Quality control of DIY designs and other speakers

[Jed]

The purpose of this thread is to review the idea of how difficult it is to maintain precision with completed DIY designs, or any speaker for that matter, over a long period of time. A lot of the designs on the web are getting pretty old now, including even some of the Mission Accomplished speakers. What happens is product tolerances change significantly over time- sometimes from batch to batch; including resistors, inductors, and caps. I recently tested my friend's completed DIY project based on a DIY design and the final measurements were a bit different than the original design. Since I have measuring capabilities, it was no problem correcting the problem, which ended up being a slightly hot tweeter. Clearly, this was the result of production tolerance of the components used. Some DIY designers offer different L-pad variances to deal with this issue, but as a perfectionist, there are many limitations of giving someone a schematic and hoping that their response will measure like the "reference" FR plot.

So, I guess I'll just have to relax and accept that there is no perfect for everyone in DIY if you are using someone else's design. For example, if one is using electrolytic caps with a fairly large value tolerance difference, that can alter the final response quite a bit too. Some tweeters, even though they might be the same model have slightly different FR (see my vifa D26 thread). I've noticed with with the D26 in particular. I'm sure there are many other tweeters that show the same issue. Someone once posted that he's tested a bunch of RS28s, some with flat top end... some without it. So, combine the diffences in resistance from someones choice to use a 18gauge instead of a specified 19-20 gauge, and in conjunction with a slight hotter tweeter than the original, and we might have +/- 3 db differences over or under the references model. This bugs me to tell you the truth, but as long as DIYers are aware that they might not get the exact thing that is posted... and rely on their ears to judge the balance of the FR... I guess that's all one could hope for.

Ideally, a designer of a particular DIY speaker should check the design with new drivers periodically. This is a rather expensive endevour to ensure the end user is getting exactly what is shown in the thread, but it hardly makes much practical sense from a financial standpoint, given the designer probably already has the completed speaker from years past, in stock.

So, just some thoughts I've been having as I look on some of my designs that are going on over a year old now.

cheers

Jed

[tktran]

You make a good point Jed, and I guess these are some of the downsides of building someone else's design without measurement equipment.

Without measurements one can never know.

But as I builder I try not to worry too much. It's just a fun hobby.

Now, having said that, even if I start changing cabinet dimensions, substitute different deadening/dampening materials, using different AWG coils or specced caps, accidentally use wrong resistors, or wiring the crossover out of phase, I'm probably still doing better than some commercial speakers.

Have you seen some of the measurements of commercial speakers lately? A quick look at Stereophile or Soundstage reveals more bad ones than good ones. Some of them are so horrible, maybe the companies should blame poor tolerances. At least that's a better excuse than poor crossover design!

[Jed]

If I was charging what those companies charge for a speaker you can be sure it would be measured before it was sent out, so that it would match the reference FR. Yeah... I've seen some really horrible measurements from the manufacturers... actually most are horrible including some from Wilson and other big names.

[Curt C]

If I was charging what those companies charge for a speaker you can be sure it would be measured before it was sent out, so that it would match the reference FR. Yeah... I've seen some really horrible measurements from the manufacturers... actually most are horrible including some from Wilson and other big names.

There's your answer Jed!
Picture the worst case scenario of a recreation of one of your designs: :g> Driver parameters have changed by 10%, wrong inductors used by the builder, modifications to the enclosure, etc. I'll suggest your design would still measure and sound better than many comparable commercial offerings out there. :T

I too have been less than enamored whenever I've listened to the top-shelf speaker products. I've always been kind and suggested it was the room etc,, but in my experiences I've yet to hear a high-end design without obvious and generally unacceptable flaws.

Unfortunately, we designers by necessity listen to the speaker, and not to the music. It's a flaw I continuously have to overcome when I just want to enjoy the music....

C

[fjhuerta]

In my very own humble opinion... I wouldn't dare design a speaker from an available design without my measurement equipment. I don't know if it's the tolerances or slight differences between one product or the other, but I've always had to tweak them a bit.

It's a mistery to me...

[Jed]

In my very own humble opinion... I wouldn't dare design a speaker from an available design without my measurement equipment. I don't know if it's the tolerances or slight differences between one product or the other, but I've always had to tweak them a bit.

It's a mistery to me...

Yes, I had this exact experience recently when I used different types of capacitors and inductors of a different brands etc from the original.

For the best reliability I would recommend building EXACTLY the same speaker as the original using ALL specified parts, and even then there sometimes are slight changes in resistor values needed to match the reference response.

[TacoD]

Maybe it's a good idea to list the manufacture date of the stuff you use in the designs. So that people can buy the drivers second hand.

But I've to say that Eton/ Scanspeak/ Vifa/ Visaton (some models are made > 10 years and T/S is the same) is very consistent.

