(Def. RTA=Real Time Analysis. For our purposes, it is the process of using a spectrum analyzer and a calibrated microphone to help you equalize your car stereo system)

As discussed in this thread, I'm trying out a downloadable demo copy of the SpectraLAB RTA software, using a Shure SM-58 microphone and their included mic calibration file for the SM-58.

Spent a couple hours with it tonight. Conclusions:

1) I need to get the patched version of the Hijack kernel from Genixia, the one that doesn't put out stuff on the serial port when you mess with the EQ settings. The sluggish response of all the EQ screens due to the serial dumping made the process maddening.

2) I need to turn off that disk indicator icon. It covers up the Q number. Always popping up at the most inopportune moment.

3) Having the remote control and a clearly-readable VFD is a great feature! Could run up the headache-inducing test tones while outside of the car. Just use the remote through the window and look at the display. Very nice!

4) I had no idea how badly the default auto-Q settings will drag neighboring frequencies when you move a frequency band. The only way this is going to work is if I set a fixed Q for every single band. With the RTA software, I could see clearly that whenever I was moving one band, it was also dragging frequencies far into the neighboring bands.

Does anyone remember if beta 11 has the bug fixed where the Q isn't right for locked L/R mode and a non-automatic Q? Or was that a post-beta-11 fix?

Anyway, I'll stab at this one again over the weekend, with the proper stuff done to the player to make it less difficult.

I'm impressed,

I've never even thought about this use for the microphone connection before, it's such a great idea!

Keep at it man,

I need to clarify that I did not use the microphone connection on the car player. It cannot sample high frequncies and is therefore not ideal for RTA.

I lugged my PC out to the garage and ran the SpectraLAB software on the PC. I transferred the sound-generated test tones as WAV files to the player and created an "Ignore As Child" playlist called "(Test Tones)".

ohh, sorry, too busy speed reading and jumping to conclusions!

Awesome! Talk about a defining need for a laptop...

I've been intending to RTA my sysetm for, oh, about a year now. Have you found that the ouput from the empeg itself is flat, or does it add it's own colorations? Right now on mine, everything through the whole system is set flat, and it sounds pretty good.

Does anyone remember if beta 11 has the bug fixed where the Q isn't right for locked L/R mode and a non-automatic Q? Or was that a post-beta-11 fix?


Alledgely FITNR, but beta12f had other issues

hijack253-eq_stutter_fix.zimage

Let me know if my paranoid firewall gives you any grief.

Have you found that the ouput from the empeg itself is flat, or does it add it's own colorations?

Although this has been mentioned several times before, I have verified this personally with the SpectraLAB software: The RCA outputs of the empeg are absolutely RULER-FLAT across the entire spectrum.

I checked this by creating a set of combined overlapping frequency tones and saving those to a .WAV file. Each tone should look like a "spike" on the spectrum analyzer, with the peaks all being in the same place.

I plugged the empeg's RCA outputs into the mic input of my sound card, and after messing with the levels to get it all right, ran the SpectraLAB program and looked at the output from this WAV file playing on the empeg. Absolutely ruler-flat across the spectrum, the peaks were all exactly at the same level.

But my purpose for doing this wasn't to verify that the empeg's outputs are flat, we've already known this for a long time. What I needed to do was to find the holes in the amps/speakers/environment of my car. This is what equalization is all about. I've been doing this by ear for a long time, I just wanted a machine to verify what I think is a good EQ curve from my player. My first attempt had some snags, and with what I've learned, I'm going to try again this weekend.

Tony, please share your experiences with a little "how to". I think you have experience explaining things to us! I'm very excited to do this and never thought of being able to do it through the window!

please share your experiences with a little "how to".

Well, basically, I did this:

- Get the downloadable demo of SpectraLAB.

- Mess with it and learn how to use it.

- Test it with a little cheapo microphone, spectrum-analyzing my voice and the room noise.

- Create some test tones. Specifically, I did a pink noise track as well as a "ten frequency bands" track. The ten frequency bands I created I chose to be the same as the ten frequency centers on the empeg equalizer. I did these in the SpectraLAB noise generator, and used the "save as WAV file" option.

- I also created a twenty-band track which was the ten Empeg bands plus ten more centered between the other ten bands (had to do a little math to find the midpoints between the bands). Then I took the center-bands track and the between-bands track and combined (mixed) them in CoolEdit. This one turns out to be a very useful track, by the way, in determining what the Q setting should be for a given band.

- Locate one of the microphones for which SpectraLAB has a reference file, in my case a Shure SM-58 (I have a friend with an extensive home studio). Loaded up the SM-58 profile in SpectraLAB's "Scaling" dialog box.

- Lug everything out to the garage. Computer equipment is outside the car. Mike cable runs into the car. I discovered that the mike has a very wide dispersion pattern and it mattered very little if it was exactly at my ear position, or just sitting on its little stand on the pocket between the seats. So that's where it's sitting, on a little stand on the pocket between the two front seats.

- Set everything on the player to flat. Loudness, Bass/Treble boost, L/R delay, etc. Turn the volume way down.

