View Full Version : A/B visual comparisons - and audio?

25-07-2009, 11:49 AM
We frequently discuss here making audible comparisons between equipment. I've just conceived an experiment about making comparisons with your eyes. We can discuss the results later and their possible relevance for audio, where we make comparisons by ear.

Using Photoshop, I think that I can demonstrate an interesting personal observation that I've made over the years about how humans make comparisons. It's easier to demonstrate this visually than using audio comparisons. The test is described in my next posting.

NOTE: Based on the early poll results I have revised this experiment to a) hide the results from voters to avoid undue influence b) verify any bias in the results is truly due to preference and not experimental technique. Also, this thread is now locked to prevent postings influencing readers and hence results.

Please vote!

26-07-2009, 09:39 AM
The more I consider this subject, the more careful I have to be in leading you through my observations. I cannot explain everything in one or two postings, so please accept my input spread over several contributions, then we can discuss our experiences together. Please vote!


HUG Visual comparison test No.1 (VCT-1) - the first step ...

Attached you will find two medium-quality pictures suitable for the internet. The resolution and size is the same. I've put the letters A or B on the right side by the sunflower in red. Once you open one image, you can use the PREVIOUS, NEXT, FIRST and LAST buttons under the image to compare them one after the other. I'd like you to compare these pictures. Do not adjust your monitor! Then choose the image one that looks best to you, on your PC, with the monitor set-up as you normally use it. Then vote for the one you prefer. What matters is your opinion, not anyone else's opinion.

The voting method has been revised to hide the poll so you are not influenced by other people's opinions. If you voted on the original poll, don't worry, I have transferred your vote to the new survey so please don't vote again.

Registered members of the Harbeth User Group can go here (http://www.kwiksurveys.com/online-survey.php?surveyID=NOOKJ_1cd4ade3) to vote. Please select A or B as your preference. Please vote once only. No personal data is requested or collected. Your results are completely anonymous.

Thank you.


29-07-2009, 08:30 AM

If you have not yet voted - please do so! I will close this thread tonight, 29 July (UK time) so that I can present my findings ... and then slip away for a few days.

Thank you.

29-07-2009, 08:27 PM
Thanks to those of you who took the trouble to vote, about one in ten who viewed the thread. I think that the preference was sufficiently differentiated that it is statistically useful but not a true scientific test .

Voters were asked for their preference between two images, but none were using colour calibrated monitors; PCs and MACs have a different gamma (mid-image brightness). Despite all these issues we were looking for a preference, viewed on whatever screen normally used and accustomed to. This image comparison is much like audio comparison; we were not at the live recording and not at the photo shoot so we can only express a preference according to our preconditioning and tastes.

Why make this test?

For over twenty years I've been aware that we humans make comparisons in a complex and often irrational way as I've cautioned readers here many, many times. Consumers often make purchase selections which wastes their precious time and money, detracts from the pleasure of music and sets them up for disappointment and 'buyers remorse'. So how can the eye or ear be deceived so easily?

What was the poll results from your responses?

19% preferred picture A and 81% preferred picture B.

Was this outcomes expected?

No. This result is the opposite of what I expected, and is why, as the early poll results started to appear, I hid the poll to eliminate potential bias.

Which picture was the 'reference'?

Picture A was the original, unadjusted technically correct reference image. Picture B was manipulated. I deliberately compressed the dynamic range of picture B. This had the effect of increasing the contrast in B, making B look more hard, punchy and vibrant and making A, the reference, look soft, weak and a little washed-out in comparison. B has a slightly plastic-coated sheen to it.

How do we know that A is technically correct?

Photoshop has a tool which allows us to inspect the dynamic range of the picture, to see how the tones and shades are distributed across the image, from the blackest black to the whitest white parts of the image. This is called the image histogram. More on this later. This is the proper way to judge a picture's technical characteristics, not only by eye.

How would a trained observer (a graphics professional) compare these pictures?

