In which the author attempts to communicate at a distance

I attempted to purchase an LED strip from a Chinese factory at considerably less than I can obtain through North American retail channels (like 25%).

This is the message thread. It’s not clear how to specify the exact item I want, so we start out:

I want to purchase a single 12 meter strip, 110v, warm white.

1. How do I place the order?
2. Does it come with the power plug or do I need to purchase it separately?

Thank you.

Response:

Dear friend
If you need 12 meters, the price is 49.5 dollars. I offer you the link, then, you can make order. I will change the price. The strip comes with the power plug.
GOOD DAY
Ming lu

I see Ming also struggles with limitations of Aliexpress’ sales platform, but has a workaround. Their product even comes with the power plug. Me:

I would like to purchase the 12 meter, 110v, warm white at $49.50 – please send the link.

Thank you.

Ming:

Dear friend
http://www.aliexpress.com/item/5050-LED-strip-110V-220V-230V-240V-IP65-Waterproof-flexible-SMD-led-strip-60leds-M-300leds/877316526.html
You can make order by this link.Then, leave me a message that you need 110V,warm white
GOOD DAY
Ming lu

I like being called ‘dear friend’. ‘GOOD DAY’ is a touch loud, but perhaps that’s just earnest sincerity. I leave several notes saying please 110v, warm white, etc. Our dance has more missteps:

Ming Lu:

Dear friend
Sir, you have payed and I could not change the price. You have to make one more order, you do not have to pay immediately,when I change the price, you can pay
GOOD DAY
Ming lu

No worries. When such good friends are thinking alike and working together, we can achieve anything. Through several additional steps we finally get payment worked out.

I am excited when I received the product. It wasn’t boxed as pictured, but it arrived intact. On opening the thin plastic wrapper it was sent in, I saw it has a Europlug, and this could only mean they screwed up and sent a part for the wrong voltage.

Of course the first thing I did was to map out the schematic of the strip, which can easily be seen through the transparent plastic insulation. The strip is a series of modules, each of which has 30 LEDs. The LEDs are grouped into 15 units, each unit being of a resistor in series with two LEDs that are themselves in parallel. Electronically there is no point in having 15 separate resistors in the chain, but I assume this is done to distribute heat evenly along the strip.

There is a single bridge rectifier at the plug end converting AC to DC, and two unshielded strands carry the rectified mains voltage down the length of the strip. The modules are soldered across these two. In the strip I have, however, pairs of these modules are in series – as one would do to adapt a 110v part to 220v.

I replace the europlug with a US plug, and the thing lights up, but as expected I can see it is too dim.

Me:

Hi, I received the order but you sent 220v, not 110v. I specified “12 meters, 110v, warm white”, but this has a 220v plug, and the LED strips are wired as pairs in series.

I will accept a partial refund to avoid shipping this part back if you will sell me the correct 110v version.

Thank you.

original

(You can see that the strip is attached to mains via what can only be artistic sculpting of hot glue.)

Carson Wong:

Dear friend
The led strip plug is do not distinguish between 110v or 220v. only led strip is distinguish between 110v or 220v .I send the strip is 110v really.
good day
Carson wong

Ming Lu has become Carson Wong, and “GOOD DAY” has become “good day”. A new friend.

Me:

So you are saying that you use a plug used worldwide for 220v on strips you sell for 110v? This is surprising.

I doubt this is a 110v strip because I have taken it apart. Your strips are made with modules of 30 LEDs each. The strip you sent has each pair wired in series. This is correct for a 220v strip, but would be in parallel for a 110v product. I did connect it to 110v, and it lights up, but each LED at reduced luminosity closer to a 3528 LED.

Thank you for taking the time to look into it.

Carson Wong:

Dear friend
I send you the led strip is 110V really. and the plug is not distinguish between 110v or 220v .LED strip is no 110-220v . LED strip is only 110v or 220v .LED strip 110v sales have been very good in my store. I don’t know how to explain for you to believe this is not problem.I hope you can understand.Thanks.
good day
Carson wong

So, maybe this guy really has no idea what they are selling. It occurs to me that maybe they have never tested their product with 110v because they don’t have it available. I take the high road:

Attached is a photo using a clamp meter to measure the current of your strip. I replaced the 220v plug with a 110v plug. The measured voltage is 120.3 volts and the current is .44 amps. The watts can’t be more than 120.3 * .44 = 53 watts.

