as it affects the downstream equipment, typically the sub amp.

It’s kinda hard to compare some of the BFD and FBQ’s specs because Behringer uses different parameters to define them.

Also, if you post a question on the ProSoundWeb or Tape-Op Message Board telling them how important input signals levels are for digital equalizers (or any other modern piece of pro-audio digital processing gear), be prepared to be roundly and soundly thrashed. They’ll tell you it’s a virtual non-issue. I know, because I’ve done it.

One tart fellow told me, “You forgot to mention that you’re an “old” digital engineer. That mindset is reminiscent of the not-so-good old days when we were struggling with first 8 bit, then 12 bit, then 16 bit a/d conversions. The noise floor was not noise at all but a signal dependant quantization distortion, nasty stuff.

"Modern higher bit rate codecs have a noise floor that is literally noise, so appropriate gain staging is very similar to analog paths (i.e. keep it out of the “dirt” without clipping).

"One remaining issue is the mental leap between FS metering where 0dB = clipping, and the more familiar 0dB at some nominal level 20dB or so below clipping, but folks can adjust to that.

"Relax, with well engineered modern gear it's not an issue (IMO).”

Here are a few other quotes I garnered:

“There's a standard of sorts that says that -18 dBFS is equivalent to +4 dBu or whatever the device's nominal I/O level may be. 0 dBFS works out to be the analog power-supply rails (or slightly less), the idea being that the ADC clips at the same time as the analog circuitry ahead of it.”

“Another thing is that the only limit to headroom in the digital domain is the architecture of the processor. Not the A/D converters. It is quite common for the DSP to operate at bit depths much greater than the A/D input.”

“It's a combination of both [the converters and the architecture of the processor]... the modern A/Ds do capture digital codes below their own noise floor. This effectively dithers the low level audio to prevent quantization distortion.
Not to get all tweaky about this, but due to the cocktail party effect we can hear signals in the presence of noise, therefore there is motivation to keep these signals clean and the distortion more than a few dB lower yet. I'm not sure what the typical target is but a few of those bits are indeed below the noise floor and productively preventing audible distortion.”

“It's also hard to convince people that the dynamic range extension is at the bottom of the dynamic scale rather than the top. In other words, 24 bit is not louder than 16 bits, it's the QUIET that is extended, insomuch as the dynamic range has extended downward, lowering the noise floor to the point where you now have 140 db of dynamic range or something ridiculous like that. So frustrating.”

Regards,
Wayne

[OvalNut]

Geez, OK now I do feel very stupidly ignorant. Wow, that's going to take awhile to digest.

One part that interests me is this last piece:

... In other words, 24 bit is not louder than 16 bits, it's the QUIET that is extended, insomuch as the dynamic range has extended downward, lowering the noise floor to the point where you now have 140 db of dynamic range or something ridiculous like that. ...

Tim

[Wayne A. Pflughaupt]

That’s probably more of a theoretical than reality-based dynamic range. If nothing else, the analog end of the circuitry will limit dynamic range to ~120 dBu best-case (of which the Behringer probably isn’t).

Regards,
Wayne