He would look for four basic clues about the image:

- How much detail is there in the shadows - if any?
- How much detail is there in the highlights (the brightest part of the image) if any?
- How evenly spread as the information in the middle-tones of the picture somewhere between black and white, the so called image gamma
- What is the dynamic range, the scope of information captured in the image from black to white? Is it like a billboard with just a few tones or like a museum painting with a millions of micro-tones? Does it look natural and believable? Life like? Over blown? Hyped up?

How did the untrained, casual observer assess these pictures?

The voting results show a strong preference for B, which is the manipulated image.

Why would viewers prefer manipulated image B compared to the reference image A?

B has the illusion of greater punch and is seemingly more dynamic. In fact (more later) B has significantly less dynamic range than A, especially in the darker tones. We know this for certain because I took the reference image A and I deliberately crushed its dynamic range to create image B. I'll show what I did later so you can replicate this experiment yourselves.

Technically, what did you do to image A's histogram to create image B?

See (histoB.jpg) attached

I trimmed from picture A those picture elements (pixels) which describe the fine details in the darker areas of the picture, such as the shadows. Thus, shadows become dense and monotonic - shadow or no shadow, no fine graduations of tone. I also removed the detail in the extreme highlights so that the picture is slightly burned-out. This is typical of an outdoor picture on a bright day where the fine detail in the clouds is erased. If you look at the yellow 'information discarded' rectangles in the histogram, it covers more peaks in the shadows than in the highlights so there is in fact more fine detail in the shadows than in the highlights; see histogram. So, my manipulation of image B has a more pronounced effect in its darker areas than the highlights. A quick peek at the picture confirms that there are no really bright, complex areas, except the lit bulb, above the sunflower - more on this later.

Can you point out the visual clues that show A is correct and where B is deficient?

Yes. Look at (A B_compare2.jpeg) attached. As I mentioned, there is an illusion that B is more dynamic, but if you look at the notes I've made on these image, you can see that the darker end of B is really horrible: no resolution at all. Dark, sinister black. (A B_closeup2.jpeg) shows a zoomed-in view. Look at the detail missing from the circuit board and the gears, and the filament detail missing in the glowing bulb in B.

Does MPEG (or similar compression system) have an equivalent effect on sound?

I suggest that a similar consequence to compressing the image may apply to MPEGs compression of sound, throwing away fine detail and only encoding the the so-called audible (visible) parts of the music:

- MPEG erases detail especially in the lower 'dark areas' of the sound, nearer the noise floor, in the decay of the notes into silence. This is justified by saying that the low-level detail is inconsequential because it is masked by the louder sounds. We've seen that in a picture, that is an erroneous conclusion

- The modern consumer seems attracted to the harder, seemingly more punchy sound, with greater contrast between tones, performers and instruments giving the illusion of greater immediacy especially in the middle and upper loudness. A compressed sound seems to get and hold the listener's attention; a softer (more natural) sound competes with the listener's attention span.

How does this relate to loudspeaker evaluation?

This is a extremely critical question. If, as we have shown here with visual comparisons, an untutored listener is presented with a neutral, natural, uncolored loudspeaker (type A) with a wide dynamic range that captures all the micro-tones in the sound, there may many more consumers who would chose loudspeaker B, with it's seemingly greater punch and heightened sonic contrast.s Only a minority of listeners would, from these results, prefer the softer, more evenly balanced tones of loudspeaker A with it's wider, easy-on-the-eye (or ear) sound and fine resolution in the quietest moments.

In a loudspeaker, what component is responsible for painting the sonic picture from black to white, silence to loud?

This is entirely the responsibility of the cone, primarily the bass/mid driver's cone. Stripping away all other tertiary factors (cabinet, crossover, stands even room) the fundamental defining character of a speaker is that of the sonic signature of the material from which the bass/mid cone is fabricated. There is no escape from this truism. Conventional cone materials (polypropylene etc.) do not resolve 'details in the sonic shadows' and cannot accurately replay shades of tone through the whole musical spectrum. This is a fundamental limitation of polypropylene's waxy, rubbery molecular structure: sound is trapped as friction heat when the molecules flex in response to the passing sound waves.

