mpeteritas

Gents,

We have had several discussions about the preamp outputs on the TSX, with many (including myself) questioning how flat the signal produced is, with treble=0, bass=0.

I have run some tests and found that the HU is overall pretty flat...especially at higher volumes...Results were dissappointing at lower volumes.

My car is a 2004 non-Navi TSX.

I burned an Audio CD with 16bit 44.1kHz White noise sample and capture the output from the preamp outs at various volumes...10, 20, 30, 40. I have spectrum analysis diagrams posted for each volume...

original clip (direct analysis):
http://rockenbach.homeip.net:8080/sh...actualclip.JPG

volume set to 10:
http://rockenbach.homeip.net:8080/sh...X_HU_vol10.JPG

volume set to 20:
http://rockenbach.homeip.net:8080/sh...X_HU_vol20.JPG

volume set to 30:
http://rockenbach.homeip.net:8080/sh...X_HU_vol30.JPG

volume set to 40:
http://rockenbach.homeip.net:8080/sh...X_HU_vol40.JPG

On all the capture clips, you can see the HU rolls off the high end at about 20kHz. That's no problem, as most of us can't hear very much above that anyway. What's a little dissappointing, as you can see, is the highs are noticably and audibly peaked when the volume is around 10 (probably gets worse between 0-10), and there is a nice little dip in the mid range. The curve starts to flatten out around 20 (50% volume), but still has a slight peak on the high end. Above 20, its really pretty flat.

I would suspect its better than many average OEM HUs. I was hoping to see flatter response in the 0-20 range, as this is where most listening is done, unless you don't like your hearing too much

So, there's my analysis.

If anyone wants to do their own analysis, the white noise clip is available at:
http://rockenbach.homeip.net:8080/sh...e_st_20sec.wav

I would be very interested in seeing a similar analysis of a NAVI Head Unit. Any takers?

-mark

elduderino

Any way to show db ranges? It's very hard to see the actual amount of variation...

And remember that CD players have a steep LP filter just below 22.5k...

elduderino

And in looking at that curve it doesn't look like the highs are peaked, it looks like the mids are dipped. The absolute HF level is the same, isn't it?

Since Pink Noise is random, what kind of averaging were you using on the audio analyzer?

jlukja

"connection refused" when I click the jpg links.

elduderino

I will measure my car again today... but I suspect you are seeing an "auto-loudness" feature.

http://www.webervst.com/fm.htm

And I run my system above 30 frequently - and once I've installed a new amp, I control how loud that actually is with my gain settings, right?

mpeteritas

The connection I have these on maps http from port 80 to port 8080, which is a problem if you are using a proxy server on your side (like from work). It should work fine from most residential networks. sorry!

-mark

mpeteritas

Cool, I'll be anxious to see your results. The procedure I used was to set the input level on the capture device to almost full vol, when HU is at vol 40. I kept the capture device's input level there for all the tests.

As far as spectrum analysis algorithm used...I am using the spectrum analyzer in Sonic Foundary Sound Forge, set to logorithmic view. The algorithm they use is Blackman-Harris alg.

-mark

mpeteritas

I have captured the spectrum analysis in a normalized view, which shows the dB range much clearer. Will post soon. I will also post the actual captured wave files in case you want to run them through a different spectrum analyzer.

-mark

jlukja

I couldn't imagine listening at above 30.

jlukja, who's volume control doesn't venture past 20 much at all.

elduderino

Remember, 30 isn't absolute - it's relative.

But with the OEM amp I know what you mean.

mpeteritas

Ok, the normalized spectrum analyses are at:

vol10:
http://rockenbach.homeip.net:8080/sh...vol10_norm.JPG

vol20:
http://rockenbach.homeip.net:8080/sh...vol20_norm.JPG

vol30:
http://rockenbach.homeip.net:8080/sh...vol30_norm.JPG

vol40:
http://rockenbach.homeip.net:8080/sh...vol40_norm.JPG

Also, the captured wav clips are at:
http://rockenbach.homeip.net:8080/sh...X_HU_vol10.wav
http://rockenbach.homeip.net:8080/sh...X_HU_vol20.wav
http://rockenbach.homeip.net:8080/sh...X_HU_vol30.wav
http://rockenbach.homeip.net:8080/sh...X_HU_vol40.wav

and again the original clip is:
http://rockenbach.homeip.net:8080/sh...e_st_20sec.wav

-mark

jlukja

Good work Mark,

Now that I'm home I can see the graphs. Interesting. I thought loudness was the artificial boosting of bass at low volume. The graphs would seem to indicate that the bass isn't boosted at low volumes relative to the high freq.

