Listening test of drivers

[ginetto61]

Hello !

Lately I am more and more interested in speakers issues.
So I start reading about speakers design.
I found at least one designer who selects the drivers for his projects by listening to them mounted on a baffle
Or better, no matter how they measure if they sound bad to him he rejects them
My question is do you select the drivers for your creations by ear as well ?
The reason for my question is that in some case bad measuring speakers are generally well judged on the basis of sound alone.
For instance, they have a very non linear freq response and then people/reviewers like them
Thank you very much
Kind regards,
gino

[CBRworm]

I know that I have speakers that measure poorly that I really like to listen to, and I have speakers that measure very well that sit in the closet.

[ginetto61]

Originally Posted by CBRworm

I know that I have speakers that measure poorly that I really like to listen to, and I have speakers that measure very well that sit in the closet.

Thank you very much for your kind and helpful reply
And this is very unfortunate ... how could it be easier to select them just on the basis of their specs.
As I think that a complete speaker cannot be better than its components now the problem is to find a list of good sounding drivers
My goal would be a 2 way speakers with enough bandwidth
I already have good tweeters I could use ... a pair of dynaudio D28
The woofer is the real problem
I understand that good ones like Scanspeak come quite expensive
Do you have any suggestion about the best methodology for selecting drivers ?
Thank you sincerely
Kind regards,
gino

[ScottG]

Originally Posted by ginetto61

My question is do you select the drivers for your creations by ear as well ?

I don't think anyone initially selects driver by ear.. unless they actually make them.

For the most part it's a combination of TS param.s for a given design with heavy emphasis on linearity (linear and non linear effects), and sometimes dispersion character (which is basically a subset of linearity). THEN you'll sometimes add-in recommendations by others on the more subjective: "does it sound good or not". Finally it becomes a matter of design integration and overall sound quality (..though for some it's more a matter of integration and overall linearity).

Frankly the "recommendations by others" and more specifically the final "overall sound quality" are pretty much the major differences between so-called subjective and objective behavior in speaker design - and even then you'll often have those who consider themselves more objective increasingly perform subjective testing (..though perhaps under more controlled conditions).

The real "kicker" to the issue of measured linearity emphasized OVER subjective evaluation is that:

1. NON-linear effects are often difficult to perceive with *music* in a domestic context, and
2. A 1 or even 2 meter reference linear (freq. behavior) for measured response probably won't be anywhere near as linear in a typical listening environment. ..and of course it's usually less linear when you factor in dispersion and in-room averaging (at higher freq.s).

[chris661]

Before looking for drivers, find a shop that will supply you with drivers at a good price, and make sure they have good customer service.

To select woofers, make sure you model the bass response. There's no point having a woofer if you get no bass out of it.

When you have a shortlist, read internet reviews, look on places like diyAudio for people that have used the drivers previously. Ask if there's anything to watch out for.

Very few drivers (of a reasonable price) sound rubbish. Some sound worse that others in different applications. A driver with low Qts will do well in a ported box, but it will hate OB. This does not mean it sounds bad altogether, it just means you're using it wrong.

[jomatecc]

Some of the most realistic sounding speakers Ive heard had frequency response that have dips in the frequency response that would be unacceptable to most designers I know, and this is from both DIY's and Commercial designs. For some measurements will always affect their final judgement no matter the sound.
As always there is more to the story. My 2 cents.

[ginetto61]

Thank you very much for all the valuable advice.
I am worried.
I understand that what measures bad can sound good
And maybe even the contrary is possible ?
As you suggest to rely on experiences of other people is the best way to gather information and avoid mistakes.

Thank you very much again and kind regards,
gino

[chris661]

Just thought of a good example - think of, say, the frequency response of a person. You'll have a fundamental at 60-100Hz, then nothing until a few kHz. That doesn't mean they can't sing.

[5th element]

Originally Posted by chris661

Just thought of a good example - think of, say, the frequency response of a person. You'll have a fundamental at 60-100Hz, then nothing until a few kHz. That doesn't mean they can't sing.

