I don't know what I don't know

[iScream]

Hey guys,

First post here. I've been lurking around for a short time trying to learn stuff but there's a huge amount of info to try to absorb and I'm having a hard time finding what I'm looking for.

Over the last 10 years I've built several subwoofers, from a 12" Adire Audio Shiva with a plate amp to an IB using 4 Adire Audio Tempest subs mounted on a baffle I created for the attic, to a custom fiberglass enclosure for the trunk of my current car.

Also in my current car I built fiberglass pods for each door for Scanspeak 10F mids and 3004 tweeters. I run them fully active through a RF 3SIXTY.3 DSP along with a 8" Dayton midbass in each door. I used WinISD to model the door pods and subwoofer enclosure for the specific drivers, though I didn't go beyond finding the volume to give me a Q and FR that looked good.

I've got saws and routers with jasper jigs and I know how to use them. And I have some electronics background but I've forgotten most of it over the last 20 years. I've got a calibrated mic and REW and basic understanding of how to flatten peaks with PEQ.

So I feel like I'm ready to tackle a nice 3 way tower for the house. But I also feel like I know just enough to be dangerous, or more likely waste a bunch of money.

I've had an audio crush on the Scanspeak 12MU ever since I saw some measurements and comments over on diymobileaudio, even though I've never actually heard it. Part of me wants to buy a set of 12MU, a set of 22W Revelators and D3004/6040-00 Beryllium tweeters, find the volumes for .707 Q on the mids and woofers and build tall/narrow/deep boxes for them. Then use a DSP to make everything work together.

But I won't before learning more.

As an example, I understand what phase and group delay are but I don't know how to make sure I account for it between the different drivers so they will blend well. Do I need to consider this when choosing my target Q for each driver's enclosure?

I've seen baffle step discussed in several threads around here but don't have a good understanding at all yet. So, as the subject says, I just don't know what all I don't know.

I've got a copy of Loudspeaker Design Cookbook that I've started working through. But what do I need to do to be sure I can come up with something that will truly sound good? Do I need to spend the next 6 months studying before even attempting something? Can I spend a couple hundred bucks on some software to cut the learning curve down significantly?

I definitely want to go active by the way, partly for the flexibility and party because I found that I enjoyed tweaking the DSP in my car. I know a smart first step would probably be to take a proven published DIY design and just try to replace the passive crossover design with an active implementation. But, honestly, that seems a little boring.

Sorry this is so long but any advice on how to proceed will be greatly appreciated.

Thanks,
Chris

 

[Mooly]

Welcome to diyAudio Chris

Speaker design isn't my field of expertise I'm afraid... and so I would perhaps suggest breaking down your questions one at a time and asking in the appropriate speaker forums.

This section is more just for intros as much as anything.

 

[eriksquires]

Hi Chris!

I cheat around all of those issues.

First, I use OmniMic and DATS V2. you can get free tools like Room EQ Wizard but having the impedance/driver measurement is also important.

Calculate your cabinet or use the recommended volume's from Madisound.

Build your box, stick the drivers in. Measure the delay and in cabinet FR and impedance.

Plug the information into XSim and you will be able to create a crossover that is exactly right the first time.

This process gets around the box simulation for baffle step, and driver delay with real data.

Also, as I've mentioned elsewhere, if you want good sound, put aside part of your budget for bass traps and room acoustics. It's better to get a great sounding 2 way than an expensive 3 way that sounds like crap. Beset of course is a 3 way with good room acoustic treatment. Then you get the high dynamic range, deep and smooth base.

Best,


Erik

 

[ra7]

Employ your time in improving yourself by other men's writings, so that you shall gain easily what others have labored hard for.
-- Socrates, philosopher

In other words, read, read and read. And learn of course. Be a sponge. Absorb everything.

Loudspeaker design is a tough field. It is a confluence of many disciplines, but mainly acoustics, mechanical and electrical engineering. Doesn't mean you need an education in one of these fields but a rational, reasoning bend of mind helps.

I would start with less expensive drivers so that your learning and experimenting won't cost much. You can do wonderful things with relatively cheap drivers if you know what you're doing. And you can create disasters with excellent drivers. Also, having a quality reference helps. I would look at Linkwitz or Jeff Bagby's designs for a reference. Not very expensive. Once you have a speaker that measures and sounds well, you can compare it against your own creations.

