I have more than a little sympathy for DIY loudspeaker builders, because that is how I got my start. I think I was about 10 when I built my first speaker with some scrap plywood and a 5-1/4″ driver that my brother gave me. How time flies…
The hobby itself can be quite seductive, especially if you have some woodworking skill. It seems like it’s a pretty straight-forward matter to build a box, buy some drivers, mount them along with a crossover or something, and then have a system that looks great and costs less than the commercial equivalent. However, once you start you will realize that good speaker design is actually a very complex affair. It only seems easy because of the relatively few parts that are involved. But I will let you discover all that for yourself–it’s part of the fun. If you do decide to go down the DIY path, you will do yourself a favor to have modest expectations at the start, and be prepared to get sucked into something that can take over your life if you let it.
Here are a few resources I recommend for the DIYer. Most of these will be known to experienced builders, but if you are new to the area you may not know all of them.
Drivers and other parts
Madisound Speaker Components is one of the best sources of drivers and other parts (capacitors, inductors, etc.) for both the DIYer and for small manufacturers. They have consistently been my first choice for supplying Biro’s own needs. Their service is top-notch, personal, and has never let me down.
Parts Express is another good source of drivers and parts. Professional service and good parts selection. It may be a small thing, but their selection of cabinet ports is quite good.
I have experience with Solen Electronique and MCM as well and cannot complain about either.
If you don’t like the idea of making lots of sawdust, Parts Express’ finished cabinets are really hard to beat. I have used these for some prototype projects, and the workmanship and finishes that I have seen have been really good. Madisound has a similar line of cabinets. While I expect that they are of similar quality, I have not actually seen any of these so I cannot say so with any certainty. (Madisound also sells a line of cabinets made by Woodstyle in California that some people really like, but the link shows nothing at the moment. I don’t know if that means they have been discontinued or if it’s just a bug in the website.)
If you are looking for information to help you decide what driver to use in your next project, Zapf Audio has a wealth of information to help you. John “Zaph” Krutke’s attention to detail in his measurement and methodology is admirable, as is his enthusiasm for sharing his findings. In addition to driver measurements, John has published a number of complete designs at his site. I have not heard any of them; however, they appear to have been thoughtfully and competently designed. While I don’t agree with the prioritization of all his evaluation criteria (perhaps the subject of a later post), I do have a ton of respect for his opinions, and they have enhanced my outlook. I have never met or exchanged email with John; in spite of this I feel comfortable saying that he is one of the few voices that are really worth listening to in the DIY speaker hobby realm.
You will need design software. And not just for cabinet design. You can’t really build a good system without software to help you with measurement and crossover design. While everyone seems to have their favorite in this area, I think LspCAD from IJData is just fine.
If you want good results, you must measure the performance of the drivers you have chosen in the cabinets that you will use them in. Because of the varying diffraction effects from different cabinets, you cannot use the measurements from some other source in your design. And you should never trust manufacturer data. Sometimes published curves are from preproduction prototypes, sometimes they are outright lies, and very rarely is enough information given about test conditions to let you extract useful information.
To measure your drivers you will need a microphone with a very flat and/or calibrated frequency response. Some DIYers build their own using Panasonic omnidirectional electret elements–some of which are incredibly flat. The only problem with this approach is that while the flattest of the Panasonic omni elements are quite flat, when all is said and done you may still be left with as much as 3dB error in the audio band. In my opinion, that’s not good enough. A few people make and sell complete mics using these elements and provide you with calibration data as well. This is the approach I recommend. The ones made by Kim Girardin at Wadenhome Sound are very good and very affordable. I’ve known Kim for several years as a result of our association with the Upper Midwest Chapter of the AES. He has real enthusiasm for the field and is one of the nicest people you are likely to encounter in the audio world. You may be able to plug one of Kim’s mics directly into the Mic input of your computer soundcard, but for the best results you will want a preamp with a controlled polarizing voltage. If this is the case, the Mitey-Mic II (or MM2) is a classic.
Most mics based on the Panasonic elements have simple two-conductor outputs and are meant to be interfaced to soundcard “mic” inputs or something like the Mitey-Mic II. If you want an instrumentation mic with a more conventional balanced output (and using +48V phantom power), the Behringer ECM8000 looks interesting. I have never seen one of these in real life, so it may actually be utterly poopy. But the specs and price look decent.
While the soundcards in desktop PCs are usually good enough for making usable speaker measurements, laptop soundcards tend to, er, suck. I have been using Behringer’s tiny and cheap UCA202 USB soundcard with my laptop when I need to make measurements with it, and the results have been just fine. It uses decent 16-bit TI/Burr-Brown converters, and the headphone output is particularly useful for making impedance measurements and is the main reason I use it rather than similar USB devices. (Look here for some test results.) Don’t mess around with the packaged ASIO driver–just plug it into your WinXP machine and let it use WinXP’s built-in USB audio drivers. If you absolutely, positively have to have ASIO with this guy, I recommend ASIO4ALL.
I regret that I don’t have time to respond directly to DIY questions. If I don’t reply to your inquiries or address your question here, please don’t be offended. With this post I really just wanted to offer what little support to hobbyists that I can. And as time allows, I will try to post other tips and suggestions.
