DELTA is an independent organisation that specialises in testing products of almost every imaginable kind. Our Q113, however, is the first speaker to go through a blind test. The testing is taking place in the specialist SenseLab department and it's going to be really exciting!

"Guuud, the new fragrance is just amazing", exclaims the woman in the perfume department. But the perfume is the same as always. It's just the name, the packaging and the advertising that have changed. In other words, the woman was smelling the marketing, not the perfume. The above is an imaginary example, but you have probably had similar experiences. For example, the red wine for DKK 400 is next to the one for DKK 50. Does more expensive wine taste better, or does the higher price just reflect that the wine comes from a smaller field? Does the price reflect marketing costs? Can we even separate out the ingenious mechanisms that help give us an impression of a product and who sets the standard for good and bad? It certainly does not appear from the product.

DELTA's foyer. Here you will meet the words that tell what is going on in the company.

We want to separate the sound experience from all other impressions

DELTA does not currently deal with scents and tastes, as in the examples above, but their special SenseLab specialises in testing audio products of many kinds. In the SenseLab you will find specially designed listening rooms, a team of testing specialists and a specially trained listening panel made up of people with sublime hearing. The four Q113 loudspeakers are now set up here, and will be the first loudspeakers in Denmark to be assessed by SenseLab's impartial and sharp panel. 10 of Ingeniørens readers are also invited to take part in the exciting test, where the sound from the loudspeakers is the only element the testers will be familiar with. By separating the sound from all other impressions, such as construction, price and design, I expect to get both an assessment of the project's speakers against each other, but also an assessment of how reliable our own test results are. The blind testing of the Q113 speakers is done entirely without my involvement, and therefore in the next few paragraphs I will hand over to Torben Holm Pedersen, who has a master's degree in acoustics and is a senior technology specialist at SenseLab. Torben will take you through an introduction to DELTA and a presentation of the facilities at SenseLab. This will be followed by a review of the actual test that the Q113 loudspeakers are subjected to and a presentation of the test results. I then return to assess the results and draw some conclusions for the project. Torben will, to the best of his ability, be on hand to answer questions on the blog. However, I would like to repeat that you are now witnessing something that is not normally seen, and certainly not in real time in the public domain! I look forward to a journey that (might) lead to a new and unknown place.

I now give the floor to DELTA SenseLab

A short presentation: DELTA is an independent and self-governing company, located in the Research Park Scion-DTU in Hørsholm, with a total of 270 employees. DELTA is the result of the merger of a number of smaller institutes, the first of which was the Sound Technology Laboratory from 1941. DELTA is an Approved Technological Service Institute (GTS) approved by the Ministry of Research, Innovation and Higher Education. DELTA works broadly with testing and consultancy in areas such as electronics, information technology, micro- and nanotechnology, light and optics, and acoustics. Most relevant in this context is the SenseLab department, where we are 6 people working full time on sound testing of various signal processing algorithms and products in audio, hearing aids, telecom, product sound and environment (noise). Read more about SenseLab here .

About the listening test itself

A test is performed to characterise the sound from each speaker type as well as a preference test. Both tests are performed as blind tests where each stereo set is compared in turn to a set of reference speakers with a fixed position in the room. The sole purpose of the reference speakers is to act as an "anchor" for the ratings. The reference speaker has a smooth frequency response in a frequency range that matches the test items and is ultimately excluded from the ratings. The reference speakers and the test items are listened to alternately and at the user's discretion. Switching between reference and test speakers is done "on the fly" with no audible effects other than the differences caused by the speakers. The difference for the attribute in question between the reference speaker and the test speaker is given as the response. This methodology prevents the result from being influenced by room acoustics when the test loudspeakers with different locations in the room are directly compared.

Blind test. The speakers hide behind the thin fabric and the test participant can only assess the sound.

A qualified objective listening test for characterising sound requires six different ingredientsand is performed with trained listeners:

1. A listening room. SenseLabs listening room with internal dimensions of L: 7,8m, W: 4,7m and H: 2,6m and built according to the box-in-box principle to provide good sound insulation. The inner room is vibration isolated from the outer room. The room meets the standards EBU Tech 3276 and ITU-R BS.1116-1 for multi-channel loudspeaker listening test. This includes very low background noise and a frequency-independent reverberation time (so there is no colouration of the sound) of about 0.2 seconds above 160 Hz and slightly increasing towards lower frequencies.

2. A listening panel. SenseLab has a trained panel that is used for the listening test on the Q113. It consists of 30 selected and trained listeners whose performance we monitor continuously. They include acoustics students, sound engineers, musicians and others who work professionally with sound, but also "ordinary people" with good ears. This test is conducted with 15 of these and we supplement with 10 interested listeners from the Engineer's readership.

3. Equipment and software for blind testing. Our listening tests are run automatically with software that presents audio samples in random order to all listeners. The listeners' ratings are made on a user interface which, in this case, is projected onto a screen from a projector outside the room. This avoids acoustic reflections from a screen physically located in the room.

4. Suitable music examples. The sound examples should challenge the characteristics/attributes being assessed. For example, you cannot judge the bass response of a piece of music with a piccolo, etc. We have had to limit ourselves but have chosen for this test: Artist: Jennifer Warnes, Album: Famous Blue Raincoat, Track: Bird on a Wire, Artist: Karina Gauvin, Luc Beausejour, Album: Bach: Little Notebook for Anna-Magdelena, Track: Willst du dein Hertz mir schenken and Artist: Paula Cole, Album: This Fire, Track: Tiger

5. Relevant, objective and well-defined attributes. The characteristics of the loudspeakers must be assessed in different dimensions. There are many ways to describe the sound, but for this kind of test the chosen descriptors or attributes must meet different requirements, e.g. a) Must be understood and used consistently by the assessors b) Demonstrate differences between test items c) Limited overlap between attributes d) Be one-dimensional, not combinations of terms e) Have relevance to preference f) Must be communicable without "slang" SenseLab is developing a general set of attributes for loudspeaker testing. The work is not complete, so we have chosen attributes based on our experience and test results rather than from a well-documented set of attributes.

6. Statistical knowledge and tools - so that the basis and significance of the conclusions drawn from the test are in place and can be justified.

Behind the fabric. The speakers are replaced while the test participant takes a break outside the listening room.

In addition to a characterisation of the sound, we also perform a preference test This test is subjective and should ideally be performed with consumers selected from the relevant customer segment. Usually more listeners are used than in the above tests and clusters will often appear which can be found in a statistical cluster analysis. It is popular to say that some people prefer red apples and others prefer green apples. We should expect that there may be different tonal preferences, for example, but the number of listeners will probably be too small for us to unravel this systematically. The test is well underway, and will be run over the next few weeks. A description of the attributes on which the speakers are judged will follow next week.