 

[adason]

...says someone who never wrote anything.

 

[Cal Weldon]

Socrates had ghost writers?

 

[ra7]

...says someone who never wrote anything.

Regardless of who said it, it is a good idea... me thinks!

 

[eriksquires]

Going with expensive drivers for your first time is risky. I did it but I had prior training and took the plunge to getting good measuring tools first.

I agree overall that using inexpensive drivers is the best way to start. You learn as much asbout your own tastes as you do about speaker making. Especially about whether you really need a 3-way.

I'd start with doing a 2-way with 5 or 6-1/2" woofers. Say a Vifa ring radiator and Peerless woofer, or SB Acoustics. Total driver cost around $60-$100. Of course others have good recommendations as well.

Another starter project is to take a pair of speakers you currently have and analyze them. Learn to measure the individual drivers, put them in XSim and evaluate the crossover. Few want to start here, but it's a great experience. You'll go through about 2/3rds of the steps you would in making your own minus the sawdust and expense.

Best,


Erik

 

[system7]

Without knowing ANYTHING, you might guess these 3 drivers play together nicely:

ScanSpeak-3W-Discovery

The 12MU is very similar to the 10F, just 2dB louder, so is also doable.

You'll get it working.

 

[krivium]

One at a time they say, and must read... ok so let's read about bsc first:


https://trueaudio.com/st_diff1.htm

Baffle Step Compensation

 

[iScream]

Hi Chris!

I cheat around all of those issues.

First, I use OmniMic and DATS V2. you can get free tools like Room EQ Wizard but having the impedance/driver measurement is also important.

Calculate your cabinet or use the recommended volume's from Madisound.

Build your box, stick the drivers in. Measure the delay and in cabinet FR and impedance.

Plug the information into XSim and you will be able to create a crossover that is exactly right the first time.

This process gets around the box simulation for baffle step, and driver delay with real data.

Also, as I've mentioned elsewhere, if you want good sound, put aside part of your budget for bass traps and room acoustics. It's better to get a great sounding 2 way than an expensive 3 way that sounds like crap. Beset of course is a 3 way with good room acoustic treatment. Then you get the high dynamic range, deep and smooth base.

Best,


Erik

Thanks. How important do you feel the driver measurements are with a DSP? Is it mainly useful for designing a passive crossover where you need to buy components or is it still worth it to get the frequencies and slopes you want for creating active filters?

-Chris

 

[iScream]

Employ your time in improving yourself by other men's writings, so that you shall gain easily what others have labored hard for.
-- Socrates, philosopher

In other words, read, read and read. And learn of course. Be a sponge. Absorb everything.

Loudspeaker design is a tough field. It is a confluence of many disciplines, but mainly acoustics, mechanical and electrical engineering. Doesn't mean you need an education in one of these fields but a rational, reasoning bend of mind helps.

I would start with less expensive drivers so that your learning and experimenting won't cost much. You can do wonderful things with relatively cheap drivers if you know what you're doing. And you can create disasters with excellent drivers. Also, having a quality reference helps. I would look at Linkwitz or Jeff Bagby's designs for a reference. Not very expensive. Once you have a speaker that measures and sounds well, you can compare it against your own creations.

But what else was there to do back when Socrates was around?

I know you're giving good advice here with using lower cost drivers for my first try at something.

As far as a reference goes, I've got a few different sets of pretty decent headphones with dedicated DAC/Amps. Doesn't help with imaging and such but I would be very happy if I can duplicate the detail and overall sound reasonably well.

I will most likely take the advice on building something more affordable to start though.

Thanks,
Chris

 

[iScream]

Without knowing ANYTHING, you might guess these 3 drivers play together nicely:

ScanSpeak-3W-Discovery

The 12MU is very similar to the 10F, just 2dB louder, so is also doable.

You'll get it working.

I suspect the drivers I mentioned would play pretty nicely together as well. But it is a big chunk of change to drop without being sure. I would love to build the Scanspeak B741 kit from Madisound but make it active. But surely I can get 90 percent of the performance for half the price, or maybe even less.

One thing I have going for me is that these speakers will be used in the bonus room where I have my home office setup. And I have a pretty tolerant wife who won't complain about bare MDF showing up here while I experiment.