Harbeth's unique RADIAL? cone is an engineered plastic produced specifically and solely by us as a loudspeaker cone material, not a shampoo bottle plastic. At a molecular level is is much more rigid and bell-like hence, details in the sonic shadows are revealed - hence a wider dynamic range.

Any other observations?

When I visit the vision gallery in TV studios, it always surprises me how soft the pictures are on the TV monitors compared with typical home TVs. Pictures are broadcast as type A but almost all domestic users set their TVs for B, with a ludicrously high contrast setting compared to what leaves the studio. TV retailers demonstrate TVs with the contrast fully up, when in fact, the correct picture balance is typically at about 50-60% i.e. very soft. So what? Well, next time you play a movie, look at how much detail there is in the shadows. Does the actor's hair look like black paint? Or can you see micro-tones?

Provisional conclusion

I created this experiment to explore my hypothesis of some correlation between the way we assess loudspeakers (or perhaps audio generally) and visual images. I anticipated that as we here as a group enjoy the wide dynamic, easy on the ear (or eye) well rounded Harbeth sound with lots of detail that there would be strong preference for image If my hypothesis was true. The opposite has been found. I am unable to say if this is because there is in fact no correlation between the visual assessment process and the audio one, if different evaluation processes apply or if the experiment is fundamentally flawed. I sincerely hope that it instils a degree of caution amongst those comparing audio equipment to be wary of the being seduced by your senses in favour of the natural.

I hope this will fuel some comments, improved techniques and further theories.


30-07-2009, 03:43 AM
Very interesting indeed. My first impulse was picture B, but after looking back and forth a few times, I choose picture A because of the details I can see in the shadow regions.

This corresponds very well to my experience with hi-fi equipments. I have not had the luxury of good sound for the past 20 years, except for a period of a few months some 10 years ago when my brother loaned me his hi-fi system.

Finally I was able to purchase my own hi-fi system some 2 months back. It was the Audio Pro Image range of speakers plus a Denon AVR. So I thought at last I could finally enjoyed some good sound. But very soon I felt disappointed by the constricted sound - and I begun to fiddle with the Audyssey equalizer system. With Audyssey turned on, the sound is more expanded - but certain details are lost, and the sound scape became distorted. One glaring example is the baritone voices overpowering the soprano voices.

So my listening session became a dilemma of whether or not to turn on the Audyssey system. That was when I realized I had to look for new speakers.

I traded in my almost new Audio Pro speakers and brought my Harbeth 2 weeks ago. Since then I never bothered about the Audyssey system, and I never looked back.

Thanks for conducting this very insightful experiment.


30-07-2009, 11:37 AM
I struggled to form a preference for either picture, I guess because it was not that important for me but give me tiny differences in sound and I'll have no trouble deciding which I like best and why.

Hi-Fi and natural reproduction has been my life for the past 49 years, so it comes naturally to me to discern small differences in sound but give me a photo or TV and I really couldn't care less about miniscule changes.

Ned Mast
30-07-2009, 01:43 PM
Thanks for the interesting "test", Alan. In photographs I always look for a the full range of values. A quick look at image B showed no details in the shadow areas, while A did. Image A was an easy choice for me. I think you're right in proposing similarities between visual and auditory preferences.

30-07-2009, 02:29 PM
Hey, Alan

Great job on this one. Choosing Photo A was a no-brainer for me, and I'll explain why.

My brother has worked for Eastman Kodak for 33 years, initially as a chemical and paper engineer and now as a digital imagery specialist. Back in the 90's when Fuji became a big competitior with Kodak, he showed me how Fuji's film/printing process created an overly vivid, high contrast image, very similar to Photo B in the experiment. The typical consumer thought this hyped image to be superior, hence, I'm not at all surprised that 81% chose Photo B.