It is a 12 meter strip, and your stated specifications are 14.8 watts/meter. It should consume around 12 * 14.8 = 177.6 watts. It is not even close.

I measured very a similar strip using smaller 3528 LEDs, same length and number of LEDs, and even it glows brighter and consumes .56 amps.

You need to either connect the modules in parallel for 110v, or stop claiming it has the 220v power specifications.

I hope this is helpful – otherwise you have a fine product.

original2

My language around watts and amps is careful to avoid derailing on a technical matter (one doesn’t just multiply volts and amps to get watts with AC, instead one gets an upper bound). “Fine product” is, of course, being charitable. Their product has LEDs mounted crooked, sloppy joints, incorrect specs, and dubious customer service.

Carson Wong:

Thank you for your professional opinion, do you think how to solve this matter

It occurs to me that I may have overestimated their respect for my opinion.

Nevertheless, maybe there is a language or cultural barrier, and they would still like to do the right thing. So:

1. Ideally, customers wanting a 110v part should get a 110v plug, with the LED modules wired in parallel. It would be much brighter and satisfactory. If you made such a strip, I would be happy to buy it.

2. If #1 is not possible, change your product description to say “Comes with the pictured europlug. Will light up if wired to 110v at reduced power and luminosity.”

3. You should give me a partial refund (because it is not the part that you specified).

I am waiting until you respond to write product feedback on aliexpress.

Thanks!

Ming Lu:

Dear friend
I am willing to refund part of the funds to you, as a solution to the matter.
good day
Ming lu

Me:

I will appreciate the refund. Please do it.

Will you also change your product description?

Ming Lu:

Dear friend
Thanks for your advice. I will sincerely consider it. The describe of our product is all correct and they are hot sale always.The comment from the customers are also very excellent.You can ask for the money back and I will return the money back to you.Thanks.
good day
Ming lu

There it stands. I’m still attempting to get the money back through aliexpress, although it doesn’t hurt much because the product was so cheap. I’m thinking of opening up and resoldering the point where each pair of modules is in series – now where is my hot glue gun?

Cheap fan death-rattle

I buy cheap stuff. Nothing bugs me so much as knowing my money is going down the drain of a middleman’s margin. Two of my best weapons to keep costs down are Aliexpress, for buying wholesale from China, and slickdeals, where readers collaboratively point out how to exploit coupons and price mistakes.

Years ago I was needing a laptop and this Fujitsu LH531 popped up on the latter. It’s out of date now, but at the time this dual i5 was a great deal for $399.

After upgrading to an SSD it has seen heavy use. I might hate this laptop if I paid more money for it. It has a cheap plastic shell and is not especially thin, or light, of elegant. In a word, adequate. However, mine has a special feature: it emits the warm glow of knowing I didn’t pay much.

It always bothered me that its fan is audible. It’s not so bad, but I’m oversensitive. Fan noise can be okay when kept at a consistent low level, but there are only two BIOS settings for this fan, which amount to “be loud all the time” or “be quiet, but constantly cycle on and off”. Third party software and manipulation of the Windows power policies didn’t help. There was also a noticeable rattle when the fan turns off.

As I said, I’m cheap. Because Moore’s law is still in effect, every month I avoid upgrading my laptop means I will be getting that much better a deal on the next one I buy. So I’ve just been putting up with the fan. After all, the only time it really bugs me is when watching a video in a quiet room.

In the back of my head every time I heard this rattle I would think… how could I fix this? I could put my own fan controller in. I could see if replacing the fan would help. I could put in an RC circuit to force it to spin up and down slowly. But this was always while I was watching a video, typically at night when I didn’t feel like starting a new project. Plus laptops can be finicky, and I didn’t feel like destroying one.