Makes me wonder if using less efficient speakers would help. The logic being that a comfortable volume would be achieved at a higher pre-amp output (and thus flatter frequncy response curve). Well, at least for those of us who are less interested in volume and more interested in SQ.

mpeteritas

Yeah, if we can draw any conclusion its that we should target our listening volume to be in the 20-40 range (assuming you have an amp and can adjust the gain). I am hoping the dude can get similar results for a NAVI HU. I will be interested to confirm whether or not the results were the same. I think he got flatter results on a NAVI HU with previous tests.

-mark

jiggaman

i set my aftermarket amp settings so that basically usually am listeing in the 25-35 range...

I'm Batman

Were you sitting in the car while running the tests?

mpeteritas

Yes, but all amps and speakers were disconnected (didn't want to go deaf), so I was simply monitoring the level on the recording device...

-mark

elduderino

I didn't have time on Saturday... I might on Monday - I'm cracking the dash open to install an Icelink anyway.

When I had the JL 300/4 under the seat, I could reach under there are grab an RCA easily.

Now that we have the DLS CA41 under there, the RCAs are on the end, under the seat rail, so it's impossible to grab them there without pulling the seat. SO I get to pull the dash for the umpteenth time, so I might as well put in the Icelink when I do : )

elduderino

At low levels (up to 20) the Nav HU tests out the same. I hadn't had it that low when I tested it before because my amp gain is low and because I was mostly verifying that there wasn't a high-pass "subsonic" kicking in at higher volumes.

But for kicks, I connected the RTA to a Nakamichi CD-400 and turned the "Loudness" control on and off. WHen Loudness was ON, the EXACT same contour appeared in the midrange as mark has posted.

So it seems pretty obvious that this midrange dip is an auto-loudness feature of the preamp IC that was used.

vidgamer

Most loudness functions I've seen boost both the lows and the highs. If you reduced the mids, I would consider that basically the same thing. :-) It's all relative....
I dunno, it seems like relative to the highs, they're not boosted as much, but relative to the mids, they are definitely boosted. Doesn't surprise me. I've run into other loudness circuits that were annoyingly harsh (because they boosted the highs along with the bass).

Makes sense, until you run out of headroom to actually run those speakers.

mpeteritas

Cool. Glad you were able to verify my findings and the explanation of the midrange dip being due to an auto loudness feature in the preamp circuit seems reasonable.

Now we know where to target our listening volumes, eh? 20+ seems to be the ticket (for those who have amp gain to control the perceived volume).

-mark

elduderino

I'd still say this about the auto-loudness and sound quality:

1) It's generally accepted that gentle variations in freq response over a few octaves are far less objectionable than spikes or holes in the response.

2) The spikes and holes created by the speakers interacting in the car are far, far worse than anything the auto-loudness control creates.

So when you actually ACHIEVE this curve as an ACOUSTIC response instead of a SIGNAL response, I would start worrying about volume levels. Until your speakers are close to this curve in actual miked-and-measured playback, I will continue to believe that your listening issues are speaker related, not signal contour related. In other words, you are absolutely right about the loudness contour feature, but I believe that until you have a VERY advanced amp-and-speaker playback system, you're close to the mountain-from-a-molehill on this being the cause of anyobjectionable sound.

I recently delivered a BMW X5 using the OEM audio system and summing the various channels together. We got this thing to sound INCREDIBLE with Dynaudio speakers, Genesis amps, and a DLS sub.

But our signal response wasn't anywhere NEAR as smooth as this one. The various bands were fighting each other in their stopbands due to phase effects of the filters. Still sounded great, though.

Or as MJ said, "It's not the shoes".