It always makes me sad to see how some people build up an emotional relation to their car. My honest belief is that a car is nothing more than a basic commodity, that only has to adequately serve its basic tasks.
No reason to waste emotions!
In short, the basic tasks of the basic commodity “car” are:
In order to serve these purposes, we can define following requirements that apply to the basic commodity “car”:
Since my contraption will fail all three of these requirements with flying colors, I finally found one substitution requirement which I felt ready to tackle:
In the following, I will describe how I tried to address this requirement.
Below, for orientation only, a clipping of the wiring diagram. It will not indicate the peers of all connections, and it may be outdated. Please refer to wiring_diagram for a comprehensive and uptodate view.
The audio equipment and wiring is comprised within the dashed lilac rectangular outline in the bottom center of the diagram. Radio, bluetooth receiver and DC/DC converter within a light green rectangle “central console - upper” within the “audio” rectangle.
As you see, the layout is an old-fashioned approach without digital signal processing and with passive cross-overs. Being dependent on components and acoustic conditions that are far from ideal, a “digital + active” approach would very much ease compensation of all the related shortcomings, i.e. mainly irregularities in frequency response and group delay. The “analog + passive” approach here clearly comes to its limits.
The radio is a simple Blaupunkt “Essen CD31”. I got it for free, and after replacing the circuit board in the operating panel and cleaning the CD player laser's lens it seems to work fine. It's main advantages are that it has an auxiliary input, and it's implicit theft protection by having zero commercial value (actually, this was indeed my main reason not to look for a more uptodate model).
The bluetooth interface (Belkin AirCast) supports streaming of audio content e.g. from a mobile phone via the bluetooth “A2DP” profile, and in addition can act as a hands-free speakerphone. It is delivered with an attached 12V adaptor built into a cigarette lighter plug.
Below photo shows the first test. The Android phone (using “Poweramp” application) is streaming into the Bluetooth interface, that is connected to the radio's “aux” input. These devices and the amplifier are commonly powered from a 12V battery.
With this setup, even though everything was arranged so tidy and neatly, there was quite some noise due to potential differences between ground of radio vs. bluetooth device. Fluctuations of this voltage offset are interpreted as signal by the radio's amplifier. Such ground loops can be broken up by either using audio transformers in the signal path, or by feeding the external appliance via a DC/DC converter with potential free output. I chose the latter option, as it has no impact on the audio frequency response. The DC/DC converter then replaces the original 12V adaptor of the Bluetooth interface.
This amplifier (Axton A4050x) drives the bass- and midrange speakers and the tweeters. Like with most car audio amplifiers, it optionally supports high-pass filtering in order to spare the lowest frequencies for a dedicated “subwoofer” signal branch. The high-pass (12dB per octave) edge frequency is adjustable.
I had chosen this particular amplifier for its small size and low weight. The power of 4 times 40 Watts appear sufficient to me, especially for high-pass mode of operation. Watching the power consumption of the the amplifier rise while turning up the volume shows that the volume becomes unbearable at a few watts already.
This is a Pioneer TS-WX77A, a particularly flat horn speaker with 2 16cm chassis and internal 200W amplifier. It allows adjustment of gain, upper edge frequency and has a phase reversal switch.
The thin doors of older cars require especially flat speaker chassis. I have chosen the AudioSystem “R165 flat” two-way speaker set, consisting for each side of a flat 165mm chassis, a dome tweeter and a two-way crossover. I was not convinced that the 165mm speaker would perform well at higher frequencies, especially what concerns its omnidirectional behaviour. So I daringly extended the front system to three ways (not counting the subwoofer), by adding midrange dome speakers (in a first attempt) and additional three-way cheapo crossovers (SinusLive CR-345, with 700Hz and 4500Hz crossover frequencies). The 165 mm chassis is now only used as bass speaker.
The following graph (source: car&hifi) shows the frequency response of the two-way speaker set (including two-way crossovers) at two different angles. The graphs indicate that the 165mm chassis becomes highly directional at frequencies >1kHz (as is to be expected for a large cone speaker). It however also shows that this effect applies to the smaller tweeter too, at correspondingly higher frequencies.
Below frequency response and impedance of the first set of midrange speakers that I installed (source: u-Dimension). To avoid disappointment, angular behaviour is not covered here.
Meanwhile, I have serious doubts that the shown frequency response has anything to do with the actual behaviour of these speakers.
I later replaced them by “conventional” cone midrange chassis (Gladen Zero Pro 80). Frequency response including 2nd and 3rd harmonic distortion (source: hifitest.de):
Fortunately, with the given crossover frequencies of 700Hz and 4500Hz, they are operated in their “sweet spot”, both what concerns harmonic distortion and frequency response including angular behaviour.