But finally I got over it and opened the damn thing up. A few prods with my fingertip showed the problems:

Bad mechanical design. There are only two screws holding the fan in place, and no mechanical buffer between the fan and the heat sink it abuts. So I made an electrical tape bridge across the fan and the heat sink, and pieces of foam tape to wedge the fan in the unconstrained axis perpendicular to the two screws that hold it in place.

IMG_20131011_214309

The rattle is gone. I gave up a few style points by not using duct tape or rubber bands. Omnia cum pretio.

Cheap-ass sub shakedown

This is a Dayton SUB-120, a dirt-cheap, value for your money subwoofer.

It’s not a product you buy to show off, but to save $$$. The scratches on mine witness a lot of use. See how I have a rolled-up sock sticking out of the port in back? That was recommended online by other users, and how I’ve used it for years. I’ve always been curious how the sock changed things, and decided to shake one down with proper measurement.

I like having a bunch of subwoofers to minimize room modes and EQ changes when walking around. I have three of these, from once upon a time when there was a good deal. (I actually bought four, and gave one to a friend.) All have performed similarly to my ear.

Inside

It’s a simple device, a reinforced particle board box with a plate amp on the back, a big driver, and two unflared cardboard port tubes. The tubes extend most of the way through the body, and the mass of air they hold makes the box a Helmholtz resonator: an oscillator tuned to exaggerate low frequencies. Blocking one or both ports changes the frequency response.

Below you can see it opened up. The back plate amp is very heavy (with a substantial heat sink) and securely mounted with fourteen screws. The electronics are well secured and padded. The driver is generic and unremarkable. The walls are adequately braced. There is some fiber glued to the walls inside, which is standard for dropping speaker frequency response, presumably by reducing the speed of sound, making the cabinet effectively larger.

The amp itself has uncalibrated GAIN (volume) and FREQ (crossover frequency) knobs. The terminals are super-cheap spring loaded.

The high-level outputs are connected to the inputs through a capacitor. This is a cheap way to get a high pass filter, but has nothing to recommend it electronically, and the filter frequency depends on the output impedance. Note that whatever the FREQ knob does isn’t setting a proper crossover point, because the pass-through isn’t affected by it. These controls are toys, and real EQ will need to be handled upstream.

On the plus side, the auto-off seems to work as one would hope. There’s a inrush thump when turning on the device, but not when it turns itself on.

Measurements

(I used an ECM-8000 placed at one foot, on-axis, and a Tascam M-164UF for ADC to a PC running REW. Tests were conducted in my untreated garage with minimum movement between measurements so room coloration is comparable.)

I have often noticed that these subs have a very nonlinear volume response with the GAIN knob, so I started by looking at that. Shouldn’t volume change smoothly with knob position? Apparently that’s too much to ask:

sub120gain

The top five lines are all equal-angle turns of the GAIN knob, and should be equally spaced on the y axis. This lurching response is pathetic. I find the frequency shift of the dip around 28hz troubling. Fortunately, in practice this knob can be usually be set once and left alone.

Moving on to the FREQ knob,

There are a lot of wiggles here, but it isn’t necessarily the sub’s fault. To neutralize the room coloration here I’ve divided the responses by the curve with FREQ set at max (all the way clockwise):

sub120freqnormalized

It appears to be a -12db/octave low pass filter (but this is a different slope than the built-in 6db/octave high pass on the outputs!). The slope of the fall-off varies with frequency, and like the GAIN knob the FREQ knob lurches between 10 and 11 o’clock. Yuck. Another knob to be tolerated by setting once and leaving alone.

Mods

I tried blocking one and then both ports, with and without extra polyfill stuffing. In theory blocking ports converts it to a sealed box, bringing the resonant frequency of the enclosure up, while possibly making it quieter at lower frequencies.  More fill should bring the resonant frequency down.  In practice one can never tell what will happen. We are detuning a (hopefully) carefully tuned system, which might be bad.

A tennis ball is a little small to temporarily block a port, but these nerf balls are perfect.

Here are the results, absolute and then normalized to the vanilla state with factory filling and both ports open:

There is nothing here to recommend blocking a single port with no stuffing, the mode I’ve been using. So much for trusting online user reviews.

Blocking both ports or stuffing produce similar curves. As expected, the sensitivity at the lower end is reduced, but not badly (-5db). In exchange the response smooths out a bit. In particular, the troubling dip at around 28hz vanishes.