- Pick a "blank" equalizer preset and name it RTA 1 or something. Don't mess up your existing presets.

- Play the test tones and mess with your crossovers, amp levels, and EQ settings. Empeg settings can be done with the IR remote control from outside the car with the doors closed (don't damage that mic cable).

Remember while you are doing this, don't blow your speakers by playing the test tracks too loud. These really work your speakers hard, so just be real careful.

Remember that the goal is not to make your car sound perfectly flat. That would sound terrible. The goal is to find unexpected holes (sharp peaks or dips) in the overall frequency spectrum that need correction. For instance, locating a dip where your subwoofer and main speakers cross over would indicate that you need to adjust the crossover points.

After you've located the holes and corrected them, then you can start to work on dealing with the overall sound. I don't know what a "pleasing curve" would look like on the graph, and I get the feeling that this is going to be different for different people's tastes.

I don't have all of this dialed in yet, I haven't done my second pass yet.

Remember that if you intend to use hijack's bass or treble controls, that these will overload channels 9 & 10 of the eq when they are active. So make sure that the most important corrections are done in bands 1-8 !
(If you are using 4 channel eq, then that is band 5 in each channel that is overloaded - use bands 1-4 for your eq)

Okay, I messed with it some more this afternoon, and had a lot more success.

Since today seems to be Digital Camera Day at the Fabris household, I took pictures. First picture: The high-tech, ultra lightweight, ultraportable field RTA system, packed and ready to go:

Second picture: The vehicle in the multimillion dollar, acoustically-perfect anechoic chamber facility, ready for testing with the portable RTA system deployed.



Note the critical acoustic-isolation mounts for the RTA system.

Final photo:

The carefully placed, very expensive, studio-grade condensor microphone, shown here on its vibration-dampening isolation mount.

So here's what I learned from this process:

All of my prior "by ear" adjustments to the EQ did not take into account the fact that there were definitely some sharp notches and peaks in the overall spectrum. I found some surprisingly narrow bands that needed correction.

In the past, with my "by ear" adjustments, a narrow peak in the midrange spectrum would sound irritating to my ear, so I would reduce the overall midrange of the system to get rid of the irritating peak. What I didn't realize until today was that I was killing all sorts of other perfectly good frequencies in the process of trying to correct that one tiny little peak.

In essence, I was making my stereo sound "less irritating" rather than "better".

Not only that, but the narrow bands that needed correction were NOT centered around the default empeg EQ frequency bands. I had to alter both the frequency and the Q of a few of the bands in the equalizer to nail the specific problems. Having the empeg's highly configurable equalizer made this SO much easier than it would have otherwise been.

The beauty of it was that I could place my mouse cursor on a specific peak in SpectraLAB, and it would give me a digital readout of the frequency. All I had to do was walk over to window, point my remote at the empeg through the window, and alter one of the EQ bands to be that exact frequency. Then mess with its level and Q until I flattened out the peak in the spectrum graph without over-affecting the neighboring bands.

I was unable to get it all perfectly flat, but that wasn't my goal. I just needed to correct major errors in the overall spectrum.

I also discovered that I had a severe hole around 100-150hz in the low end because I was overzealous with the crossover on my rear deck speakers. Correcting this by lowering the crossover and re-EQing that range made the bass sound much more natural and less boomy. I was pretty proud of how non-boomy my system sounded already, and this made it sound even better. The bass frequencies are now more natural overall, while still being loud and punchy.

Now there are whole areas of the mids and low-mids that are back in my music that weren't there before. Everything seems to have more presence and clarity. The stereo image seems better (taking into account the fact that my car's stereo image is pretty poor to begin with).

I haven't had a chance to live with it in the long term yet. I haven't even taken it on a drive to see how road noise enters into the equation, but so far, I like what I'm hearing.

Sounds good.

How many bands of eq did you end up using?

Eight. I used all but the highest two bands, because I wanted to have the option open of using your bass/treble adjustments for broader "overall tone" controls.

Something I forgot to mention about this whole process:

I get very little response out of my system above frequencies in the teens (13000 and above). Not sure if this is a limitation of the microphone, a limitaiton of my sound card's available sampling rates, or if my amps and speakers just don't do a good job of the highest frequencies. So I didn't really try to do any corrections up there, and it made it easier for me to leave those last two bands alone.

Hmm. Looking at it a little more closely:

Technically, if I leave a band at 0db, then regardless of its Q and frequency, I'm not actally using that band, right?

Although I fiddled with all eight bands during the process, at the end, I actually only corrected five bands. So I've still got three bands free and ready for future corrections.

I gave it shot today as well, but had some really weird things happening with the EQ settings. I had the EQ set to 2 channel stereo (10 bands), locked left and right. I was able to make some corrections and flatten out the curve quite a bit (it really is amazing how some of the peaks and valleys are in such a narrow frequence range). The only problem is that it seemed to only affect the right channel. I tried resetting 2-channel mode, locked L/R, etc. I have no explanation for it... anyone have any ideas? I did not try setting L and R independantly (yet).

