MVCalypso

I got curious this afternoon and spent some time in the RX-8 in the garage... I'm replaceing a BMW 328 with the RX-8 - the BMW has a highly custom stereo system I designed for the car. I'm not really unhappy with the RX-8, it's just not anywhere as nice as what I have gotten used to - so, I thought I'd take some measurements of the BMW frequency response and the RX-8 to compare them - mostly for my own curiosity.

So I measured the frequency response of the Bose system and plotted the results - figured I'd share the data in case anyone is interested...

Here is the measurement approach info:

Measurements taken 8-16-2003
2004 RX-8
GT package Bose stereo w/6 disc changer

Reference tones - Recording Studio Test CD
Tejera Microsystems Engineering

Radio Shack Digital SPL meter
Placed at approximate head level in drivers seat (see pic)
Levels measured relative to 70dB at 1kHz

Settings: Audiopilot off; tone controls per plots; windows & sunroof closed

Measurement tech (me ) sitting in passenger seat

and these are my first impressions:

The front woofer amps are way too hot !
31.5 Hz to 200 Hz is 23dB (!) to 4dB above reference level.
Remember that "flat" is generally accepted as +- 3dB…
also remember that the graphs are labeled at 1/2
measured SPL levels

System has significant crossover holes?
Dips at 315Hz and 630 Hz
Interesting that they are one octave apart...

I'll try to post the pics and the response plots as soon as I figure out how to get them into a format I can attach to a post...

MVCalypso

Here is the "ear"...

MVCalypso

THe composite curve plots...

MVCalypso

the bass control curves...

MVCalypso

and the treble control curves...

ProtoConVert

Uhhhh... thats awesome that you did that.

But could you tell me what all of this means? I'm always lost when reading audio posts

MVCalypso

Wow - that's one of the hardest, simple questions I've seen in a while... I'll take a short crack at it, but a full answer would be a book.

In the process between the recording of a sound (let's consider music) and the reproduction of that sound later in time, a primary goal is to not introduce changes. This allows the listener to hear what was originally played. That ain't easy...

One of the ways that audio systems introduce changes is that they don't always amplify all tones the same. In theory if you put a known sound into an audio amplifer system, you should get the same sound out - only louder - amps and speakers, being imperfect, make this hard to do. Sounds systems tend to amplify differnet pitches different amounts - generally this is considered bad.

The plots I posted are graphs of the output of the Bose system (Cd in, speakers out) as a function of pitch - where all inputs started as the same volume (from the CD). in an ideal system, the output would be a dead flat, horizontial line - which never happens. This system is a long way away from that ideal.

When you use tone controls, you are intentionally introducing change - more/less treble and/or more/less bass. This changes the graphs (as can be seen).

Perception wise frequency response graphs are one way to get a quantified handle on "why does it sound flat or muddy?". Given this data, one can begin to consider how to modify the system to get a desired tonal balance.

OK, I'm going to stop there - if you're interested in the topic - there are any number of Hi-fi stores willing to teach you more as they part you from your $

eccles

Given that someone has already determined that the tweeters fire almost directly across the car at each other, rather than at the audience, it might be interesting to repeat the process with the meter in the "line of fire" to give some indication of what could be achieved if somebody came up with a way to reposition them.

Wing

Wow those graphs are HORRIBLE! Man my home system is near linear coming up at 20Hz at 75db straight across to 20Khz with VERY few dips.

That thing looks like an earthquake graph!

seems like the mids is more of an issue than the tweeter or the sub.

MrWigggles

MV,

Great work.

I have some suggestions that might make the response more level.

1. Use the meter with "c weighting" enabled. (It looks like you are already doing this)

2. Try to point the meter straightforward instead of up at a 45 degree angle. This might help level off the high frequencies a little.

3. Make sure the meter is in slow response mode and I would read the value right from the meter (IF the meter has a "peak-hold" feature, that might give some inaccurate readings on the low end.)

4. Is the CD that you are using doing pink noise? or are the test signals sinewaves? Pink noise (sounds like static) is much better for this test.