I'd guess that most of us have highly developed senses in one area or another, so perhaps we Harbeth owners have greater hearing acuity than the average person. My brother, on the other hand, would likely choose a "hotter" speaker over a Harbeth in a listening comparison.

I've sent your concluding post to my brother-I'm sure he will be fascinated by it!

Thanks again-for everything that you and your team do!

Bob LaBarca
State College, PA

30-07-2009, 04:45 PM
Interesting test...
OK, I will admit to being one of the unsophisticated crowd who chose B! As someone with quite a lot of photography experience I did so in the full knowledge of what the differences between the pictures were and why.
The comparison with music and hi-fi preferences bears a bit more teasing out though. I gave my preference for image B on the basis of a quick comparison. Which makes a more immediate and pleasing impact on me? B. Which would I prefer to live with over a period of time as a poster on the wall? A. What the test shoes I think is the folly of making short-term consumerist decisions, which is obviously particularly problematic when choosing a hi-fi system that needs to give years of long-term pleasure! I perhaps wouldn't choose Harbeth speakers if I listened to music every now and then in 2 minute impressive bursts (punchy picture B). But I have chosen Harbeth as someone who wants a speaker to settle down with for the long-term (natural picture A).

30-07-2009, 06:32 PM
Interesting test...
OK, I will admit to being one of the unsophisticated crowd who chose B! As someone with quite a lot of photography experience I did so in the full knowledge of what the differences between the pictures were and why.
The comparison with music and hi-fi preferences bears a bit more teasing out though. I gave my preference for image B on the basis of a quick comparison. Which makes a more immediate and pleasing impact on me? B. Which would I prefer to live with over a period of time as a poster on the wall? A. What the test shoes I think is the folly of making short-term consumerist decisions, which is obviously particularly problematic when choosing a hi-fi system that needs to give years of long-term pleasure! I perhaps wouldn't choose Harbeth speakers if I listened to music every now and then in 2 minute impressive bursts (punchy picture B). But I have chosen Harbeth as someone who wants a speaker to settle down with for the long-term (natural picture A).

Hello Tozen,
fully agree with you. I chose picture B exactly the same mistake as with the purchase of my previous speakers. After few month I started with the merry-go-round.
The speakers were not bad, but something missed here, some redundancies there..... Lucky me, I discovered Harbeth.

30-07-2009, 10:17 PM
I have had a life-long interest in grading images, and in the office I'm notorious for my sensitivity to an images colour hue (regularly needling Trevor, our PR man about colour cast that he claims he can't see until pointed out! His excuse is that he's a wordsmith ...) and I am completely familiar with your comments about Fuji (positive) film. In fact, a few years ago, our then retained Harbeth photographer was using wet (positive) medium format film. he's all-digital now. I forget which model we were photographing in his studio (I always like to be present: lighting is so extremely critical) but I asked him if we could try alternative films.

In those days he had a dark room, and in the red light I waited whilst he loaded the exposed films into the developer and about thirty minutes later we had the developed positives to study on the light-box under the loupe. It's exactly as you say: the Fuji was slightly over saturated compared with Ektachrome positive (E6 I think) and (if I recall correctly) had a slightly yellow cast (hue). If you set your reference as the Fuji (the B image) and then studied the E6 (the A image) under the loupe you'd say that the E6 was significantly softer (lower contrast), 'cooler' and had a slight blue tint. I preferred the punch of the Fuji.

It was only when we took the light box out onto the studio floor where the speaker was still set-up under the lights that the Fuji was revealed as 'better than true life'. The Kodak E6 was in fact correct.

P.S. Do you remember that 3M (negative) film which was often given away free with mail-order processing? Ghastly magenta hue.

31-07-2009, 08:00 AM
Now we've shared some common experiences concerning the evaluation of colour images, we can move to the second part of this discussion. Actually, I conceived Part 2 first, but realised that some prior explanation was necessary, hence Part 1. There is no voting this time round - I'm actually going to tell you what I've done to the images; all I want you to do is look at them and try to connect what you see with my explanation of how I've manipulated them. Then in Part 3, next week, I'll explain how this may (or may not) directly relate to the evaluation of hi-fi equipment.