Thanks,
Chris

 

[iScream]

One at a time they say, and must read... ok so let's read about bsc first:


https://trueaudio.com/st_diff1.htm

Baffle Step Compensation

Thank you for the links. I think I get it well enough now. It seems this is a case where going active and taking real measurements at my listening position would be a big advantage.

But it also seems that trying to make it as smooth as possible in the first place by considering how the driver is mounted in relation to the baffle edges would make the measuring and PEQ config much easier. Especially since I pretty much only try to cut peaks, not boost valleys in response.

So is this something that can be modeled in software affordable to a hobbyist? Or do I have to just offset my drivers some and experiment?

I got the impression that one 7" driver mounted in the same spot as another model 7" driver would end up with very similar ripples from this? Is that a safe assumption?

Thanks,
Chris

 

[eriksquires]

Thanks. How important do you feel the driver measurements are with a DSP? Is it mainly useful for designing a passive crossover where you need to buy components or is it still worth it to get the frequencies and slopes you want for creating active filters?

-Chris

Assuming you have the correct box size and shape, all you need is the FR (or impulse response) for DSP based crossovers but having that is critical to the DSP filter design process. Some systems like Hypex prefer impulse response, others like FR I think. It's worth downloading the software to play with it before committing to the DSP if you can.

They don't just spit out filters though. The process is iterative. You examine the FR and phase in real time as you add filter sections. For the output of that to be accurate you must start with accurate measurements.

Technically, you CAN just create a DSP filter that outputs a filtered signal that is of a specific theoretical type. Say 2nd order LR high-pass that is -6db at 2800 Hz without any measurements at all, and electrically it will be correct.

Does this make sense?

The problem is that the output of a speaker system is the boundaries + cabinet + drivers + electrical signal and that is what a crossover designer must take into account, so it's impossible to overcome these multiple simultaneous effects just by going with a DSP and picking a simple crossover.


Best,


Erik

 

[krivium]

It seems this is a case where going active and taking real measurements at my listening position would be a big advantage.

Yes you are right and this is the way to do it right. But if you have a nice pair of headphone that you listen regularly (that you know it's sound on tracks you know well) you can adjust by ear too (to adjust BSC)! You will probably be astonished by results, our brain is relatively accurate! But be carefull in doing that that level should be matched between your headphone and speakers, and you should keep a reading of Fletcher Munson curves with you (another set of technical stuff to study! ).

Especially since I pretty much only try to cut peaks, not boost valleys in response.

Right. That's one of the biggest secret you just discovered about our brain and eq: we are less sensitive to cut than to boost. No need to take care of all the valley and peaks with BSC, what is really important is the shelving either to boost the low end or to attenuate/cut the high (see the lats graph on rod elliott article). Attenuate as another benefit: you gain HEADROOM.

So is this something that can be modeled in software affordable to a hobbyist? Or do I have to just offset my drivers some and experiment?

From what you wrote i think you are talking about baffle edge diffraction rather than BSC. This is related but not the same problem. Yes you can simulate it (it can be important for tweeter location) and you can find freeware or shareware to do it. Try googling frd consortium. For Bsc this is only related to the shortest dimension of facia of speaker (usually width). EG: box 70cm high/ 50cm width. Which freq for shelf eq? 115/0.5m: 230hz. You must center of shelf eq at 230hz and take measurements or adjust by ear...

I got the impression that one 7" driver mounted in the same spot as another model 7" driver would end up with very similar ripples from this? Is that a safe assumption?

Well yes and no... It all depend from frequency and some physical layout. Try to see a D'apollito (also know as m-t-m, vertically aligned drivers with tweeter in center of the woofers) for exemple. You must know too that some rounding of edge or truncating can have a positive effect on baffle edge diffraction. But i repeat this is an other subject than baffle step correction.

 

[system7]

I did stumble across something with the SEAS Idunn Kit.

SEAS offer the DSP filter file as a download!

It works with the Hypex AS2.100.

Interestingly Hypex's demonstration speaker is a SEAS Idunn with an auxiliary Bass Radiator rather than a simple reflex tube. They must be collaborating with SEAS on this. You would guess the SEAS Bragi wouldn't be hard either.