One thing I found very helpful was to run the mic into the PC's right channel and patched into one of the empegs line outs to the left channel. Then by looking at L and R separately, I could see a clear picture of how the EQ curve was set and the effect of the "Q". It took me a little while to get a handle on "Q"...

Cheers!
Phil

Oh yeah, I forgot to re-iterate this to everyone, sorry:

There is a known bug in Beta 11. If you set a specific Q value in the equalizer, it's only working on one of the two channels. (Good catch discovering this one on your own. )

If your Q is set to "auto", you're fine, the only time this is a problem is in Beta 11 and with specified Q values other than "auto".

This bug is fixed in later betas (none of which you folks have seen yet).

I assume this bug goes back to 2.0 beta 3 as well, not certain. If you want to spec Q's, either wait for the next beta release or go back to 1.03 I think.

Bummer - well at least I know it's not operator error . Does it work properly if the L/R are unlocked and each channel adjusted independantly (to the same values)? I guess I'll have to wait until the next beta... (any hints on an ETA?).

Phil

Does it work properly if the L/R are unlocked and each channel adjusted independantly (to the same values)?

Yes, actually, I think it does work that way. So that would be another work-around.

Just now reading your RTA summary. I have no problem hearing differences above 13kHz in my system. I have some pretty nice amps and speakers (although no guarantee it makes a big difference). The default 18kHz band is only noticable with large EQ deviations and a track with a lot of high hats, etc. Although Your effort to share results is greatly appreciated.

Cory

Do you know whether it is fixed in beta-13?

That's why we write release notes..

Rob

I have actually downloaded them and the developer build, but I just zap them straight to CD for looking at when I get home. I suppose I could spend a few seconds and not be such a lazy get when in the office, but I have just had a superb bacon and brie pannini and a Kronenbourg and motivation/consciousness are low right now

Edit:-Quick flick through using a crappy text viewer at work which gives no line breaks!! And I see it is indeed done. Cheers lads.

I just finished running it on mine ... I used the Radio Smack Analogue (for the Brits) SPL Meter (33-2050) ... I found a mic calibration for this meter, but it sure didn't match mine. I went in and calculated it by hand. Dood, this meter sucks as a mike ... frequency fall-off after 5Khz was quite significant, I was at least hoping for something through 10Khz. Building my own calibration file, I got it more or less flat, but by 10Khz, I was bumping it by 20db off center. Nevertheless, it sounds better than it did (hard to believe it could get better!)

I recommend this to anyone, not just the audiophiles. It's a good experiment at any rate, expand your horizons.

I would still like to see someone do a "crudimentary" RTA/EQ-Fixer on the Empeg. Yes, we know the mic input sucks at higher frequencies, but it'll be better than most people's current EQ settings. And that input can't suck as much as this SPL meter.

Any thoughts? Yay, Nay?

Greg

I just finished running it on mine ... I used the Radio Smack Analogue (for the Brits) SPL Meter (33-2050) ... I found a mic calibration for this meter, but it sure didn't match mine.

When I downloaded the demo version of SpectraLAB, it came with a calibration file for that particular mic built-in. What software did you use?

There is an issue in doing this RTA work that I haven't yet resolved, I'm wondering how you felt about it:

At first I thought the mic's exact position in the vehicle wasn't important as long as it was somewhere close to the driver's seat. That's still true in terms of its height placement and left-right placement in my car. But I noticed that there were significant differences as I moved it forwards and backwards in the cabin. It seems that standing waves resulting from the differences between the front speakers and the rear speakers can make a significant difference in the frequency plot. As you move the mic forward and backward, you can see sharp peaks and valleys appear and disappear.

Other than completely crossing over all speakers so that they don't ever intrude on each other's frequency bands (impractical in my car's setup without some additional expense for different kinds of speakers), is there any way around this?

And of course, Greg, I'm sure that you've discovered that (a) it's impossible to get it perfectly flat, and (b) anything close to a perfectly flat response line doesn't even sound that good. The RTA is a "baseline" from which you tweak to taste.

I've found that I like the way Genixia's Bass/Treble controls work off of the EQ. I can use the first eight bands for all of my RTA corrections and leave bands 9 and 10 open for Genixia's code. Then I can increase the bass and treble in Hijack to give the overall response a decent amount of punch. Of course, then I have to make sure to keep the volume control below -6db to prevent clipping.

On mic placement, I usually just position it where my head would be, if I were sitting in the driver's seat, using a tripod. From the front to back sonic issues, having the mic simply in front of my face, while sitting in the driver's seat definitely alters the levels.

I don't think crossing over each speaker would prevent this ... it just has to do with fall-off after a few feet away from the source. The material in the car, echos, obstructions further cause this deviance. But, that's the fun of having audio in a mobile environment ;-)

I was using spectra lab, but the included radio shack file was for a different mic than the one on the SPL meter.

g

You mentioned about keeping the volume below -6 dB to keep it from clipping. What are you using for measuring the signal to see if it is indeed clipping or not? Are you using music software or pink noise to determine clipping?

No, I'm simply dealing with this bug.