What are the downsides of DSP?

[Enstip]

DSP capability seems to be everywhere now and potentially a means to hide any number of audio equipment or room shortcomings. But, I wondered if it’s all upside. You don’t get something for nothing, right?

So, if I use REW to identify problems with my room, and load its recommend EQ adjustments into my device (in my case RME ADI-2 DAC FS) to produce a flatter response, have I just turned water into wine, or has something I’m not aware of been compromised in the process?

 

[fpitas]

Every point in the room can't be "right" at once. REW can make things better in the vicinity of your ears, though. That's not the fault of DSP, just the facts of sound in a room.

 

[DVDdoug]

You don’t get something for nothing, right?

Sometimes the only "cost" is money, time, or processing power, etc. There are not always trade-offs in sound quality. And you can make very-slight changes (like a very-small volume or EQ change) that changes every-single audio sample without anything being audible.

All kinds of DSP is used every day in audio production without a second thought. Digital mixing is DSP.

There are limitations and it's not always just digital signal processing but any kind of processing. You can get into trouble if you try to boost the bass but your speaker can't put-out the deep bass. You can end-up pushing your amp or speaker into distortion with frequencies you can't hear. You might even burn-up your speaker with subsonic bass.

Similarly, if you try to fix a standing wave node (dips in frequency response where the soundwaves cancel) you can easily end-up pushing your amp or speakers onto distortion without fixing the problem. But standing wave anti-nodes (peaks in frequency response) CAN be smoothed-out with DSP.

DACs are hard-limited to 0dBFS and most program material peaks at-or-near 0dB. So again, if you try to boost any frequencies you can get digital clipping if you are not careful to reduce the overall level. (Analog equalizers normally have plenty of headroom so this is much less of a problem.)

 

[amirm]

DSP needs to be performed at higher resolution than the samples and then properly converted back to sample size. If this is done and there is no overflow/clipping, there is no down side to it other than expense and some additional complexity.

 

[Enstip]

Thanks all, for your replies. I think I understand.

I have only EQ’d for my listening position which is quite specific and I’m not worried about the rest of the room, so no worries there.

The REW EQ’d measurements look better and it does sound different, but I’m not sure it sounds better. The proof of the pudding is how often I listen with EQ on versus EQ off ….. and it is mostly off.

The reason for the question was because I don’t know why I dislike the EQ’d flatter response. It doesn’t bother me - I’ll just listen to whatever I prefer - but it made me wonder whether the EQ was fixing one problem and creating another and I like to understand why.

 

[TK750]

Depending upon implementation and hardware you can encounter some additional latency, usually pretty miniscule though. Obviously the impact of this depends upon exactly what one is doing, how much additional there is and how sensitive the activity is to said latency. It can range from unnoticeable to infuriating. Please note I'm speaking about DSP in very general terms here. For your specific example it's rarely much of an issue.

 

[droid2000]

The REW EQ’d measurements look better and it does sound different, but I’m not sure it sounds better. The proof of the pudding is how often I listen with EQ on versus EQ off ….. and it is mostly off.

I think there are two common reasons for this:

1. It can sound weird if it's the first time you've heard a flat response.
2. Most people like a bump in the bass region. That's why RME has a loudness function. You can also add a house curve to REW and it'll create filters according to the new target.

 

[kemmler3D]

@DVDdoug already hit most of these, but possible downsides or unwanted effects of using DSP, especially EQ:

• Can add time delay (usually not enough to be a problem)
• Can add phase distortion (group delay, usually not enough to be a problem)
• Wrong use of EQ can easily make sound worse (e.g. trying to boost a null, correcting to an improper measurement, etc.)
• Over-use of EQ with speakers that don't have even dispersion can improve on-axis sound but make overall sound way worse.
• Over-use of FIR can add unwanted pre-ringing (i.e. impulse response starts before it's supposed to and sounds weird)
• Might cause digital clipping if used improperly
• Can easily boost bass frequencies beyond the capability of gear, causing distortion

My rule of thumb advice is to use the moving mic method with pink noise in REW to take a listening position measurement if you're not just sitting in one chair (i.e. your LP is a couch or whole room) and only correct below maybe 400hz depending on room size.

 

[DVDdoug]

A couple more thoughts...

If you were to build the best speaker you could build within a certain price range, or with certain component's or size limitations, etc., almost every speaker could be improved with some EQ.

If you make the speaker active you can build-in those tweaks in. And once you have an active speaker (with a power supply, circuit board, etc.) it's costs virtually nothing to add some analog "tweaks". I don't think DSP circuitry (plus an ADC & DAC) can be added at an "insignificant" cost, but it is generally less expensive to build the amplifier into the speaker than to have a separate amplifier so the cost is defrayed somewhat. Or if you have a speaker with digital inputs it already has a some kind of "digital processor" and a DAC and it might not be that much more expensive to use a more powerful processor for DSP. DSP can be "more precise" than analog EQ.

Almost all small studio monitors are active and I assume they all have some built-in analog or digital EQ. (I don't know much about "big" studio monitors.) Lots of big concert PA systems also seem to be using active speakers. Home subwoofers are usually active but because of tradition, and for practical reasons, most home speakers are passive. And in most homes it would be "messy" to plug 6 or more surround speakers into power outlets.