Over the range of a tenor, the fundamentals lie between ~125hz to ~500hz, with harmonics of several orders occurring at intervals after the fundamental.

A soprano ranges from ~250hz to ~1000hz with the high coloratura pieces going up to around 1500hz.

A bass extends down to around ~70hz.

So to cover the fundamentals of all the voice types you're looking at ~70-1500hz. With the harmonics extending to around 10000hz for the top notes.

Loudspeakers that measure badly can sound good for a couple of reasons.

1) The 'bad' results aren't actually as much of a problem as you'd initially think. Narrow dips in the frequency response usually wont annoy you in any major way for example. Also you could have high distortion at low frequencies, this doesn't typically sound that bad and you only realise what you're missing when you hear something that does it right.

2) People can be very tolerant of certain flaws in a loudspeakers performance. For example a loudspeaker with very little bass output can be very enjoyable to listen to if it does everything else really well. Also you can have an obvious honk in the sound, or a colouration over a small frequency range that you always hear with certain voices, but because everything else is great you don't mind it.

You cannot get around a drivers limitations if you're trying to work to a specific specification and the end result won't ever exceed what the drivers are capable of. However you won't get fantastic results even if you use the best drivers in the world, if your crossover is duff. Likewise you cannot listen to drivers without any crossover and make a general assumption about whether they are good or not.

For example you could place a scan speak driver, say the scan speak 15W/8530 in a box and listen to it without an xover and it would sound pretty grim. It'd probably be a little shouty, no real treble and sound quite forward.

On the other hand you could place another cheapo drive unit with a much higher voice coil inductance in the box and find it sounds far more pleasing. Why? The high inductance rolls off the top end helping to reduce the effects of a mid bass playing too high. Not only this but the high inductance helps to shape the response such that baffle step is somewhat compensated for so the sound isn't as thin any more.

You could also try listening to a Seas Excel W15CY001 in the same way. It would share similar characteristics as the Scan speak driver, maybe a bit less shouty perhaps, but it would 'hiss' at you when run full range due to the massive peak in the response at ~8khz. This doesn't typically sound 'bad' as per say and can give a false impression of actually have some real treble, but neutral it most certainly isn't.

The 15W and the W15 are both excellent drivers when used properly, but could possibly sound worse then a cheap driver when compared full range.

[ginetto61]

Thank you very much for your very interesting reply
I accept without problems that speakers that measure badly can give a good sound
But someone told me that also the opposite can be true .. this is so much more difficult to accept to me
Is it possible ?
If so I could think that measures are not only unuseful to assess the sound quality of a speaker but even misleading

Nevertheless I extracted some passages from your reply

the end result won't ever exceed what the drivers are capable of.
However you won't get fantastic results even if you use the best drivers in the world, if your crossover is duff

This is perfectly clear to me and tells about the paramount importance of the drivers selection procedure
Of couse a bad crossover network vanishes every effort to make good things

Likewise you cannot listen to drivers without any crossover and make a general assumption about whether they are good or not

This is what Mr Lynn Olson, a well known speaker designer, states. He mounts the drivers on a standard baffle and listens to them without any filtering crossover, for peakes in response, timbral accuracy, transient response and so on.
His point is that if the driver has flaws they should be apparent.
He does so even with tweeters, filtered with a quality cap in series
He tells that the ear must be educated but it is possible to ear the overall performance of a driver
Then he goes on listing a number of good and less good drivers
It would be greatly useful to have a recommended drivers list, for each type of drivers of course, cones, domes, ribbons, ecc.

But the very problem for me is that good measuring drivers can sound bad (flat, boring, opaque, …)
Because it implies that traditional measures are indeed useless
Nevertheless I can easily accept the contrary. A lot of cases confirm this situation
I hope that one day a set of measures that correlate well with listening results will be available
Thanks very much again and kind regards,
gino