Summary thus far

In Part 1, I selected a reference image (A) and manipulated it (B) to give B with lower resolution and limited detail in the shadows and highlights was preferred by 81% of voters.

Explaining the second test

This time, I have locked the resolution in the shadows and the highlights and I've manipulated the image in the middle tones - the opposite situation exposed in the first test.

Attached you will find picture (CDE2.jpeg). This is the familiar BBC test card, latest version version W (widescreen) which has been used since the introduction of colour TV to the UK in the 1960s. It's very well designed and was the primary tool for the TV technician to align a customer's set, especially in the analogue CRT era. Note even how even the X in the game is the precise geometric centre of the image. It's a masterpiece.

The reference image is D, in the middle. We know it's the reference image because if we look at the technical characteristics of the image in Photoshop, we can see, for example, that the background grey is exactly 128 points on a scale of 0 - 255. That means it is exactly in the middle of the tone map between the blackest part of the image and the brightest part.

What have you done to images C and E?

I've chosen this image because the predominate colour, by area, is "mid grey". I've locked the black and white points and I've manipulated gust the energy in the middle of the picture's tone curve. In other words, I've manipulated the image where the greatest effect will be visible because the mid-tone occupies the largest area of the image.

In picture C I have reduced the energy in the middle of the image tone curve by 10%. In D I have boosted the middle tones by 10%*. There should be no difference detectable in the blackest or brightest parts of the image - only in the middle tones.

Histograms of the image

I can show how I manipulated the image if we look at the image's tone curves. The graphs shows the linearity between input and output - that means, between the images latent characteristics and how it is presented on your screen. For D, the reference image, the blue line is exactly 45 degrees which means that for every change of shade in the actual image file, there is an exact change of brightness on your screen (hopefully). So, a 17.3% increase in the image from one shade to another would be reproduced with exactly 17.3% on your screen. However, for images C and E I have 'dished' the linearity curve. Both the black and white points remain locked (unlike in the first test) and pass through 0 and 255. So the input-output linearity is what we're interested in here.

What to look for in images C, D and E

Concentrate on the shade of grey comparing C with reference D, and D with E and then C with E. Also note how Carol's skin tone has changed, more so than any other picture element - which is why skin tone is so useful in establishing image quality.

Why I chose this particular image

As you will know, our founder Dudley Harwood was the senior engineer in the BBC's loudspeaker development dept. in the 1960s. As the time he was actually developing the BBC's famous monitors (from which all Harbeths are derived), the video department in the rooms adjacent were working on colour TV. So the BBC Test card was the video department's direct equivalent in precision and philosophy to Harwood's loudspeaker evaluations. Here (http://en.wikipedia.org/wiki/Test_card) and here (http://www.meldrum.co.uk/mhp/testcard/bbc_test.html) you will see the uniqueness of the BBC test card is that it uses skin tone (a picture of a human) to give even the untrained eye a good clue as to overall picture balance and colour cast, essential to achieve a natural picture balance. All part of the BBC's 'natural sound, natural picture' emphasis in the 1960s.

What's the relevance of this test to audio?

I'll leave you to mull that over whilst we drive to Northumberland for a few days. All will be revealed in the final part!

* 10% is approximately 1dB.


16-08-2009, 10:46 AM
We've talked often about comparing electronics. There seem to be two ways of making comparisons:

1. The strict objective comparison where every conceivable variable that could influence the listeners (undoubted) preference is removed. For example, an instantaneous switch-over would be employed or the listeners could be blindfolded.

2. Subjective comparison, perhaps minutes, hours or days apart where little or no effort is made to remove variables and the listener is able to touch and see the equipment and make adjustments.