Whether stuffing or blocking ports is the right thing depends on all kinds of things, not least being the room one is in. This particular sub I chose to leave stuffed and block both ports. I’m interested in musicality and not high volume, and use an external DSP equalizer to flatten the curve. This makes the most important constraint the least sensitive frequency. The dip at 28hz bothers me more than the roll-off at 20hz, because I can hear 28hz more than 20hz. I also didn’t like that measurements suggested that dip frequency varying with gain.

Here you can see it filled and then closed up again. Although this is a pretty trashed exterior, even my low standards couldn’t stomach leaving Nerf balls sticking out the back, so I glued some wooden squares primed black over the ports.

Based on these numbers I also went ahead and closed the ports on the two other subs I have. I didn’t bother to stuff them.

Conclusion

If I had to pick one word for this sub, it would be: adequate. It is solidly built and reproduces low frequencies at moderate volumes, but also looks and measures cheap. The built-in crossovers are not to be trusted. For musicality they really need upstream DSP to correct the response (which all subs need anyway to remove room color).

I’m a cheapskate and am going to continue using them – in fact, I kind of enjoy the duct-tape, anti-audiophile aesthetic. If you have a favorite cheap sub, let me know.

Tang Band T1-1942S speakers

I bought these speaker modules from Parts Express for near-field use in an industrial post-and-wire shelving mobile standing desk.  I’ll talk more about the desk and why I would choose these speakers in a later post, but wanted to share some rough and ready measurements.

As described by the manufacturer:

Tang Band’s line of patented speaker modules are designed for modern audio applications such as sound bars, desktop multimedia speakers, flat panel tv speakers, and portable audio systems. Don’t let Tang Band’s diminutive driver modules fool you; each is built with a passive radiator that extends midbass frequency response, allowing these to mate to a subwoofer system quite easily.

The T1-1942S speaker module from Tang Band and is built around a passive radiator assisted, polypropylene-composite 46 mm dome driver. This allows for a very full frequency response from a single, small driver. Advantages of this type of speaker design include the elimination of crossover components in the critical midrange, superb off-axis response due to the use of a small driver (without hot spots), and perfect time-alignment.

The specs looked promising, claiming 78Hz through 20kHz.  The response graph looked like it would usefully reach down to 100Hz, at which point I’d cross over to a sub anyhow:

T1-1942S-spl

This is a pretty deep range for such a small package, achieved with the passive radiator.  Tang Band has a bunch of patents, some based on the illustrations clearly covering this unit, but nothing that a quick skim showed especially interesting to an unwashed end user.  They directly form parts via injection instead of gluing discrete pieces together, and so on.  I was curious to see them in person.

Here’s what the units look like in my hand, front and back:

They feel solid, but have no way to disassemble nondestructively.  I immediately hated the crappy, tiny connector which forced me to resolder the driver itself; no big deal, I’m clearly not using as the manufacturer intended, probably in a TV:

Tang Band incorporated mounting tabs and a two-pin wiring harness jack into the T1-1942S to facilitate mounting and amplifier connection to the optional KIT-0041 speaker module mounting kit with amplifier. For custom applications, two polarity-indicated solder points are built onto the driver PCB.

IMG_20130918_140146

I wired this up through an 80Hz high pass filter as the manufacturer recommends, and made five frequency sweeps with a cheap, generically calibrated ECM8000 microphone:

  1. 12 inches from center of module on axis
  2. 1 inch from tweeter on axis
  3. 1 inch from passive radiative on axis
  4. 24 inches from center of module on axis
  5. 12 inches from center of module on axis (desktop)

The first four had the module suspended from wire shelving with about a .25in gap.  The last had it lying face-up on a butcher block surface.

I’m in-between workshops right now so these were not careful tests.  Just to make sure you don’t get the impression that this was tightly controlled, here you can see it sitting right next to the acoustically reflective monitor for the desktop 12″ test, comb filter galore:

IMG_20130924_095316

Here is the suspension mount, and perspective from ear level.  The shelves are of wire of varying thicknesses all less than .25″ wire, well below a 1/4th wavelength for the frequencies of note.  I doubt diffraction is playing a big role.  The geometry of the desk is also such that there is no direct reflection off the monitor or rear wall when suspended.