The AudioSystem crossovers, that have very generous adjustment possibilities for the tweeter branch, remained in the dashboard and connect to the tweeters, that are mounted on top of the dashboard. Their lower frequency output, which is now set to transparent “straight through” mode, is connected to the inputs of the additional three-way crossovers mounted in the doors.
These “cascaded” crossovers further split the signal for bass chassis and midrange speaker, with crossover frequency of 700Hz and mid-range end frequency of 4,5kHz. The tweeter branch of these crossovers is left idle.
Main reason for this strange set-up is that for each side, there are only two wires available that connect dashboard and door.
In both crossovers, I have removed those components that are not used, i.e. the low frequency branch for the dashboard crossover, and the tweeter branch for the door crossover.
In each of the front doors' stacking tray, I provided a small black box with wire connections to bass and mid-range chassis'. It allows to add e.g. reisistive voltage dividers to reduce the volume or to reverse the phase of the mid-range chassis. Please note that the drawing on the photo is outdated.
Resistors in parallel to the speakers will also support the operation of the crossover, since the impedance of the speakers alone is strongly frequency dependent. In particular, the impedance rise at higher frequencies, due to the speaker coil's inductivity, may prevent a clean cut-off at a given upper edge frequency.
Currently, the chassis have no resistive dampening yet, which may leave further room for improvement.
Below photos (quoted from the “dashboard” section) recap how radio and bluetooth interface are arranged and wired in the central dashboard console.
The plastic film-can above the radio, with blue writing on it, contains the DC/DC converter that powers the bluetooth interface.
The picture shows the high-pass amplifier mounted beneath the glove compartment. The plastic cover had to be cut out a bit to fit over the amplifier (giving it some extra ventilation). At the second attempt, I immediately got it right.
The tweeters allow mounting both in a cutout or on top of a surface, with an angled base.
I have chosen to place them on top of the dashboard, avoiding the original tweeter position in the doors. The tweeter point upwards, so that the sound is reflected by the windshield. Regarding acoustics, this has several advantages: The distance between tweeter and ear is higher. Therefore, the perceived sound level does not depend so much on your actual position and is better balanced between left and right. Same applies to sound delay, which in addition becomes more aligned with the delay from mid-range speaker and bass chassis. And most positively, the sound now appears to come from the front, and from the height of your ears - and not from below and from the sides.
Main disadvantage is that stray sound directly emitted from the tweeters into your direction may interfere with the reflected sound, potentially creating irregularities in the frequency response. This is why real audio freaks dislike this approach.
For the bass chassis, I had to re-use the spacer rings of the original speakers (commercial rings were a bit too wide). Destructive, dusty, smelly job…
The new speaker mounted into the original spacer ring, with aluminium coated butyl rubber voodoo dampening around.
Front door with the bass chassis mounted, and more butyl rubber dampening mats. On the center bottom of the picture, the crossover for bass and midrange speakers is visible.
The purpose of the butyl rubber dampening mats is to deaden vibrations, by absorbing energy when they are flexed. They however do not absorb acoustic waves from the air, but just reflect them.
Before mounting the door panels, the apertures in the door must be closed with a foil acting as a humidity barrier. Instead of thin plastic foil that may cause noises, I used 3mm foam foil. Second photo shows the backsides of the door panels, loose parts fixed up with glue to avoid noises and with some pieces of foam rubber attached to them. The foam rubber is intended for dampening resonance in the door's cavity, by absorbing sound from the air. It is a compromise: Foam rubber is actually too lightweight and there is also not enough of it to be efficient, but heavier material that has better acoustic properties might too readily absorb water and cause rust issues.
The mid range speakers were mounted on wooden spacer rings, but due to their depth also required cutouts in the door panels.
For the cone speakers that I installed later, I used wedge-shaped adapter rings to slightly tilt the chassis back - and upwards.
The ugly blain in the cover below the cutout even was intentional, though it came out misshapen: The new bass chassis have a slightly higher diameter, so I had to mistreat the plastic cover with the hot air blower to free some space for them.
Re-assembled front door with bass chassis and midrange speaker - first attempt …
and the second, final version:
Experts say that the rear system of a car does not contribute much to the sound - and experience shows that passengers on the back seat will anyhow only ask to please turn down the volume of that terrible music, you darn little sonofabitch. So the rear speakers are “low budget”!
Again, I have re-used the spacer rings of the original speakers. The new ones (Sony XS-F 1037) are coaxial 3-way 10cm speakers, that were fitted in via wooden adapter rings. With their small diameter, they also had sufficiently low depth to fit into the original mounting place in the hatch door.
Dampening material in the hatch door and on its cover.
Rear speakers mounted on the hatch door. Additional foam rubber pieces are placed within the door to avoid cavity resonances. The foam rubber ring around the speaker is intended to seal it against the cover, to prevent “acoustic short-cuts” from the speaker's front to its backside within openings in the door.