Both approaches are valid, within their limitations. No.1, the objective approach, lost favour in the 80s when it was discovered that under controlled conditions, it was much more difficult to decide if A really was preferable to B. Our point here is not that approach 1 is necessarily better than 2 but that objective testing may reveal that differences are smaller than previously believed. I do believe that audio equipment can and often does sound different under 'normal' (subjective) comparison. I have heard this myself. I know that listeners truly do hear differences. I truly believe that I did. But why? Are these preferences validated by repeatable tests? Is there a danger that under slightly different subjective circumstances the listener could actually reverse his opinion? Yes, that does happen. Often. With carefully constructed objective tests perhaps inexpensive amp C actually outperforms expensive amp F, or expensive amp B is proven worthy of the price as it is definitely superior. And even if B is sonically superior, if it's known to be unreliable perhaps amp K would be a better all-round deal? You get the idea.

The biggest advantage of objective testing is that it is repeatable, probably anywhere on earth. This empowers the buyer to make the best overall performance-ranking decision and then if necessary re-rank according to cost and his budget. If the conclusions are truly valid then they will stand the test of time.

This thread has shown how sensitive we are to small manipulations in our sensory perceptions. We've concentrated on our visual sense but exactly the same applies to our ears. Take a look at the picture I've taken of a few amplifier volume knobs. A, B, C are scale-marked. D, E, F have no scales. E and F are not real physical volume controls - they're rotary encoders that send 'volume up, volume down' commands to the microprocessor; there is no way I can describe to you how to set a particular listening level with these two knobs as they just go round and round. All these amplifiers have different power outputs, different gains. Furthermore, there is no international standard for graduating the loudness at your ears with positions on the volume control. At a quick check, position "7" on knob A gave the same loudness at my ears as position "4" on knob C. Would you know that by just looking? No. If you were making the comparison between A and C, would you set both to the same scale mark? Or a compromise position, say 12 o'clock? If you did you would have negated the comparison entirely: you would not be comparing like with like at your ears.

Hence, without test equipment to measure loudness at your ears it is impossible to reliably compare any audio equipment - in this case, amplifiers - subjectively because the loudness at your ears is not normalised across the items under test. Even if you had an instantaneous switch-over, the conclusions would be completely worthless. Loudness, as is well known, is a listening variable which has a huge effect on perception of quality and balance*.

Of all other variables that impinge upon audio comparison, difference in loudness between A & B must be eliminated for the comparative deductions to be valid: controlling level (loudness) between items under test is super-critical to allow reliable conclusions to be drawn from audio equipment comparison. If I achieve nothing other than making you fully aware of this factor when you compare equipment my mission will have been fulfilled.

*Obvious example: listen to your hi-fi and normal volume; then dramatically lower the volume and evaluate the bass response. There seems to be no bass, all middle and top. This is because the ear's perception of bass is highly sound-level dependent.


18-08-2009, 11:01 AM
This item appeared on the Institute of Broadcast Sound's forum today. It concerns the precise electrical/mechanical graduation of the resistance element (hence the volume steps) of two power amps fitted with volume controls and of the same output power.

> Back in the 1960s, just about the time that Marshall guitar amps
> were being made famous by Jimi Hendrix, their main UK competitor was Hi-Watt.
> Both products knocked out about the same power - 100 watts RMS from
> valves but Hi-Watt fitted LINEAR Volume pots rather than the conventional
> LOG ones used by Marshall.

> It meant that in showroom side-by-side comparisons, at volume 1 to
> 4, the Hi-Watt ALWAYS sounded much louder than the Marshall.
> It was only when you got on stage and ramped it up to 11 that the scam
> showed up and it was obvious there was little difference.

I already planned to illustrate this point in more detail, but this suffices for now.


You cannot compare amplifiers properly unless you ignore the volume control knob markings and measure the loudness at your ears. I illustrated that in the previous thread. I've noted that the volume graduations printed on the front panels are non-standard and have no meaning regarding actual loudness at your ears: there is no agreed standard for rotational angle and loudness. None. Never has been.