Finally, here are the measured SPLs.  The first graph is unsmoothed and the second smoothed to 1/3 octave. The absolute numbers aren’t comparable, so I added offsets to set the single octave smoother curve equal at 1kHz.

There’s no way that the practical range goes down to 78Hz as claimed, but it’s as good as I was hoping for.  There’s no suprise that the manufacturer’s curve looks better – it’s their job to measure in a way that it will.  Maybe they aimed their microphone at the passive, but the way I measured wasn’t fair either.  Suspending in the air lost some bass extension.  Touching the modules with my fingertips while hanging indicated pendulous modes at lower frequencies, which may be one reason they performed better lying on the desk.  One might be able to get that back when suspended simply by attaching more mass.

I was much more disappointed by the high end.  My lame measurements can’t be blamed for the consistent fall off after about 5kHz, which doesn’t look anything like Tang Band’s graph.  You might be able to push this speaker to 10kHz at low volumes, but it can’t possibly reach 20kHz.

I have a sub on a lower shelf of the desk that handles the low end.  The midrange is very clean and I get good imaging.  My ears can only hear up to about 15kHz and I experimented with correcting the high end, pushing up gradually around 10k to get a little more high frequency extension, but AB testing showed me it just wasn’t significant.

Regardless of the curves, subjectively I’m finding myself quite satisfied.  I’ve been listening to them in the suspended position at ear level on either side of my monitor while working for the last two days.  In this fixed-position, near field application my impression is similar to wearing a good pair of headphones, without managing a cable or getting sweaty ears.   I will be using these.

Not getting high

I’m playing with some small speaker modules that the manufacturer recommends using with a 2nd order (-12 dB/octave) 80 Hz high pass filter.  Reading the manufacturers data sheet I couldn’t be positive they couldn’t be damaged at low frequencies, so I went ahead and bought some passive filters from Parts Express.

Here’s what they look like in the catalog, and the circuit such parts have:

Here’s what arrived.  Because of the low frequency the inductor is huge which makes the whole thing pleasingly heavy.

At first I was just going to keep them on the workbench, and then I realized the high probability of them accidentally touching metal, attracting a magnetized screwdriver, or some similar diaster.  So I decided to put them in an enclosure.  I had to break one module apart to place the inductors at right angles to avoid coupling:

I got all the way to hot gluing them down when I thought, I’m going to all this work, so let’s make sure I can also use this as a low pass filter.  After all that’s just a slightly different wiring.  Had I purchased a low pass filter instead, it would look like so and have this diagram, swapping the capacitor and inductor:

I set out to print myself a little circuit diagram to glue on the front of the enclosure, and at this point looked closely at what I was wiring up – not what was on the label, mind you, but what was actually in front of my eyes.  Something was utterly wrong.  Either I didn’t understand passive filters at all, or I had been sent the wrong thing.  I consulted the photos I took of the label before gluing them down (seen above) – no, that definitely says high pass filter.  I went back to check my shipping receipt.  Parts Express also claims they were sending high pass filters.

After head scratching I realized what had probably happened. The low-pass and high-pass designs are nearly identical.  They are the same size and use the same components.  The traces on the PC board being the only difference.  Whoever assembled this put the wrong label on the board.  I wonder how many of these they shipped? Caveat emptor.

Fortunately, in my case it was super easy to fix, since I could still wire it up any way I want.

Now it works as intended, and should be handy to have around for speaker testing.

Between the notes

As a teen I sang and played drums with a show choir, sometimes “going on tour”, to performances and choral competitions out of state.  We once performed at the National Cathedral, an impressive venue.

It was an ear-opener.  The organ pipes are located considerably farther from the audience than the choir.  Wikipedia says the full length of the hall is 457 feet – that’s a full half-second for the first impulse of sound to reach the back, and of course much longer for echos.  To keep the audience hearing the organ and choir in sync, the organist would have to play significantly ahead of the choir.  A pipe organist has to learn to play only against music in one’s head, and discount the notes still arriving seconds later.  I’ve always marveled at this ability.