[Undefinition]

You could make yourself crazy

Jed,
I wish I only had to worry about variances in QC of components. In my experience with other people building my designs, it seems that builders like to substitute drivers, change cabinet dimensions, etc. "But it's not the same speaker," I think to myself. I often talk to the builder, insisting why I designed things the way I did, and chose the parts I did. But a lot of guys have their heart set on making some mod. I guess DIYers are just the type who like to put their own spin on things.

How to not get frustrated with it? Well, when I first got into DIY, I spent a lot of time studying Wayne J's designs, and noticed that people were altering cabinet dimensions, driver layouts, and even swapping out drivers without making XO changes!
Second, I have written quite a bit of music for video games, percussion ensembles, and also a few albums of electronic music on the side. Now I see (or people point out to me) that people are using my music in various places, and no one ever asked me for permission (or in the case of the Percussion Ensembles, never paid me for the score!). While a lot of people have suggested that I pursue legal action, I just don't have the time or interest. I have a day job, you know?

I'm not a religious man, but it conjures up the serenity prayer: especially the parts about "accept the things I cannot change ... and the wisdom to know the difference." Someimtes, it's just easier on the nerves to just step back and let people have their fun, even if it's not the way I would have done it.

-Paul

[jkrutke]

Yeah, QC of posted designs is a serious issue. The only way to be sure is to continuously test drivers as time goes by. I actually do that whenever I get a chance, and I get help from the community when I can - I have people borrow drivers to me for checkups.

I've pulled a few designs due to QC issues. Maybe QC is not the right term, but batch consistency issues is a better designation.

I just got in a complete set of new drivers for the ZDT3.5 tower, and noted they all looked good, except for the tweeter. The current batch of ND20FB's is about 1 dB higher sensitivity, 1.5 worst case. As such, I've added some level options and commentary to adjust.

Driver sensitivity is the number one issue. Breakup node shifting is probably a distant number 2 issue. With sensitivity, we can at least include level adjustment options in all our designs, something I make sure I do lately.

When I did the ZRT revelator tower design with Madisound, part of the deal was that they would check new batches for consistency, and I'd tweak the design as needed. I wish I had that with all my designs.

Over the years I've tested lots of Seas L18's and 27TBFCG's. Those are probably two of the most consistent drivers ever made. I never had to change anything, and the breakup never budged.

[Jed]

Which begs the question, How much are people willing to pay for a DIY design that ensures consistency?

It does look like Seas is one of the "best" in terms of consistency, yet around here the best values seem to get more attention. Not a bad thing but I think it comes at a price, just maybe not one with a dollar sign. I mean, lets just look at TB for a minute. Seems like they change a basket or a surround or a this and a that constantly. Luckily the last time I checked the TB W4-1337 everything was close within specs to the previous version. It's definitely not just the tweeters we have to be concerned about.

Next time I post a design I'll be sure to include different Lpad variations.

[Jed]

Someimtes, it's just easier on the nerves to just step back and let people have their fun, even if it's not the way I would have done it.

-Paul

I hear ya Paul. It would really upset me if someone stole my art/design and put their name on it though.

[fjhuerta]

Over the years I've tested lots of Seas L18's and 27TBFCG's. Those are probably two of the most consistent drivers ever made. I never had to change anything, and the breakup never budged.

I borrowed your 28TBFCG crossover design from the Bargain Aluminum MTM on my Monolith design... it measured *exactly* as it did in your graphs. This has never happened on any other design I've tried.

[oldbar]

I have a seas L18 driver which i bought early last year and it has been sitting on the shelf for awhile and it was made in 2004.
I emailed seas because I noticed there was small changes in the TS parameters.
They told me that my driver was an from an older batch and that they actually changed the membrane since then.

[jkrutke]

I have a seas L18 driver which i bought early last year and it has been sitting on the shelf for awhile and it was made in 2004.
I emailed seas because I noticed there was small changes in the TS parameters.
They told me that my driver was an from an older batch and that they actually changed the membrane since then.

It may be an L18, but it was likely an older H model rather then the new H1224. The model designation changed to H1224 when they redesigned the cone profile. That wasn't a consistency issue, it was an improvement. (and one that I wished for and welcomed) The older H1142 straight cone version was obsoleted, and vendors with slow moving stock were stuck with the old version. Nobody wanted the old version when the new one became available, myself included. Can I ask where you bought it? It wasn't Madisound I'm sure.

edit: oops, some H number confusion in there.

[oldbar]

I bought it in Australia from Aranmar Acoustics who use to be a seas distributor.
It has on the label H1224 2/04 which i assumed was the new model which you indicated was an improvement.
I actually emailed them at seas and asked and they emailed me telling me that the 2/04 was the older model.
Maybe they misunderstood my email I am not sure.I might email them again and see what they have to say.

