Designing a speaker with an iphone

[RayW]

Based on a thread started by someone new to speaker geeking, I started pondering whether I could design a speaker using my iphone and free apps. This thread is the design process of the IP1 speakers.

For any design you have to start somewhere. Generally this means evaluating what you want from the speaker. Low budget? High efficiency? Lots of bass? Some combination of these? This is something you'll have to sort out yourself. Figuring this out will let you choose appropriate drivers.

For the IP1 I had a head start. I have always been fascinated by overachieving small speakers and had some drivers on hand that I had been meaning to try out. I suppose if I had to put a theme to the design it would be a small, low budget monitor type speaker. The drivers involved are the Dayton ND105-8 mid/woofer and ND20FA-6 tweeter. I have used the ND90 and was impressed by it so I wanted to see what the 105 could do. The tweeter is the same one used in the Mindignias and Incentria (search for the Indignia thread) and know it's a good tweeter for the price. It appears that retail price for this box o' drivers would be around $90.

Having the drivers picked out we can start designing. The first thing to do is figure out the box. Using the phone means making some leaps of faith. The first of which is that the published specs for the drivers are reasonably close to reality. There are ways to measure T/S parameters using a DMM and there are apps for the iphone that will allow you to measure impedance and extract T/S parameters from it but for this exercise we're going to give Parts Express the benefit of the doubt.

I tend to like ported boxes when using small woofers like this so using the published specs for the ND105-8 and any one of the gaggle of box calculators available on the net (actually I cheated on this one as I have BassBox6 and I am very much used to it), I settled on a box of .2ft^3 (5.75L) tuned to 50Hz using a 1.25" x 5.5" port. It gives a bass boost up to about 100Hz that should help give the illusion of deeper bass from this little woofer. This should result in an F3 of 48Hz and F6 (my measure of choice for usable bass in a ported design) of around 40Hz.

Putting one of these together and mounting drivers looks like this:

 

[RayW]

Next we need to start working on a crossover. For this we're going to need a few things. First up, is impedance information for the drivers. As I said before, we're going to take PE's word on the parameters and that will include Re (aka DCR). In order to make the crossover easier to deal with we will use a Zobel network (impedance equalization) on the woofer to flatten the impedance curve so that we can use the Re as a ballpark figure for impedance. This means we won't have to deal with a rising impedance in filter calculations. There are a bunch of online calculators for these networks. I used this one: http://www.diyaudioandvideo.com/Calculator/ImpedEqual/
Using the published numbers of Re=7.6Ω and 1.37mH gives this result:

Next up we will need frequency response plots for both drivers in the box. This is where test equipment can get expensive. I cut my teeth on LMS/LEAP which goes for a couple thousand dollars and then moved to TEF20 which was about $5k for the base system and each measurement package was an added expense. For this exercise we're going to be using a freebie RTA program. There are several out there and I chose one called "RTA Lite". The full app is here: https://itunes.apple.com/us/app/rta/id298839433?mt=8. There are also several free FFT apps that will work but things were easier to see on the RTA so I went with it for this design.

**a note on the response of an iphone - we are not going to get laboratory grade results out of a phone. It just ain't gonna happen. In the following measurements you will note that the upper 2 bands (above 16kHz) are always rolled off. This is a result of the phone, not the tweeter. Likewise the bass rolls off around 200Hz. The speakers have more bass than that, the phone just doesn't show it. The range we're worried about here is the crossover region where the phone is most sensitive. The phone's response will also not be ruler flat but we're going to ignore that here and just run with what we've got.

To get response curves I created 2 files in Audacity (free download). One is 1 minute of 20-20kHz pink noise and the other is the same file run through a 1.5kHz 2nd order high pass filter. The latter will be used to test the tweeter and the HP filter will protect it from low frequencies. I dumped the 2 files to my ipod and used a Lepai T-amp for power. I put the speaker on a stool, hooked up the woofer, ran the pink noise file, and held the phone out about a meter in front of the speaker. The spacing isn't critical as we're just getting an idea of the shape of the curve. Keeping the phone steady is a bigger issue. This is a screenshot of the result:

Note that the free version of the RTA includes advertisements. Also note that I did not win the $250.

We can see that the response starts dropping off around 3kHz and then makes a bit of a rally.

Doing the same with the tweeter (and using the band limited file) gives:

This also shows a roll off at around 3kHz (keep in mind that the source material is rolled off at 1.5kHz for this plot). It looks like our crossover should be around 3kHz.

 

[RayW]

The next step is to hit up one of the myriad crossover calculators on the net. The one I chose is from the same site as the Zobel calculator, here:
http://www.diyaudioandvideo.com/Calculator/XOver/

Putting in our values (and using Re for the woofer since we have a Zobel network for it) gives:

We will have to do some rounding to get available values so C1 = 5uF, C2 = 3.3uF, L1 = .5mH, and L2 = .8mH. Having done this a few years, I have a nice stash of bits and pieces to design with. If you don't, these parts are not all that expensive if you use nonpolar electrolytic (NPE) caps and 18ga inductors. Scavenging is also fine as long as the caps are nonpolar.

If we look back at the driver parameters, we will also see that there is a discrepancy between them in sensitivity. The tweeter is rated at 90dB@1W1m and the woofer is in the basement at 83dB@1W1m. The woofer's number is a product of Hoffman's Iron Law which means that you have to give up efficiency to get bass out of a small package. To equalize things out we will need an L-pad on the tweeter. Once again we will go to the DIYaudioandvideo site for an L-pad calculator here:
http://www.diyaudioandvideo.com/Calculator/LPad/

Using 7dB as a starting point and the tweeter's Re of 5.2Ω we get:

Again we will have to round and use 3Ω and 4Ω. As with the caps and inductors, I have a bucket of resistors. If you don't, they're about $.80 from PE. If you scavenge, make sure you get non-inductive resistors of about 10W or more.

