harsh audio and a real time audio spectrum analyzer

[jhw]

How much precision do I need?

I have a really cheap full range worth about a nickel (always was worth about a nickel,too), and I want to see how well I can tinker this thing to perfection with cost and time as a factor.

Was thinking of a real time analyzer to assist in this deed and became stumped on what level of precision is practical to fight the harsh frequencies. Will 1/3 octave measurements cut it or should I opt for 1/24 octave measurments.

I checked out examples of 1/3 versus 1/24 at True Audio and I'm kind of intimidated at the high level of precision offered by 1/24 measurements because at these more precise levels, one can see--charted--that the audio spikes start to become more numerous and razor thin.

Is it practical to think that an audio expert would be able to subdue these individual razor thin spikes--using 1/24 data--economically while not butchering the sound quality (via collateral damage to neighboring frequencies); getting sound very pleasing, flat and/or accurate.

Or could the same expert take the same task (but with 1/3 data) and get the speaker almost as good as with 1/24; just not quite as accurate; but much cheaper, and much quicker.

And how bout if my speaker budget went from a nickel to upwards of $300 for the speakers for three-way designs. Does this change the answer by much i.e. 1/3 vs. 1/24?

[ErinH]

Re: harsh audio and a real time audio spectrum analyzer

keep in mind the use of an RTA; it's not ideal for driver/speaker measuring for quasi-anechoic purposes. it's a system response measurer... and by 'system', I mean speaker+room. Essentially the same thing as not gating an impulse response and letting it run out in to the 100's of milliseconds.

That said, what is your goal? It sounds like you're wanting an impulse based measurement system rather than an RTA.
In that case, there are many free programs with all the level and precision you'd ever need. If you want to go with TrueRTA for RTA purposes, then it also depends on what you want to measure. For room modes, you need software that will measure high Q values... something at least on the order of 10 (the higher the Q #, the more narrow the mode/boost/cut; think of a parametric EQ) to view modes. So, using 1/6 octave resolution will get you to 8.6 Q. This means that with 1/6 octave and below, the highest Q value you can measure is 8.6. If you want to measure higher, you'll need finer resolution. 1/12 will get you a Q of ~17.
On the flip side, if you don't need that kind of accuracy then 1/3 will probably suffice... especially if you're using most modern DSP's.
Higher resolution is always a great thing but it can be confusing. I use 1/12 or greater to figure out where there are resonance issues below the transfer function of the room. Above that, I just use 1/3 to EQ with my 1/3 equalizer.

http://www.sengpielaudio.com/calculator-bandwidth.htm

YMMV... and everyone seems to have different takes on what is useful and what isn't. In my humble opinion, until the low end response is right - which can't be done without taming high Q modes - you'll never quite experience what is really possible with a high end audio system. Sometimes an EQ ability isn't enough and you have to go mechanical (helmholtz absorbers or bass traps). Blame it on the room. They suck. :(

[Whitneyville1]

Re: harsh audio and a real time audio spectrum analyzer

JHW, a "perfect flat" FR speaker will probably sound pretty bad in most real-world rooms. 1/24th octave may give you so much data, you can't "see the forest for the trees". 1/3rd octave is too smooth, and makes everything look nice, including the bad stuff. People much, much smarter than me use 1/6th and 1/12th octave the most for their speaker measurements. If I understand you right, "harsh audio" is a form of non-linearity (distortion), and like we all hate, usually goes back to the drivers/crossovers not matching up well, or not matching the room acoustics. (I HATE that!) Good luck. Trust me, I've messed-up almost every way possible and invented some new ones on the way.

[jhw]

Re: harsh audio and a real time audio spectrum analyzer

Thanks for replies

To clarify my intentions, I have problems with transients on my pc speakers so I thought I could eq them away. I got to playing with my computer's ac97 10-band eq and I could definately improve things from terrible sound to "surprisingly ok" performance, but all my confidence ended when I came about this big distraction: I could drastically turn up a certain frequency, like 1k hz, and this made the audio drastically quieter and absoluting notihing to my ears got louder. And, this is not consistant. It's in relation to whether or not adjacent frequencies are up or down. Is this a phase issue, a phase issue with a single driver and no crossover as in, the room is definately at play here?

I'm puzzled on this so that's when I decided I needed a spectrum analyzer. I thought it could help clear up what seems to be nonsense.

So, then it seems I should measure for speaker+room?

Or another approach?