For the cathedral concert we were given a choral piece to be conducted by the composer himself.  As soon as we rehearsed I could tell there was going to be trouble.  It was prissy and experimental – written for the composer himself to flesh out some statement he wanted to make about composition.   There were multi-octave pitch changes that were hard for amateur voices to land.  Rhythmically I couldn’t find any groove.

This is not knocking experimental music.  I have a soft spot for musicianship that pushes boundaries, and I’m glad musicians do it.  Sometimes this leads to greatness.  Composers we revere now, even regard as stereotyped and quaint, were often regarded as experimental in their own time.  But often music that is a challenge for the performers and the audience isn’t to be enjoyed so much as intellectually appreciated.  Great music comes from emotional power located outside the intellect.

As a drummer, I’m particularly sensitive to this.  I find I have two modes of playing.  One is practice mode, where I think of individual notes and the technology of playing.  In the other, music is felt and flows, thought is at a higher level than the notes, and emotions are accessible to me.  Music that is too technical for the skill of the performers gets everyone stuck thinking rather than feeling.  When the performers can’t even get into the music, God help the audience.

Here’s what I might be thinking when in the technical mode: “this note now, count out the beats, two-and-three-and-a-now, wait for the bass (I messed this up earlier), an eighth rest and now.”  And in the flow mode it is just listening and observing: “That feels right.  Let’s play more aggressively.  The piano is playing behind everyone else’s beat, bear down and help her out.  The bassist just created an opportunity to groove together, ha ha, that was nice.”

Back to my story.  Being the drummer, I was asked to play the composition’s scored tambourine part.  The piece was irregular and I really didn’t get it.  I wanted to understand what the composer was trying to do, but the note placement just felt too random to absorb.

We barely had any time to practice, but did have one small-audience, low-stakes public performance before the big cathedral performance to polish things.  When the time came, I lost my place in the score.  When this happens one has three choices:

  1. stop playing,
  2. play a regular beat until things improve (usually quarter notes), or
  3. improvise.

Since I had been playing, stopping would be noticeable, so #1 was out of the question.  The composer was conducting us as if he thought there was a beat, but since I couldn’t find it, I couldn’t do #2.  So I improvised.  Since the music seemed sort of random to me, I tried to fit in playing “randomly”, hitting the tambourine every now and then, until the song ended.  Lame, but one does what one must.

After this show the composer came up to me.  Shit, I thought, I ruined his piece.  “Nice job!” he said, smiling.  He seemed genuine.

Huh?, I thought.  And then: I have it made.  He doesn’t recognize how he scored the percussion, so at the big performance I’ll just do this again.  No worrying about the nasty arbitrary notes, I’ll just stare intently at the music while playing whatever I like.  Relax.  There is no need to practice.

And then came the big performance, and trouble.  Just before the concert, I’m introduced to The New Guy.  The composer thinks it would be even better if there were two tambourine players, and had found someone else to join me.  This won’t work at all, because we have to play the same notes at the same time.  No improvisation.  I am screwed!

(It is only as I write this now that I spot the irony.  Mr. Composer may have actually liked my improvisation at the first concert, and had I not pretended it was his work, he would not have doubled down.)

Fortunately, an actual performance often goes better than rehearsals.   Sometimes it has less to do with practice, and more with the focus that adrenaline brings.  Whatever it was kicked in and I was able to play, more or less, the music as it was written. Which is to say, I played the same thing as the other player.

Until the triumphant climax of the song.  I don’t know what happened.  Perhaps it was the sense of relief, letting guard down, or maybe the revenge of withheld errors accumulated throughout the piece.  The last two notes were played utterly out of sync.

They weren’t even close.  Nobody said anything, and likely nobody cared.  But I’ve thought back on those two notes more than any others I’ve played.

Sometimes we have to improvise.

Compensating for equal loudness contours

So, is it possible to compensate for the effect discussed in the previous posts, in which the apparent loudness of low frequencies is affected, among other things, by the volume setting?

I do this manually when watching TV at night, and I’m sure others do too.  Half the time the TV goes on I find myself compelled to turn the subwoofer up or down.  I do it by ear, and it depends on the content, but usually within around a 6 dB range.