To make matters even more complex, the increase/decrease in loudness for small rotations of the volume knob is not standardised either. So changing from position 12 o'clock to 2 o'clock may give a 2%, 10% or 50% change in loudness depending upon the track resistance characteristics deep inside the volume control as mentioned in the Marshall/Hi-Watt amps above. Only measurement can remove these variables. No human can accurately detect and remove these variables when making subjective comparisons in their ear/brain as our ears simply don't have the resolution to do so. Our ability to detect changes in loudness is very poor. So dramatic are these variables that, in my honest opinion, it's really an exercise in time wasting to compare amps without measuring equipment to be at least sure that the loudness is the same at your ears and then that you are comparing like with like. Otherwise you will hear differences between amps due to differences in loudness. That's guaranteed.

Once this loudness issue is put to bed can we move on to the core task of properly evaluating amplifiers. I don't doubt that whether by design or good luck one amp is inherently better sounding than another - but which one? There might be some real surprises.

18-08-2009, 06:50 PM
It meant that in showroom side-by-side comparisons, at volume 1 to
4, the Hi-Watt ALWAYS sounded much louder than the Marshall.
It was only when you got on stage and ramped it up to 11 that the scam
showed up and it was obvious there was little difference.

I already planned to illustrate this point in more detail, but this suffices for now.


You cannot compare amplifiers properly unless you ignore the volume control knob markings and measure the loudness at your ears. I illustrated that in the previous thread. I've noted that the volume graduations printed on the front panels are non-standard and have no meaning regarding actual loudness at your ears: there is no agreed standard for rotational angle and loudness. None. Never has been..

Nigel Tufnel: The numbers all go to eleven. Look, right across the board, eleven, eleven, eleven and...
Marty DiBergi: Oh, I see. And most amps go up to ten?
Nigel Tufnel: Exactly.
Marty DiBergi: Does that mean it's louder? Is it any louder?
Nigel Tufnel: Well, it's one louder, isn't it? It's not ten. You see, most blokes, you know, will be playing at ten. You're on ten here, all the way up, all the way up, all the way up, you're on ten on your guitar. Where can you go from there? Where?
Marty DiBergi: I don't know.
Nigel Tufnel: Nowhere. Exactly. What we do is, if we need that extra push over the cliff, you know what we do?
Marty DiBergi: Put it up to eleven.
Nigel Tufnel: Eleven. Exactly. One louder.
Marty DiBergi: Why don't you just make ten louder and make ten be the top number and make that a little louder?
Nigel Tufnel: [pause] These go to eleven.

18-08-2009, 07:17 PM
Now, there is a way to really, precisely control volume and to make every volume control behave exactly like every other one. It was used by the BBC in previous generations of control desks when they used rotary volume controls (before slide faders).

I have found in the BBC Engineering Division Training Manual (published 1942) that the BBC were very concerned about controlling volume, in a regulated way consistent from studio to studio. This involved a beautifully crafted arrangement of brass studs and a wiper that slid across them as the control was advanced or retarded. Between each stud was a fixed, precision resistor soldered into place (I've seen pictures of boxes of resistors individually hand tested to find the exact value they want).

This arrangement had the advantage that the volume level could be described precisely according to the angular position of the black/white skirt around the base of the knob. You could see from the other side of the room how open the fader was, and how loud one channel was compared to others. There was precision about sound level. The cost of making such a control now frightens me. But the certainty. The accuracy of control. Incredible.

Control of loudness in audio is fantastically critical. The listening level is a critical part of the emotional experience of listening to sound. How would you feel if your girlfriend SHOUTED sweet nothings in your ear at 120dB? The same words at 50dB would have an entirely different effect. Same sound wave, different replay level = totally different emotional response.

Once again, I'll wager that there isn't an audio subject relating to perception of quality which the BBC hasn't properly researched and documented - all thanks to the British taxpayer.


19-08-2009, 09:39 AM
Very intriguing findings:


Understanding dB for sound

Our ears expand when it is quiet to hear detail and contract when it is loud.