8 thoughts on “DIY 1”
Congratulation for this excellent post.
My english is too bad for more words. 😦
I hit your tests of UCA202 after I started my research why it returns output singnal on input. While your results are great (and I don’t doubt them), there is one measurement missing – how much of the output signal returns on input – with only USB connected, no RCA cables in or out. And the truth is quite ugly.
I sent 0dB 1000Hz to it and input level was -65dB, higher freq, even better.
But 500Hz->-60dB, 200->-54, 100->-48, 20(ok, that’s extreme)->-36dB.
Guess how much of some rock drums you can hear in between your singing for instance. The only workaround I see is to lower your master when you record another track.
Yeah … that’s quite a bit of crosstalk. If I find the time I will test this on the ADS RDX-150 as well. I am curious as to whether this is a shortcoming of the BB codec chip or a layout and/or power supply and/or grounding isue. Does the amount of output-to-input crosstalk change if you change the headphone level and/or turn on/off the “Monitor” switch?
The strange thing is that there is less crosstalk from right-to-left in playback than in output-to-input.
Maybe, just maybe, I made something wrong. I can’t measure this with analog audio because I have no tools to do so + it is not related to RCA cables. I used REAPER as a DAW, ASIO driver is Behringer’s, tone generator is free VST plugin by mda (http://mda.smartelectronix.com/effects.htm). I created two tracks, I run tone generator on track #1, track #2 is armed for recording, using any channel as input. When I run mda tone gen on 0dB I can clearly see signal on input.
I wanna check it out with some other ASIO capable DAW, but I have to try some demo version or so. I tried messing with Monitor switch and/or with headphones too, no significant change – but I’ll rather go through various combinations again. When I run the same REAPER setup with Creative’s ASIO (oldish Audigy), no problem… so it must be beyond REAPER. I doubt it is in digital realm, so I blame part of the UCA behind DA converters. When I learn more, I’ll add new information. MAYBE it’s just my piece, I don’t know. 😉
Thanks for the extra info.
P.S. You might want to try ASIO4ALL instead of Behringer’s ASIO driver. The Behringer ASIO driver that came with my UCA202 was a repackaged version of someone else’s very old ASIO driver. I found that ASIO4ALL gave me better results, but I can’t remember the details. 🙂
I ran a subset of your tests with my UCA202 and can confirm your findings. I got the following results:
Noise floor: -75dB
This is using the standard Windows XP drivers, Audacity for tone generation, and Paul Marshall’s Level Meter V.1.1 (darkwood.demon.co.uk) for input level sensing. Nothing was connected to either inputs or outputs
Very strange … and disappoining.
When I tried the same with my ADS RDX-150 I get:
Noise floor: -78dB
IIRC, the RDX-150 and the UCA202 use the exact same TI codec chip, so the differences here are almost certainly due to layout and supporting circuit issues.
Thank you very much. I have good experiences with current Behringer’s drivers, so no problem there (and I guess it doesn’t affect this particular type of crosstalk ;-)). Now I was discussing this with Behringer, they have good support (at least this first experience was good in terms of responses, attitude, etc.). They generally talk about -77dB on 1kHz in this Out->In crosstalk, but I made the man to measure 200 and 100Hz as well. His unit gave -62 and -58dB respectively – that’s better but still worse than stereo crosstalk which is curious.
I don’t know how expensive is the other device, truth is that UCA202 is crazy cheap. When I lower master in REAPER (or whatever DAW for that matter) you can fight the problem and still monitor on good levels – as the monitor phones have quite strong amp or you can use mixer monitoring which gives you enough sound. So… I can bear with it, hopefully there are no pieces worse than mine. On the other hand, man at Behringer told me that I should let my unit examine at local reseller and… I don’t know, maybe switch for a better one? 🙂
Thank you for your time and test. UCA202 has its twists obviously, but I still can’t just put a poo on it – for the price is extremely affordable.
I agree: the UCA202 is very cheap and despite this crosstalk shortcoming it is a good value. I ran the tests against the RDX-150 to see if the problem with the UCA202 was related to the codec chip or to design issues. The RDX-150 was (and still is) about 1.5 the cost of the UCA202, has optical S/PDIF input as well as output, but lacks a headphone amplifier and input monitor. The one I have has a PC board that is fully shielded (all “blank” areas are filled with grounded copper), whereas the UCA202 is fully unshielded. This one small factor may be the reason for the added crosstalk in the UCA202. If I have time this summer, I will try to trace out the UCA202’s circuit a bit more carefully to see if there is a circuit-design change that will fix the problem.
I was surprised to see the lack of shielding in the UCA202’s PC board because it doesn’t cost more to design a shielded board and AFAIK they are not more costly to manufacture. (None of the vendors I have worked with has charged me extra for fully flooded boards nor offered a discount for not flooding; but then I am not having boards made in the millions.) Maybe Behringer reclaims the extra copper that is etched off their boards and the copper has more value than the chemicals used to remove it … I really don’t know. In any case, the UCA202 is still a good value, and it’s good to have discovered this “bug” so people can work around it.
BTW, the current version of the RDX-150 seems to be the RDX-150-EF. I have no idea what changes have been made in the EF version.