5. Try and raise the volume to a more suitable level than 70 dB. I would say 90 dB or higher would be a much better representation.

6. See if the audio pilot changes the response any when engaged or dissengaged.

7. Was the car idling during the test or was everything completely off?

8. It looks like your +/-3 dB window (which should be 6 dB total spread) is only +/-1.5 dB. You have a note saying: "measured dB values are 2x plotted values" Is that where the difference is? Just checking.

Excellent work,

-Mr. Wigggles

Ps. the dips in the frequency response are probably from "nulls" created by the geometry of the vehicle's interior. Try doing the test with all the windows and sunroof open to see if the dips are still there. Also it might help if you moved the meter in a circular pattern when you did the measurement. Just a thought.

MrWigggles

As far as the increase in bass is concerned, the response curve should have a some peaking in the lower octaves. This is because road noise is also quite bass heavy. By your measurements at 70 dB, the Bose system does seem a little too heavy, but during my test drive I didn't think it was nearly that bad. (I had the bass at +3 during my test drive and the volume playing pretty loud.)

-Mr. Wigggles

MVCalypso

Some additional notes re your comments above:

1) The meter was on C weighting for all tests

2) I wanted the meter Horiz, but the LCD angle is such that it made it very hard to read the meter from the passenger seat when I did that (see note below re head position)- so I compromised.

3) Meter was in slow response mode - started doing the tests letting the meter take a 15 sec average (each test track is 30 sec)- but the meter has to be put back in that mode each time - given the 30+ tomes and multiple runs, I settled for watching the level and "eye averaging" - I figure the meter reading error is at most 1 dB.

4) test signals were sine waves - the test CD also has pink noise tracks but I haven't messed with them yet - pink noise would let me find nulls I guess, how does one use pink noise to measure frequency response? (I suppose if I had some FFT equipment..)

5) Yea, I got surprised - I set the system to 75dB at 1 kHz per CD instructions - only to find out (afterward of course) that that there is a slight dip about 1k Hz... In any case, the volume was not changed during the tests so relative info should be ok.

6) I could not see any effect from the audio pilot.... and I have the same experience when driving - my ears can't tell if it is on or off - weird - should the effect be easily noticable?

7) car was off, just sitting in garage (wondering why I was not out driving).

8) yes the real 3dB window is +- 1.5 db on the chart. I had to plot the values at a compressed scale to fit the "paper" I had. The semi-log chart came on the CD as a PDF file, I was going to hand plot to change the preprinted scale, but when I called the local drafting supply stores, I was told that no one makes log or semi-log graph paper anymore!! Apparently this is a casulty of the computer age - grumph - so I settled for scaling the data to fit the scale on the page. Real measured values are 2X what's on the chart - hence the bass is 23dB high from reference level!

9) re windows - yea I want to try that, but I got tired sitting in the car. Note that intereference patterns were very much present - I had to hold my head in the same spot on the head rest during measurements - as moving would change the readings at higher freqs.

I'll prob try some more combos next week - but for now, I'm taking the RX-8 up to the marina to play boat person for the rest of the day (see avatar). :D

Dave

Wingnut

Awesome work. I'm glad to see a little science in the audio impressions. One thing I noticed, however, is the setup of the "ear." I suspect the issues with low frequency hot zones are the result of the mike sitting on a trash can (that's what it looks like to me) resonating. You need something solid there for the best results. I'm not an audio expert but that "stand" is a big variable thrown in.

I look forward to more results.

WN

MVCalypso

Hum, you make an interesting point- I had not tought about resonance of the "stand".... Yes it was an inverted trash can - the first thing I found in the garage about the right height. OK, when I get a chance, I'll run a sample sequence with some other mounting arrangement (once I figure out what that might be).

Dave

jonalan

Will the meter fit inside the opening in the back of the seat (headrest area)? Maybe just stuff it in there, since that is approximately ear level. Also, where did you get the test CD? Is it available online? I'd like to do the same test with my cloth seats to see the difference between the cloth vs. leather. I would assume that the cloth would absorb more of the sound reflections.