The next step is to build a prototype crossover. I did it by twisting wires and using a bunch of alligator clip leads. It looked something like this:

Note: try REALLY hard not to short things out. Amps don't like that much.

 

[RayW]

So now we have a prototype speaker and we need to test it. For this we'll go back to the pink noise source and the phone. Playing full band pink noise through the speaker and measuring gives this:

In this plot we can see that the high frequencies (above our 3kHz XO point) are lower than those below it. It looks like the L-pad is doing a little too much. So we can recalculate it for 5dB:

Rounding down to 2Ω and 6Ω and putting those into our XO circuit, we get:

This shows that the treble is now too high (yeah, I overshot on purpose to make the point). So we backtrack and go again.

This is the point where ears come into the picture and we start seasoning to taste. I changed to a 6dB pad and then played with the parallel and series legs a little to get what sounded good to me. I ended up with 2Ω series and 4Ω parallel and a response like this:

Listening to some Melody Gardot now and it's sounding pretty nice. Now I have to build up the other speaker, make some more permanent crossovers, and have a stereo pair ready for finishing.

 

[RayW]

A couple of notes on design in general...

First, this design happened to work out with the "textbook" crossover point. That doesn't happen all that often. What will usually happen is a response or impedance anomaly near the crossover will leave a dip or bump in the overall response. When this happens I take whichever driver has more leeway (flat response around the crossover frequency) and adjust the frequency accordingly.

For instance, if there had been a 3dB hump at our crossover frequency then I would have moved the tweeter filter up by about 1000Hz and tried again. If this left a dip then I would move back by 500Hz. You get the idea. The point is that having a crossover nominally at 3kHz doesn't necessarily mean that both filters are nailed down there. They can be moved around independently to get the final response you need/want.

Second, I went with a generic 2nd order Linkwitz/Riley crossover for this project. Sometimes things don't work out that way. If the woofer has a nasty breakup (after it starts to roll off on the high end it comes back up higher than the average level, this is due to cone ringing) then you may need to go with a higher order filter on the woofer. See the Indignia project for an example of this. If the crossover is low in relation to the tweeter's Fs then it may start distorting at higher levels. If that happens then you can go with a higher order filter on the tweeter. The point here is that you don't have to have matching slopes on the woofer and tweeter, you just need to pay attention to how they play together in the crossover region.

I hope that's enough to get some new folks started. I have a lot of fun playing with this stuff. I hope you will, too.

 

[dnewma04]

Ray, this is great stuff! It looks like you maybe missing a couple of attachments in post 4, however.

 

[RayW]

That's weird. The pics that are linked show up in the post but not as attachments. Wonder if I attached them to the wrong post or something?

Now I've got them all attached, just not in the right order. Ah well.

 

[Dan ʕ•ᴥ•ʔ]

Very cool project. Sadly, I own a stupid-phone...not a smart-phone...

 

[StevenZ]

Good work, Ray!

 

[cubby01]

Great stuff. I'm always get a little lost in the process as it is and I think this little step by step summary will help my understanding, with or without iPhone as the case happens to be.

 

[ejfud]

That's cool Ray. Thanks for sharing the process.

 

[dnewma04]

I'm going to follow this process for a speaker build and post the results.

 

[djnagle]

Good work Ray. We sure have progressed since Paul Klipsch earned his B.S.

 

[RayW]

I'm going to follow this process for a speaker build and post the results.

Since I know you have the Audiotools app I'll let you in on a little secret...I was using the FFT from that toolbox to check the free RTA used in this thread. It's a lot more precise and easier to deal with if you know what you're looking at.

 

[dnewma04]

Ray, I do have audio tools, but it would be fun to follow your technique to see if I can make something pleasing with lesser tools. I have some vifa mids and hivi ribbons I could use.

 

[eddantes]

Hey! That's what I did last night! I used Audiotools on the iPad to have a look at my Silver Irises... and my RS full rangers in big cardboard baffles. Even though the tools are limited (mostly due to the user and the environment) they certainly do illustrate what I would say is "the general gist" of what is going on.

After measuring the Silver Irises using RTA tool with pink noise, I went on line to look at other folks measurements and they were darn close (and that is in my noisy - read furnace and kids upstairs - basement). From that exercise, I learned that you really can get decent info using these tools... and that the Silver Irises have a giant hole at the crossover.

Anyways - another reinforcement of the idea that building (playing with) speakers isn't as inaccessible as I had thought.

 

[audiozaz]

Ray,

Great thread, BUT what can you do for people that can't cut a straight line with a radial arm saw? You may want to say that you need some minimal woodworking skills. I won't even get into soldering -

Jes Saying!!

You could use the phone for what nowadays seems to be a secondary if not tertiary purpose: make a call (to a local builder).

az

 

[thefragger]

Neat-o!

Sticky?

 

[240sx4u]

Nice job Ray, very fun and easy to understand read. You're making me a bit less hesitant to dive in head first.. which is really saying something. I have a pair of DIY speakers I made and I just tossed an off the shelf crossover in there.. this may be my motivation to do it right.

Evan

 

[dnewma04]

Good work Ray. We sure have progressed since Paul Klipsch earned his B.S.

Ray does lay down some pretty advanced BS.