[tktran]

Hi,

The changing designation numbers are confusing. I can see that even John may be a little muddled up with it.

The H1224 is the designation of revised L18RNX/P, which was what he used in the SEAS "All Metal System" 2-way.

The H1142 was the older L18RNX/P.
The H1085 was the L18RCY/P.

2004 was indeed the transition year for the L18RNX/P, so it makes things a little tricky for you.

One way to tell is from the curvilinear cone profile of the H1224, but that's not always easy to judge. Another way to tell the difference is the inspect the rubber surround, and the junction where it terminates at the cone.

See attached- the H1142, and H1224.
Take note of the H1142- there is more rubber where the surround meets the cone. (x marks the spot)

Hopefully your label is correct, and you have the newer one from Aranmar Acoustics.

regards,

Thanh "Trainspotter" Tran.
(who hasn't managed to finalise a project since 2005)

[Kevin Haskins]

This is a problem for all companies, let alone DIYers. Driver variability in manufacturing is one of the larger issues so if you sell something, you have to keep an eye on the inevitable variation in production runs. Most of this is due to soft parts. The motor is simple to keep in spec, voice coils are wound to 1% spec's. Making them the same batch to batch is easy. I've heard a lot of nightmare stories about soft parts, cones mainly, as production runs change the tolerances, thickness and sometimes even the paper stock used for paper cones has different properties.

Suspensions have large tolerances too but they tend to affect the final response in the application in a meaningful way.

Crossover parts are usually pretty tight unless you use electrolytics. MKP caps, inductors and resistors are all high tolerance parts so you don't see a lot of change.

I know of one company, that makes an active monitor that measures every finished unit on the production line and trims the crossover to keep them within a +/- 1dB window. That is about the only method of keeping 99.9% certainty that the product comes out as designed, irrespective of manufacturing tolerances.

[jkrutke]

The changing designation numbers are confusing. I can see that even John may be a little muddled up with it

Oops, yup. My design was for the H1224. Also known as "the good L18". The only one with a curved cone. Ok, going back to see if I can edit my mistake. I generally don't use the H numbers in Seas product lines and instead use the normal model numbers. I wonder why Seas needs both designations.

[jkrutke]

The motor is simple to keep in spec

Gap tolerances and centering - yes I agree, but I see companies having issues maintaining the same magnetic charge. I don't know enough about loudspeaker manufacturing to know why, (burnt out coils? motor pulled early?) but it's pretty easy to see in woofer T/S numbers and in tweeter sensitivity.

This tweeter for example has a deficient magnetic charge. The result is that it's Fs is the same, DCR is the same, but it's sensitivity is 5dB lower and there's now a high Qts bump on the low end. This is pretty common, particularly with low end drivers.



I've actually got a whole load of examples like this. More seem to be with woofers though.

[Kevin Haskins]

Gap tolerances and centering - yes I agree, but I see companies having issues maintaining the same magnetic charge. I don't know enough about loudspeaker manufacturing to know why, (burnt out coils? motor pulled early?) but it's pretty easy to see in woofer T/S numbers and in tweeter sensitivity.

This tweeter for example has a deficient magnetic charge. The result is that it's Fs is the same, DCR is the same, but it's sensitivity is 5dB lower and there's now a high Qts bump on the low end. This is pretty common, particularly with low end drivers.



I've actually got a whole load of examples like this. More seem to be with woofers though.

Yea... about +/- 5% is common. I've seen some spec's at +/- 10%. The grade of the steel, magnet and how the magnetizer is set are probably the culprit. Also, if the VC is offset and you have a parabolic BL curve the T/S measured BL is based upon an average measurement of the BL. Offset coils may put the coil slightly out of the parabolic hump of the BL curve and give you a lower than designed BL measurements from looking at the measured T/S parameters. Certainly should be less of an issue with a tweeter than a midbass or a woofer since you see such small excursions in a tweeter.

It is hard to see how they could have a 5dB difference without changing the part. It would have to be a gross error, not a simple manufacturing tolerance issue.

I don't monkey with the low-end parts because when your limited to a $5 part companies don't give a hoot about maintaining any kind of standard. That is fine for mass market items but my entire market is high-end, either 2-channel or home theater and it seems foolish to cut corners on transducers when people have thousands of dollars invested in the results.

[jkrutke]

Just a quick note, that tweeter was a "worst case" magnet charge issue. Most are not that bad but this one sure helps show what I mean.

Offset coils is a good point, probably more of a concern with tweeters.

$1K Scan Speak designs aside, I do favor the high value stuff. I just retest the drivers often and if it gets bad enough, I pull the design. (like the Silver Flute W14 woofer system I used to have posted) There were mild sensitivity changes, but the Qts was all over the place, wreaking havoc with proper tunings and enclosure sizes.