The dips at 315Hz & 630Hz don't surprise me. This is an issue with most Bose systems. That's what makes them sound so thin.

MrWigggles

The wavelength of 315 Hz is about 3 feet. It is easy to get nulls in frequency response at those frequencies. Those type of nulls are common in many audio systems and they aren't that easy to get rid of, nor do you want to pump up the equalizer by 10 dB to get rid of them.

MV,

4. I have 31 short WAV files with 1/3rd octave pink noise instead of just pure sinewaves. It is about a 33 Megabyte file zipped. Private message me if you know a way I can send them to you.

5. I also think the relatively low volume level has triggered some sort of loudness function in the headunit or in the Bose system. I think 90 dB or so will work better. Remember car stereo systems have to play louder than home systems due to the increase in background noise.

Your methology is very good and I think using pink noise at a higher output level will improve your results.

Oh, and nice boat. Or, should I say "ship"?

-Mr. Wigggles

MVCalypso

Mr Wiggles -

I've sent PM re email contact info.

Re the loudness function theory - maybe, but I don't remember standard loudness curves being anywhere near +23dB for boost. I was real suprised at the magnitude of the lower end - but something is not adding up right yet - as it does not sound that boosted to me when in the car... I'm thinking that the audiopilot is acting up - but I did turn it off - could it be that it never really goes "off"?

And now I have to ask - just what does one call 1/3 octave pink noise files? Since white noise is equal across all frequencies, and pick noise is equal across the audio range, is pink noise that is limited to 1/3 octave called red noise? :p

Dave

astrlsrfr

Hey MV - thanks for sharing the fruits of your efforts.

Auto interiors are notorious for their resonances, boominess, peakyness, etc. But each is different and yours are the first hard numbers I've seen on the 8.

I think pink noise is just an alternative to a pure sine wave so that you can still measue individual freqs, but don't get constructive and destructive interference as prominently. I think someone already referred to "NULLs" - that would be an example of destructive interference.

You can also use white noise. White noise is meansured accross the band. The quickest way to measure it is with a spectrum analyzer. I have one built into this old graphic equalizer in my home system. It lets you hook up a mic & adjust the EQ til the response is flat. Barring something like that - pink noise maybe the best option.

jtdwab

First MV, Great Work!

I'm not really an audio file but I do like to tinker from time to time (drives my wife nuts). Anyway related to loudness issues I was playing with the sound today trying to work out some of the issues and found that turning the volume up a few more levels helped to even the sound. I don't know why, but it did. I used to listen to the sound at around 3-5 setting and fight the controls for it to sound good. I bumped the sound up to around 10 and found it was not only louder but fuller.

I tend not to listen to things loud becuase it bothers me (i'm only thirty). I been this way my whole life, I guess my ears are to sensitive or something. In short I would recomend boosting the volume and see if it works better for you.

Thanks again for the work your doing a great job.

MVCalypso

Hi - I saw a post asking where to get the test CD - you can get one from http://www.TMEnet.com

Dave

ELK

I've played only briefly with the audiopilot feature. It does increase bass tremendously. It may somehow have been on when you were measuring.

Bose systems are notriously inaccurate. They are tuned to quickly impress the neophyte listener. At least in Bose home systems, they typically have very hot highs and a large bump at about 120hz that provides the illusion of powerful bass, but is boomy. Unfortunately, most people think Bose equals good audio. Oh well . . .

However, the relative bass levels you measured are astonishingly high. I suspect that it is a function of confines of the car (listening to audio in a car is like listening in a partially carpeted closet full of shelves).

Thanks for posting your efforts. Interesting stuff. I have a real time analyzer, calibrated mics, etc. for use in my home system but it'd be a pretty good trick to get it set up inside the car.

rotarygod

If there is anyone in Houston who is curious about the response of their car I will arrange to have an RTA (real time analyzer) measue it with pink noise. That way we can see everything at the same time over 30 different points. I was a custon stereo installer for 8 years and competed nationally on the IASCA and USAC circuits and have won several IASCA and USAC world finals in a few different power classes. I know how this stuff works.