The active subwoofer is mounted on top of the rear traction battery box cover. It is held in place by aluminimum rails and steel mounting brackets. Power is fed directly from the rear distribution box, which is connected to the 12V battery via 10mm² gauge wire.
Since the manufacturer had applied dampening mats only sparingly to the car's body, I added some more so as to reduce drumming noises of the sheet metal. Now, if you for instance crash the car into a tree (or slam a door), it will give a cultivated “thud” rather than a cheap “bonk”. And - as it was the primary target of the exercise - audio perception should benefit from reduced rolling noises.
I have already spent some time tuning the sound via the numerous setting possibilities of the dashboard crossovers (tweeter frequency response and volume), high-pass and subwoofer amplifiers (edge frequency and gain), and by trying different resistor combinations for the mid-range speakers' voltage dividers. For the mid-range speakers' polarity, I decided (rather a guess) not to reverse it.
After a while, I was inclined to believe that the radio sounds somewhat ok - mission accomplished!
Later however I had to realize that the sound was actually not so good, particulary in relation to the effort I had invested. The initial, positive impression probably came from always listening to the same few “test tracks”, and from underestimating how quickly and how much you can get used to a speaker system's linear distortions. It also seems virtually impossible to derive the right “corrective action” to a perceived flaw in the frequency response without the help of an objective measurement.
Another surprise was to see (or hear), by how much the sound perception changes under the influence of rolling noises. It seems that especially when listening at low volume, the noise floor will first compete with the “dips” in the frequency response, masking out the corresponding parts of the spectrum and thus making the “peaks” much more prominent.
Rolling noises, by the way, have their highest portions at low frequencies, asking for a deliberate enhancement of the bass and sub-bass bands to keep them audible at low volume.
After quite some fruitless efforts to get the system to sound well, I decided to objectively record the frequency response.
Please note: The following measurements were all done before I replaced the mid-range speakers. With the second version that I used, I was able to get a proper sound out of the system, by just adapting the levels of the different chassis. I so far have not repeated the measurements with the new setup.
This eigthies' electret condensor microphone is placed beneath the passenger seat's headrest.
A sinusodial test tone generated by an android tone generator is fed to the radio via bluetooth streaming. The DJ mixer acts as microphone preamplifier. The amplified signal is “quantified” by a volt meter.
For each of a row of frequencies, a fully automated test support chimp sets the tone generator to a certain freqency, restarts it in case it has another “buffer overflow” and enters the AC voltage reading into a spreadsheet.
The spreadsheet transforms the readings into a logarithmic scale and generates the chart below.
It shows the results from the Astra's car stereo, and two reference curves. Their “polygonal” appearance is due to the low number of samples taken. The car stereo measurement has not been corrected by subtracting a reference curve. So, a “perfect” speaker system should not yield a flat curve, but rather a curve resembling reference 1.
Generally, it was a bit disappointing how much irregularities an acoustic frequency response seems to have, also compared with curves one usually gets to see. Obviously, a higher number of tests points would have been useful.
pink curve (reference 1): frequency response of the microphone, as from it's datasheet.
blue curve (reference 2): frequency response of my “living room stereo” speakers (Marantz LD200), old, but still sounding very well. Note that the bass chassis once have been replaced (by a professionist) and the bass reflex system has apparently not been matched to the new speakers' properties.
For this measurement, the microphone was placed approx. 1m in front of the speaker, in a moderately dampened room. Apart from the bass resonance peak and dip, quite some irregularities are visible.
dark red curve: frequency response of the Astra car stereo (somewhen during the “optimization” process). In the car and at higher distance from the speakers, the sharp irregularities are much more prominent that in the “living room” case. This is is probably due to reflections within the car.
The most ugly peaks are at 50, 100 and 400Hz. The 400Hz peak might be due to a cavity resonance from inside the front doors. The 50Hz and 100Hz peaks are apparently caused by cavity resonances of the car's interior itself. The 50Hz peak was additionally enhanced by a rattling noise provoked by that resonance. Both peaks are probably not easy to tackle, except by electrically suppressing these frequencies.
The range between 700Hz and 4kHz is just too low by some 5..6dB. This can be easily corrected by turning up the volume (i.e. modifying the resistive attenuator) of the mid- range speakers. Also raising the high frequency end is only a matter of some settings of the crossovers.
Apart from that - it seems that here I am at a point where I can either pretend that I am happy with the sound, or I have to bear the effort of integrating a digital sound processor. Starting with this, probably also the passive cross-overs would have to give way to a few more amplifiers…
Please note: I had drawn above conclusion before I replaced the crappy midrange speakers. With the second set of speakers, the sound is now much better, without any large investments like sound processor or active crossovers.
Before the speaker swap, I even had tried to integrate a sound processor. I returned it shortly after, since the sound got “artificial” rather than “good”.