 

[iScream]

Assuming you have the correct box size and shape, all you need is the FR (or impulse response) for DSP based crossovers but having that is critical to the DSP filter design process. Some systems like Hypex prefer impulse response, others like FR I think. It's worth downloading the software to play with it before committing to the DSP if you can.

They don't just spit out filters though. The process is iterative. You examine the FR and phase in real time as you add filter sections. For the output of that to be accurate you must start with accurate measurements.

Technically, you CAN just create a DSP filter that outputs a filtered signal that is of a specific theoretical type. Say 2nd order LR high-pass that is -6db at 2800 Hz without any measurements at all, and electrically it will be correct.

Does this make sense?

The problem is that the output of a speaker system is the boundaries + cabinet + drivers + electrical signal and that is what a crossover designer must take into account, so it's impossible to overcome these multiple simultaneous effects just by going with a DSP and picking a simple crossover.

Erik

I think I get it well enough on creating active filters. My first experience was with the IB sub I built for my DIY home theater in my previous house. It was pretty crude as I just played low frequency test tones and measured each with a simple SPL meter at my listening position. I had to choose a few peaks I could smooth with the limited number of PEQ filters available in a Behringer Feedback Destroyer. Even with that limited use of DSP that IB sub was incredible, both in shear output and in quality.

The DSP I run for the 3 way setup in each door of my car allows 30 really flexible PEQ filters per individual channel. For that I enabled one individual driver at a time and played frequency sweeps with REW capturing the response. I don't remember what fraction of an octave I had it set to but it was at least 1/24 and I think I actually went more granular than that. Then I manually created PEQ filters with center frequencies, levels and slopes that matched the shape of peaks as closely as I could. After a couple passes on each individual driver I ran through again with both sides enabled then with everything on and kind of averaged things together.

For the crossover points I just looked at the response graph predicted by WinISD for my drivers in the enclosures I built and picked frequencies that looked like good spots to transition. I did try to avoid beaming when choosing the starting crossover points but what I didn't consider through any of the process was phase relationships or decay at any particular frequencies. Other than flipping the phase of my sub 180 degrees and playing with the delay by ear to get it blended a little better that is.

-Chris

 

[eriksquires]

Steve, Master of crossovers and Google strikes again!

Cool find.

Erik

 

[iScream]

Yes you are right and this is the way to do it right. But if you have a nice pair of headphone that you listen regularly (that you know it's sound on tracks you know well) you can adjust by ear too (to adjust BSC)! You will probably be astonished by results, our brain is relatively accurate! But be carefull in doing that that level should be matched between your headphone and speakers, and you should keep a reading of Fletcher Munson curves with you (another set of technical stuff to study! ).



Right. That's one of the biggest secret you just discovered about our brain and eq: we are less sensitive to cut than to boost. No need to take care of all the valley and peaks with BSC, what is really important is the shelving either to boost the low end or to attenuate/cut the high (see the lats graph on rod elliott article). Attenuate as another benefit: you gain HEADROOM.



From what you wrote i think you are talking about baffle edge diffraction rather than BSC. This is related but not the same problem. Yes you can simulate it (it can be important for tweeter location) and you can find freeware or shareware to do it. Try googling frd consortium. For Bsc this is only related to the shortest dimension of facia of speaker (usually width). EG: box 70cm high/ 50cm width. Which freq for shelf eq? 115/0.5m: 230hz. You must center of shelf eq at 230hz and take measurements or adjust by ear...



Well yes and no... It all depend from frequency and some physical layout. Try to see a D'apollito (also know as m-t-m, vertically aligned drivers with tweeter in center of the woofers) for exemple. You must know too that some rounding of edge or truncating can have a positive effect on baffle edge diffraction. But i repeat this is an other subject than baffle step correction.

I get you on the difference between baffle step and edge diffraction, I think. Both were discussed in the link you provided so I guess I thought they were commonly combined together.

So, basically, baffle step is just the energy from your speaker being spread around and behind the enclosure as wavelengths increase. So less energy is directed forward as the frequency decreases, reducing the SPL that actually makes it to your ears while sitting in front of the speaker.

Baffle edge diffraction is a disturbance of the sound waves caused by interaction with the shape of the baffle and sharpness of the edges. I would have guessed that it doesn't come into play until the wavelengths get long enough to start wrapping around the enclosure kinda like baffle step. So the tweeter thing confuses me a little.

Thanks,
Chris