Speaker Quality Control During Manufacturing

[PristineSound]

I thought about starting this thread for a bit, but was looking for an existing thread, didn't find one. So here I am.

I want to start a discussion about speaker manufacturing and quality control. I would argue in the last 5 to 7 years, there are so many great speakers being made. Many are even publishing their own CEA2034 or are sending it to the likes of ASR, Erin, Audioholics and Soundstage knowing that they will measure it.

Many of these speakers comes back with jaw dropping measurements. . . and so many of us flock to buy them.

But here is something I don't think we talk enough about:

1. Each speaker that is made to the same exact design have fluctuations, sometimes 5+dB, this is largely due to the driver and crossover tolerance. While some driver manufacturers have very tight tolerance, but when you start to add multiple drivers and the crossover components together the end result can compound.
2. The speaker that the manufactures send to the reviewers, would of course have been vetted out by the manufacturer. So what you see is the vetted speaker, it is not always indicative of the actual speakers you get, again sometimes you can get as much as 5+dB variation from measurements that's published.
3. Some manufacturers will pair match the speakers, rarely from 20Hz to 20kHz, typically from one to few hundred Hz to 10kHz. The quality of pair matching isn't always consistent across different manufacturers, such manufacturers say they pair match, but if you take a high-resolution anechoic scan, it's poor-quality pair matching. But nevertheless, pair matching will help significantly with imaging.
4. The speakers with very high level of QC are match to a reference within a very tight tolerance, sometimes within .5dB. That is the high-end stuff and often times where a lot of cost goes into. My understanding is that pair matching with passive speakers are harder to do, hence driving up the cost even more. Speakers in this category are the likes of the KEF Blade and Reference models and I believe Revel Salon.

Open discussion. Do you think speaker to speaker fluctuation is a big problem? Are you willing to pay extra for tight tolerance pair matching or tolerance against a reference? My experience, after hearing the Neumann KH120-II, it's very hard for me to go back to anything that isn't matched to that tolerance level.

 

[Ktacos]

The speakers with very high level of QC are match to a reference within a very tight tolerance, sometimes within .5dB. That is the high-end stuff and often times where a lot of cost goes into.

Don't have to spend much to get tight tolerances. Most of the drivers I use are under $50 and a lot of them just measure like overlapping lines in REW. I recently measured four peerless FSL because one user posted distortion measurements showing very bad tolerances. I was skepitcal so I did mine own on four drivers. The drivers are very well matched. The scale is big because we mostly cared about distortion but I did trace these and throw them into REW and they match eceptionally well up to like 5k.

My understanding is that pair matching with passive speakers are harder to do

I'd say it's more that active has an advantage here but IME getting a good match between speakers passively is not difficult to do. I don't really feel it's something one should really worry about to be honest. It's kind of the driver manufacturers problem and if you're buying from reputable brands it SHOULD be ok but there are definitely lemons here and there from everyone. So far I've only had to exchange one driver for being a bit too far out and it was like a $15 woofer so big whoop. Out of the probably hundreds I've bought by now the rest were interchangable.

Crossover components seem hilariously insensitive to tolerances IME. I just buy whatever is cheap and at the value I need, sometimes mixing and matching brands between speakers and my impedance graphs always look like overlapping lines.

 

[egellings]

Placing two nearly identical speakers in different positions in the listening room may result in sonic differences due to variations in room interactions based on speaker position in the room.

 

[levimax]

Open discussion. Do you think speaker to speaker fluctuation is a big problem? Are you willing to pay extra for tight tolerance pair matching or tolerance against a reference? My experience, after hearing the Neumann KH120-II, it's very hard for me to go back to anything that isn't matched to that tolerance level.

If you build DIY speakers and measure each driver individually you get a good idea of how much variation there can be, like you mentioned up to 5 dB which is a lot.

I think "matching" is an under appreciated advantage of high quality active speakers like Neumann. They don't just match pairs but they claim every KH 310 (and I assume it is true for most of their line up) is matched to any other KH 310 past, present, or future. To try to do this with passive speakers borders on impossible at any cost. From Neumann's web site:

Production consistency due to individual hand alignment in final testing

• Any KH 310 is “pair matched” to any other KH 310

 

[Arindal]

Do you think speaker to speaker fluctuation is a big problem?

Depends on the application in question. In most of cases, fluctuation is rather narrow-banded, which is expected to mostly effect localization stability, or interaction with a waveguide/horn. For nearfield studio monitoring, this is definitely an issue.

sometimes you can get as much as 5+dB variation from measurements that's published.

+-5dB is actually a lot, but more important is the particularly band in which tolerances are occurring (1-4K is the most sensitive band in my understanding) and of course Q factor of the deviation. The broader, the worse.

Planars and AMTs in general tend to show broader deviations, so impact without any QC is expected to be more dramatic, but on the other hand easier to mitigate, if active EQ or some sort of level-adjustment in the passive crossover is available.

Are you willing to pay extra for tight tolerance pair matching or tolerance against a reference?

For nearfield monitoring, definitely yes. I remember the best of all solutions being every single unit of a speaker being measured after manufacturing, with individual reciprocal DSP curve being implemented. KSD were offering such service, not sure if that is still the case with all their affordable active monitors.

 

[kemmler3D]

I think pair matching has obvious benefits but demands a very different process on the factory floor. Speakers can be tested with a very quick sweep to verify performance within a given range, pass/ fail.

Matching drivers requires another testing station per driver and limits production to how quickly you can make matched pairs. Alternatively you can have your driver supplier do that, but it won't be for free. Either way there's a noticeable cost increment.

Onboard anechoic EQ makes the most sense to me, as you could very quickly generate and upload a correction for each speaker as it comes off the line.

Overall I think it only makes sense as a manufacturing process for very expensive passives, and DSP speakers.

If we want pair matching for cheaper speakers, we can do gated measurements at home and do the correction ourselves with PEQ.

 

[Speedskater]

In the old 'Newsgroups' days, Dick Pierce would somtimes write about quality control problems.

CartChunk - Welcome

Audio Topics

Dick Pierce Anecdotes

groups.google.com

 

[PristineSound]

I think "matching" is an under appreciated advantage of high quality active speakers like Neumann.

I am of the same opinion. On Neumann's website, they attributed their "razor sharp" imaging to their reference match; I believe it's within .5dB from 20Hz to 20kHz of the reference, if I am not mistaken. The imaging is on my KH120 is uncanny compared to any speaker that I have heard.

 

[Ageve]

Some of the speakers I've measured:

(Ignore the small wiggles. I measured F228 in my living room, without moving them away from walls)

 

[Audionaut]

Each speaker that is made to the same exact design have fluctuations

I hope this isn’t really an issue with mass-produced speakers.
I’ve been building my own speakers for decades, so I can only speak from experience with individual drivers.

The biggest variation I encountered was a few years ago with an ESS AMT driver. The two units were completely different, and I sent them back. Fortunately, the German distributor now only offers drivers that are matched in pairs. With established companies, differences hardly have any effect anymore.

Another factor that can lead to very different results is the operating condition of the speakers. A speaker that you just plug in after two weeks in a cold room will measure differently than a speaker that has already been playing for a few hours at a midsummer party.