In the room I do most serious listening in I use a DEQ2496, a nice, cheap, Swiss army knife (or maybe spork) DSP with pro audio features.  I’m only using it as a DAC and to EQ and delay for phase matching the subwoofer, although it is capable of other tricks. I have it set up to adjust the low response based on volume, and thought I would share the settings.

These settings use the dynamic equalizer (DEG) to track the volume, and progressively boost the low end by up to 10dB when things get quiet.  There was nothing scientific about these settings.  At first I kind of eyeballed the Fletcher-Munson curves, and then ended up using my ears and picking 10dB as a nice, round, safe number.

There are many caveats:

  • There’s no compensation for changes in loudness as I move around the room.
  • Settings depend on the relative volume of the equipment downstream, such as speaker sensitivity.
  • Loudness is only approximated by the average recent dB.   There’s a delay in changing the EQ when dynamics change.
  • A song with soft and loud sections will have inconsistent EQ.

More generally, this is an impossible problem.  Recordings of instruments played quietly do not sound like music played loudly but with the volume turned down. You could easily tell the difference between a piano played pianissimo and one played forte if the sound pressure levels were matched. The sound system has no idea what the original recorded volume was, or the intended distance to convey from the source. The process of recording and mastering sound already took its best guess at how to take loudness contours into account, and that step can’t be inverted.

Having said that, this compensation does seem to work in the sense of doing what I tried to make it to do. But I have found I usually prefer it off. I’m conceptually bothered when it is on, because I keep wondering how it might be coloring things at this moment, and I can only satisfy that itch by running over to the little DEQ2496 screen to see, which ruins my listening experience.

A distant rumble

The last post talked about how we perceive low pitches as varying more with absolute pressure levels than high pitches seem to.  It’s a big effect, a major perceptual bias.  Why does such a bias exist?  What is it for?  It’s too big an effect to be an accident, so there much be some evolutionary advantage.  Here I’ll lay out a possible explanation.

One might expect sound from a point source to decrease by 1/r^{2}, like light or any sort of other wave expanding outwards in the shape of a sphere with area r^{2}.  And this is true – as we move away from a sound the volume decreases rapidly, with sound pressure for an ideal point source outside dropping by 6 dB as distance doubles (dB is a logarithmic measure).

If sound were immune to friction creating heat, and there is nothing near the point source to reflect sound (perhaps a speaker hanging from a crane), then spherical expansion would be all we need to know.  But reality is very different – higher frequencies are attenuated with distance more.  Much more.

We have all heard boom cars and thunder from far away, so this is intuitive.  However, the physics is not.  Searching for the reasons for this brought up a ton of very bad, misinformative web pages.  (It appears that lots of people use Yahoo answers for their physics homework, and the kids answering clearly don’t know they don’t know the answer either.  It’s a virtuous cycle – as a teacher I might choose to pose questions for which answers blindly copied from the internet are obviously lame, making it easy to discern whether the student has a true understanding.)

There are many mechanisms at work.  Besides absorption in air due to friction varying with frequency, there are differences in how longer wavelengths refract around local objects, bending due to temperature gradients, foliage, ground absorption and reflection, wind, and so on.

There is a careful comprehensive ISO standard for modeling these effects.  True to form for a useful document prepared with public money, it can’t be downloaded for free, but I found a copy hosted by the FAA as part of an environment impact study.  Here’s a table showing some sample values of attenuation, in dB per kilometer.

Attenuation

Absorption by the air alone is a very significant effect at high frequencies.  At 8khz, 15 degrees C, 20% RH there is a 202dB drop in a single kilometer.  (Twenty orders of magnitude, is that to be believed?  I think they are expressing are dB in power, which means only ten orders of magnitude in pressure.)  This is an exponential effect that quickly swamps the loss due to mere spherical expansion of energy.  But low frequencies don’t feel it so much.