1dB = x 1.26 power change we hear as smallest loudness difference (laboratory only).
1dB sound change is too small and not used in most calculations (laboratory only).
9dB is not used because the 1dB difference to 10dB is too small to hear.

3dB = x 2 power change we only hear as a bit louder.
6dB = x 4 power change we only hear as a bit louder again.
10dB = x 10 power change we only hear as double as loud.

Audible difference
Difference between 100 - 200 Watt sound system, is minimal (3dB)
Difference between 100 - 1,000 Watt sound system, is twice as loud (10dB)
(who needs BIG amplifier?)

(see attached graph)

0dB SPL (Sound Pressure Level) 0.0002 dynes/cm? (20micro Pascals) at 2kHz
Reference for, threshold of hearing the smallest modulation, of atmospheric pressure.

By paying attention to this graph, it can be seen that at the threshold of hearing, our ears are approx 60dB (one million times) more sensitive at 2kHz than at 40Hz. This is the reason a loudness switch is put on most domestic sound equipment. The loudness switch boosts the bass to compensate for our hearing at low level. But at high power, all frequencies tend to be heard at approx the same level.

After the telephone was invented, it was noticed that a ten times power change (10dB) was only heard as double or half as loud. A two times power change (3dB) is noticeable. Later, these measurements became called the Fletcher-Munson curves. The large energy variations in our environment (light, sound, touch, taste, smell, temperature etc) are compressed by our sensory system, in a subjective and complex manner.

19-08-2009, 07:03 PM
Studded faders, as they are known in the trade, made it beyond the days of rotary controls into the era of sliding (quadrant) faders. But they weren't without their problems. Under certain, highly critical conditions it was possible to hear the signal level jump from stud to stud, especially at the bottom end of the volume range where the jump between studs was >3dB.

The final generation of this type of fader had 30 studs covering the whole attenuation range, studs 30-15 probably changed level by about 1dB per stud, below stud 15 the level drop between studs gradually increased. Ten studs up from full attenuation gave about 40dB signal attenuation, five studs from the bottom and the signal was 60dB down and absolutely mute at 0, the very bottom stud (this last point is vitally important in broadcasting where a channel may contain a high level signal which, for reasons either technical or editorial, should not be heard at all).

The traditional remedy for faders where dirt or wear had made the jump more obvious than it was on new and unworn components was to rub your finger on the side of your nose to obtain some of the natural greasy residue that our bodies place there, then rub the grease onto the exposed studs. Worked a treat.

These stud faders had another 'up and over' quadrant version (the forerunner to the modern sliding fader), and were sometimes found in stereo: that is, with two audio paths through the same fader, each subject to exactly the same attenuation at all times. At least, that was the theory. In practice they were built so that the A & B (i.e stereo left and stereo right) paths changed in an alternating pattern (one for you, one for me, etc.) which tended to make the stereo image wander around in a most odd manner. Another solution was to strap two adjacent mono faders together with a mechanical clip which, unless the fit was very rigid, was even worse at ensuring that the A & B legs of a stereo source tracked together. If the fit was rigid enough to ensure good tracking, the whole assembly was likely to be so stiff that you needed a significant push to move it.

Talking of push, while the BBC made its own kit (or at least, had sufficient clout to tell manufacturers what it wanted), sliding faders got louder by pulling them toward the operator. Bear in mind that this decision, although contrary to present day practice, was made well before the world at large had ever seen a sliding, or linear, fader. The exact reasons for this choice are lost in the mists of time (there are many rumours which purport to offer an explanation, but nothing definitive AFAIK).

The BBC had a preference for quadrant motion where the fader arm operated around a radius of about 12cm over a 60? arc. This fitted in naturally with the way the hand/wrist bends and required less obvious movement than is needed for a true linear fader - which came into existence several years after the studded quadrant, using much cheaper conductive plastics as its resistive element. Electrically, a standard potentiometer.