Thanks again

[billfitzmaurice]

Re: harsh audio and a real time audio spectrum analyzer

How much precision do I need?

What are you using for EQ? There's no point in measuring with higher resolution than what your EQ can address.

[Sydney]

Re: harsh audio and a real time audio spectrum analyzer

so I thought I could eq them away. I got to playing with my computer's ac97 10-band eq and I could definately improve things from terrible sound to "surprisingly ok" performance,

EQ cannot fix all issues and 10 band is not sufficient in control; in professional scenarios 31 band ( 1/3 octave ) are the minimum.

[ocool_15]

Re: harsh audio and a real time audio spectrum analyzer

EQ cannot fix all issues and 10 band is not sufficient in control; in professional scenarios 31 band ( 1/3 octave ) are the minimum.

I agree completely.

There are some players with better EQs built in. I believe Winamp would still have 1/3 octave eq available.

[jhw]

Re: harsh audio and a real time audio spectrum analyzer

Points all taken, so nothing's stopping me from upgrading my eq, except, I still lack the understanding why--using my software Ac-97 Audio Configuration Eq (utility came with my computer)-- boosting a frequency would make the perceptible audio quieter and why cutting that frequency would make everything seem louder. How can I go further with alleged nonsense in the equation. Can someone clear this one up? Is it just software failure or is this a real issue?

Technically, with the phenomenon I experienced, is this remotely possible: I build a notch filter to cut a frequency and once incorporated because it slightly missed its' mark frequency wise or because the values of components were 8% off, does this mean that perceptible audio can become louder.

[whatatrip]

Re: harsh audio and a real time audio spectrum analyzer

Thanks for replies

To clarify my intentions, I have problems with transients on my pc speakers so I thought I could eq them away. I got to playing with my computer's ac97 10-band eq and I could definately improve things from terrible sound to "surprisingly ok" performance, but all my confidence ended when I came about this big distraction: I could drastically turn up a certain frequency, like 1k hz, and this made the audio drastically quieter and absoluting notihing to my ears got louder. And, this is not consistant. It's in relation to whether or not adjacent frequencies are up or down. Is this a phase issue, a phase issue with a single driver and no crossover as in, the room is definately at play here?



Thanks again

What happens when you reverse the polarity of one of the speakers?

[jhw]

Re: harsh audio and a real time audio spectrum analyzer

What happens when you reverse the polarity of one of the speakers?

No change; still a problem.

Here's the specifics: I always start off resetting the the eq starting from scratch with all bands at the zero mark. Raising any single band will raise volume as expected and vice versa when lowering levels. The failure occurs when I lower one band and additionally lower a second band two bands over; skipping one. The problem now becomes that the band I skipped over, inbetween, gets this bizarro opposite effect, so when I try raising it, the overall volume level sinks and vice versa, if I were to lower the band. Not all bands are affected in this way; I'd say a couple a frequencies out of the 10.

It actually took me several minutes of failure before I could recreate this situation, so I almost started to think I was imagining the whole thing. But now I am most certain it is real. Is this a phasing issue?

[fbov]

Re: harsh audio and a real time audio spectrum analyzer

Key piece to me is:
"...using my software Ac-97 Audio Configuration Eq (utility came with my computer)..."

Do you know what the SW is actually doing? This could be an auto-leveling feature of some sort, like a peak limiter but for the average level. The other thought is wide frequency bands, where the adjacent filters overlap to an extent that renders them no longer independent.

Sometimes, things don't work the way we expect them to... you may be chasing a SW bug.

HAve fun,
Frank

[Dirk]

Re: harsh audio and a real time audio spectrum analyzer

Do you know what the SW is actually doing? This could be an auto-leveling feature of some sort, like a peak limiter but for the average level. The other thought is wide frequency bands, where the adjacent filters overlap to an extent that renders them no longer independent.

I'll bet a it's *exactly* that. Let's say the sound card has a max output of 1. By default, the preamp sets the output of the mixer at .8. You raise that slider by 4 steps of .1 each. Now the max output needs to be 1.2 in order for the unclipped waveform to leave the sound card. But it can't, since 1.0 is the max. So the sound card is taking it upon itself to create the headroom by making everything quieter.

Does that EQ have a "preamp" slider? If you are EQing a channel up by 4 notches, you need to take the preamp slider down by the same amount. It should NOT need to be additive, I don't think. Raise two of them by 4 and you should not need to lower the preamp slider by 8.

Mess around with it.