A perfectly flat line response sounds dead and like crap! The key isn't so much that it doesn't slope downwards as frequency rises but rather you don't want any big valleys or peaks in the response. Completely flat works to a point. I had one competition car that was completely flat between 500hz and 5000hz. Below 500 the rise was 3-4 db per octave while above 5000 there was a decrease of 3-4 db per octave up to 20000hz. That car sounded REAL. However the same response curve in a different car may not work at all due to interior shapes and differences. I don't believe in the bs about not being able to sound real yet it sounds accurate. What is accurate? I would think that accurate is closest to the real thing. Accurate is actually closest to true flat and this sounds dull. Our hearing is much more sensitive to high frequencies than lows so we boost the lows. If you were to hear a system that was a flat line you would hear overbearing highs, muddy midrange, and a severe lack of lows. The general slope downward of the Bose system is not bad. What is bad is how erratic it is. This could also be due to the method of testing. Remember that there are no 2 places in the car that sound alike. The mic is in one place. You have 2 ears though and they are several inches apart pointing opposite directions. Our ears will somewhat average what we hear and send it to the brain. If there is a severe null at one ear it becomes very annoying. Also our ears are much more sensitive to frequency spikes above the rest of the range yet we are almost impervious to dips. Strange how that works.

MVCalypso

If you do get an RX-8 tested with a spectrum analyzer, It'd be great if you could post the results. I realize the limitations of the tests I did - they were limited by the equipment I had at hand.

I'd find it really interesting to have more info from someone that has done this at a competition level.

Dave

MrWigggles

I should already have my car but I will gladly take your offer when I get it. It is supposed to be hear in a 1-2 weeks.

When I do RTA readings I set the system to the slowest response setting and I move the mic around slowly in a figure eight pattern to help average out any nulls in the frequency response.

Like you said, in a car completely flat is usually not a good idea due to the high level of low frequency road noise. 3dB per octave below 500 sounds about right. That would end up being +12dB at 31Hz.

I will PM you when I get my car.

Thanks for the offer,

-Mr. Wigggles

rotarygod

No problem. We'll get it done.

I find that the best way to measure is to not move the mic at all and then get a reading. Do this in a couple of different locations (like where your ears are). I generally use the fast setting just to get an idea of what it is like. The slow is good to see a more accurate shape and this is how I would do it to print it up. Unfortunately I don't have any way of printing it up anymore. We can snap a picture of the RTA though and post it. I used to use software such as LEAP and LMS and print up everything. Unfortuately all I have access to anymoe is just the common Audiocontrol RTA that every other shop has. Since I don't do this for a living anymore it would only take me about a days notice to arrange it at Expressive Audio in Humble (where I used to work).

Just to give a little bit of info as to how Bose designs their systems. Bose takes measurements in 8 different locations within the vehicle. They have a certain frequency response curve that they aim for. Too bad it isn't the same one that I like and that their speakers are too crappy to do it anyways! The home speakers actually have made in Mexico stickers on them attached to the magnets! Anyways back on track. They shoot for a certain curve but at 8 different locations it takes some serious averaging for the best overall compromise. The biggest problem is that you can physically average 8 numbers but this doesn't mean that it sounds good. Most of the time it creates more problems than it fixes. I always tuned a car to sound the best from the drivers seat (where I sit!) and still be quite nice from the passenger seat as well. Any other seats are irrelevant and most people aren't discriminating listeners anyways. Bose wants a good sound from every seat in the car at once. We aren't in a house or auditorium so good luck. Each Bose speaker is amped at the speaker. Each speaker is also in its own tuned enclosure. Built into the amp is the equalization circuit for that speaker. This is the reason why you can't just swap Bose amps from other cars or locations out and have it sound right. This is also why people can't add EQ's to a Bose system and have them sound right. You would have to add several new EQ's. Anyways sorry to ramble. Just thought the info was relevant.