I can imagine other effects that might apply in a natural setting.  Outdoors among trees, each branch is a potential reflector of sound.  When we are very close to a point source, the reflections will be far away, relatively small in energy and therefore not significantly affect any frequency.  As we move away from the source, the reflections make up more and more of the total energy from the point source that reaches us.  Furthermore, as we continue to get farther away we will hear more and more 2nd, 3rd and higher order reflections.

Adding lots of reflections in the limit becomes convolution with a Gaussian.  Crisp impulses in the original signal turn into a smoothed-out lumps.  It’s easy to see that this attenuates higher frequencies.  I only brought up a Gaussian because it is algebraically special by being its own Fourier transform, so it makes a low pass filter, as seen here (image taken from this pdf):

GaussianLowPass

Ok, I’ll admit that last part is pure handwaving.  In any event, we know low frequencies carry farther.

So, my theory is just this: in natural environments, being farther away 1. makes things softer, and 2. attenuates high frequencies; therefore, our nervous system evolved to compensate.  Lower tones and softer volume will both let me know that the mammoth is bellowing far away, but it should continue to sound like the same animal as it approaches.

The modern world has volume knobs, and these confuse our sensory apparatus because the world we evolved in didn’t have any.

I find this idea simple and elegant.  It’s therefore probably wrong – I’d love to hear what others think.

Equal loudness contours

An interesting fact: as volume increases, low pitches seem to get louder than high pitches do.  To put it another way, turning up the volume on the stereo enhances the bass, without touching any other controls.  Conversely, turning down the volume on the TV late at night to not disturb someone sleeping can make the bass go away, and the audio sound tinny.

The general effect is called equal loudness contours or, more often and little confusingly, Fletcher-Munson curves.  This graph is copied around the web a lot, because the actual standards document isn’t reproducible for free:

EqualLoudnessContours

The first time I saw this graph I had to stare at it a while to understand.  The solid lines here are levels which are perceived as the same volume.  They zoom upwards at the high and low ends for frequencies we can’t hear well.   There are also some wiggly bits in the midrange that have to do with how we hear speech, but none of that is what I want to focus on.

Here’s an easier way to understand the data, from NIH, that inverts the first graph, showing directly how perceived volume changes with sound pressure.  It also normalizes by A weighting, which gets rid of the swoopy effects at each end.

nih_loudness

The lines show how loudness is perceived as volume is increased.  The key thing is that they aren’t parallel: increasing sound pressure makes the low end feel louder faster than it does at the high end.

Suppose your hand is on the volume knob and you are listening to a note two octaves above middle C (1khz).  You turn the knob and hear it go from the equivalent of a whisper (30dB) to level of a normal conversation (60dB).  Now you play the lowest B note on a piano (31Hz) and set the volume to be at a whisper.  The graph says this level is objectively at about 45dB, and by turning the volume knob the same as before (+30dB more), you’ll make it seem as loud as city traffic (as loud as 85dB at 1khz).

Now forget the graphs and listen to the effect yourself.  A good test is a movie with a sound track containing big explosions played at both high and low volumes.   At low volume the explosions will be wimpy, but you’ll be able to hear speech fine.

I was so surprised when I first got serious about listening and ran into this substantial and under appreciated effect.  It has all kinds of consequences, such as that perfect EQ is impossible, even for a perfect sound system playing perfectly recorded music in a perfect listening room.  It isn’t actually possible to once and for all correctly balance highs and lows, except for a phrase of music having particular dynamics.

Worse yet, the effect varies between individuals.  Perhaps this contributes to why some people often listen to music with volume turned way up – they are unconsciously EQing.

Of course, if I’m selling you audio equipment I’m not going to focus on how relative everything is.  The important thing in marketing is how people feel, and showing flat dB graphs make people feel good about their purchases.

The question is why should this effect exist at all?  What could be the evolutionary advantage of such a huge perceptual bias?  I’ll muse about that in the next post.

Antikythera fridge

Last year I bought a copper-colored vinyl decal with a loose schematic of the Antikythera mechanism on Etsy.  (For bonus points, it was even shipped from Greece.)  Originally intending it for a wall, I decided it looked better cut in half behind the stainless bars of the refrigerator.  The bars contribute to the mechanical vibe.

I was worried the thin strips would not hold up to the rough environment of the kitchen, but nine months later it still looks great.