If you read Amir's headphone reviews he is almost always able to improve performance with EQ. I haven't read them all but I don't remember any headphone where he didn't recommend some EQ for improved frequency response. Again "tradition" calls-for passive headphones that are interchangeable with different headphone amps. "Theoretically", a headphone could be sold with a specific-matching headphone amp with the necessary EQ built-in. But here, the economics are reversed because it costs almost nothing to build a headphone amp into a receiver, phone, or computer, etc. The matched headphone & amp would be more expensive.

 

[Prana Ferox]

The reason for the question was because I don’t know why I dislike the EQ’d flatter response. It doesn’t bother me - I’ll just listen to whatever I prefer - but it made me wonder whether the EQ was fixing one problem and creating another and I like to understand why.

Not sure if you're accounting for this but the Harman response curve (that people seem to like) isn't flat, it's a gentle downward slope. Also, a 'flat' signal tends to sound lifeless at low volume because how your ears work, and generally needs an equal-loudness contour applied to compensate.

As far as downsides, everything in the signal chain adds noise and distortion, although it's quite likely an A/D-> D/A conversion is inaudible. Also, any active tweaks you do to the sound (from the bass and treble knobs through to complex PEQ) can screw sound up far more easily than they can improve it, if the operator doesn't know what they're doing.

 

[Keith_W]

There are two big downsides of DSP:

1. the learning curve
2. easy to get bad results if you don't know what you are doing.

DSP software varies with the amount of automation (and therefore ease of use) and features that it gives you. I have most experience with Acourate (which I use myself) and Audiolense (which my friends use). It takes me several hours to do in Acourate what Audiolense does in 15 minutes, because of the amount of automation in Audiolense. Acourate has some automation, but not much. It otherwise forces you to go through steps and make decisions. I can tell you this was a MASSIVE learning curve for me. Although I am now reasonably proficient at it, there are still things that Acourate can do that I have not explored.

 

[boxerfan88]

Every point in the room can't be "right" at once. REW can make things better in the vicinity of your ears, though. That's not the fault of DSP, just the facts of sound in a room.

Exactly right. This is my observation too. I tuned my MLP nicely, but then the left seat got a one note boom, and the right seat got elevated bass. I do tell my visitors, sit in the middle for the balanced sound, left for boom boom bass, right for enhanced bass.

MLP gets tuned well with the REW/EQ, but other seats may have some crazy amplitude response as a side effect. Well, it’s the room…

 

[dasdoing]

Every point in the room can't be "right" at once. REW can make things better in the vicinity of your ears, though. That's not the fault of DSP, just the facts of sound in a room.

this was like the go-to "EQ doesn't fix things" answer back when I started EQing.
it is not wrong, BUT without EQing every point in the room is even wronger lol.
I mean, you optimize to one area, most other points will still be better than before, especially in the bass

 

[Svend P]

If DSP doesn't have any real downsides, what is the reason the majority in this poll only uses it to the Schroder frequency?

Should we correct to Schroder, or full range?

I would like to gather opinions from ASR members as to whether we should apply room correction to the Schroder frequency, or whether we should do full range correction from 20Hz - 20kHz. There appear to be two schools of thought: - Correction to Schroder: aims to only correct bass peaks and...

www.audiosciencereview.com

 

[Sokel]

The downside I can think of (apart from devices that messes with it as we saw in the other thread on some occasions) is what Mr Toole says all the time,if someone gives you a hammer everything looks like a nail.
People filling dips far beyond amps and speakers ability,people forcing curves above Schroder to speakers EQ un-friendly,people leaving system with no headroom,etc.
Best thing,is ASK,either experts you pay or people you trust online.

 

[Bjorn]

Biggest pitfall IMO with DSP is applying EQ to what's non minimum phase behaviour (like for instance most of the room response). Something that's done a lot and with quite severe audible artifacts in many cases. I believe this is one of the main reasons DSP is largerly rejected by the high-end community. Phase distortion doesn't sound right.

 

[flipflop]

If DSP doesn't have any real downsides, what is the reason the majority in this poll only uses it to the Schroder frequency?

Should we correct to Schroder, or full range?

I would like to gather opinions from ASR members as to whether we should apply room correction to the Schroder frequency, or whether we should do full range correction from 20Hz - 20kHz. There appear to be two schools of thought: - Correction to Schroder: aims to only correct bass peaks and...

www.audiosciencereview.com

Because, if you read past the poorly phrased poll question, OP is actually asking about equalizing the in-room response.

 

[Hayabusa]

DSP capability seems to be everywhere now and potentially a means to hide any number of audio equipment or room shortcomings. But, I wondered if it’s all upside. You don’t get something for nothing, right?

So, if I use REW to identify problems with my room, and load its recommend EQ adjustments into my device (in my case RME ADI-2 DAC FS) to produce a flatter response, have I just turned water into wine, or has something I’m not aware of been compromised in the process?

latency

 

[Ken Tajalli]

DSP needs to be performed at higher resolution than the samples and then properly converted back to sample size. If this is done and there is no overflow/clipping, there is no down side to it other than expense and some additional complexity.

An excellent point that many don't realize.
Many assume: If it is digital, then it is perfect!
But this brings us to the old question:
How long is a piece of string?
How much Higher resolution? How to convert back (or decimate?), before it becomes audibly transparent?

 

[vicenzo_del_paris]

I do apply full range correction with Dirac instead of limiting to Schroder frequency.
Most of the enhancement (FR, impulse and decay) is obviously below 1khz.
My room is untreated and doing full range vs Schroder frequency only raises down high frequencies (>10khz) a bit (1-2db) and globally reduce decay. Apart that there is no noticeable difference (to me and REW results) in my environment.