Thứ Ba, 12 tháng 7, 2011

Speaker Testing and Analysis



  Now that you have built your speakers, it is time to test them. You will need a microphone and a computer with a microphone input & headphone/line-out output. You will also need software that generates test tones at different frequencies and then reads back those tones using the microphone connected to the computer. 

  If your don't have a microphone for testing, you may want to build your own instead of spending a lot of money for one. Check out our page onbuilding your own microphone for information on building a quality microphone for under 5 dollars. Note that your own custom microphone will not be calibrated, so expect a ±3-4dB error vs. ±1dB for a calibrated microphone. Most of this error is above 4kHz and is a soft error. It will not cause lots of random narrow spikes in your tests, so it can be used to try to find any spikes in a response curve. It will not help in identifying speakers that gradually increase or decrease in SPL over a wide frequency range since that could be a result of the microphone itself. So what does this mean? With an uncalibrated microphone you can still design a series or parallel notch filter, but you should not add a contour network based on results from an uncalibrated microphone. If you still want to purchase a calibrated microphone, check out the combined calibrated microphone + software deal from True Audio. 

  Use a laptop for speaker testing if you have one. It will make things a lot easier. Plug your microphone into the microphone input jack on your computer. On a computer this is typically a 1/8" (3.5mm) mono plug. You will need an adapter if your microphone has a 1/4" or a standard professional microphone connector. Typically, a professional microphone will have the jack shown on the right, and the microphone cable will convert from that 3 pin connector to a 1/4" plug. You then need a 1/4" to 1/8" adapter. You can find this online or from Radio Shack. 

Microphone Connectors

  This is what the jack looks like on the computer. The jack on the computer may be pink. 

Computer Audio Connectors


  Place the microphone in front of the speaker. If you are testing a single driver, then place the microphone right in front of the driver, centered on the cone and aimed directly at the driver. Shotgun microphones work well for these tests. Place the speaker off the floor and away from any walls (including behind the speaker). This will reduce reflections that could impact the test. 

Microphone recording tweeter


  If you are testing a speaker box with multiple drivers then you will need to place the microphone a couple of feet from the speaker - or a couple of yards away if using a shotgun microphone. This is necessary so that the microphone can pick up the signal from all of the drivers. 

  Your other option is to setup a in-room listening test. Place the speakers where you want to have them permanently located in your room and place the microphone in your most likely listening position. Do not use a shotgun microphone for this type of test since you actually want to account for the in-room reflections in the test. This type of test is used to find the ideal speaker location, height, and angle for your room. This test can be done with speakers you have constructed or ones you buy off the shelf. 

  In any case, you should fashion some type of microphone holder for your tests. Do not try to hold the microphone in place with your hand. You will not be able to hold it stead enough. This might be OK for the subwoofer frequencies, but it will have a dramatic impact on the higher frequencies. You don't need to buy a real microphone stand. Use any type of clip/clamp that will hold it in place. 

Microphone Holder

  Next, plug a 1/8" (3.5mm) to RCA Adapter into the line-out jack in the computer. This will be a green jack on the computer. Then run RCA/Composite cables from the adapter to your stereo receiver. If your computer has a digital output (either optical or coaxial) then use that connector instead. You could also use a DisplayPort or HDMI connector if your stereo supports it. These digital cables carry both audio and video.
3.5mm Computer Line Out to RCA Adapter


  Finally, hook the stereo up to the speakers that you are testing. Make sure that you don't have any active crossovers or equalizers enabled if you are using a receiver. For example, any modes like Hall, Stadium, Movie, or Game will affect the speaker output. Your receiver may also automatically enable a high pass filter if you are using a subwoofer. Disconnect any subwoofer and turn off any filters in the receiver before starting your testing. Note: If you need that high pass filter because your speakers aren't full range, then leave it in place. Also, if you are performing an in-room listening test to determine speaker positioning, then you should connect and run all speakers (including the subwoofer with the active crossover) as you would normally. 

   Now for the software. For free software I recommend AUDio MEasurement System (AUDMES). On the paid side, TrueRTA from True Audio or ETF Acoustic. TrueRTA is cheap enough and very popular. There is a free version of TrueRTA, but it can only take 1 measurement per dB which is not adequate for real speaker testing. ETF has a free demo that should do everything you need. 

  All of these tools generate a series of tones from your computer and then measure the volume (SPL) of those tones using the microphone. They then display the results in a nice chart so that you can easily see any spikes and dips in the frequency response. With every tool, you should first raise the volume of your computer and lower the volume of both the stereo/receiver. You want the volume during the test to be loud (better signal to noise ratio), but not more than your speaker drivers or stereo can handle. Start low and then raise the volume once you are comfortable. Note: In my own testing, I've seen higher noise levels when the computer volume was lower and the stereo was higher - even though the net effect was the same volume. 

  Also, most tools detect your input (microphone) and output (speakers) at startup. It is best to plug in everything before starting the program. 

  How to use AUDMES: This tool is very easy to use. It should automatically detect your sound card and your input/output channels on start. If not, you can select/change your Input/Output devices from the Tools->Select Sound Card menu option. To begin your test, select the Frequency Response tab from the bottom of the window and then click Start. It takes about a minute to run each test. 

  How to use TrueRTA: There are 4 levels of TrueRTA. The highest level (and most expensive version) allows for 24 measurements per octave. The program will default to 1 measurement per octave regardless of what version you purchase, so the first thing you should do is change the RTA Resolution on the right side of the screen to whatever level you have purchased. Click the Quick Sweep button on the bottom left of the screen and within a second you have a full frequency response chart. Note: In my testing, I changed the Ampl. (Amplitude / Computer Volume) on the right side of the screen from -10dB to 0dB. 

  How to use ETF: From the top menu, select File -> New Measurement -> Normal Transfer Function. On the popup window, check the Sweepradio button in the Measurement area. From the top right of the popup window, click Level Check to check the sound level. If the sound level is green, then click Start Test. The program will not produce two quick frequency sweeps - One for the Left speaker and one for Right. If only one speaker is connected then you will hear only one sweep. Since you only have a single mono microphone, you will get only one response curve. ClickOK at the top to get your frequency response curve. Enlarge the new window with the chart to make it readable. Check the Linear Frequency Response icon at the bottom to get the right chart. 

  Here are some sample frequency response charts from TrueRTA using a B&W 602 with a custom wired Panasonic WM-61A Microphone without a shotgun tube. For the first three charts, the microphone was mounted within 1" of the tweeter, woofer, or port of the speaker. I tested the port just to see the effect it was having on the system overall. Notice the nice flat response on the tweeter measurement. Not bad for a $1.90 microphone cartridge. 

One Inch from Tweeter
Frequency Response Chart from B&W 602 One Inch from Tweeter


One Inch from Woofer
Frequency Response Chart from B&W 602 One Inch from Woofer


One Inch from Port
Frequency Response Chart from B&W 602 One Inch from Port


  In the fourth test the microphone was placed 6" from the speaker centered between the tweeter and woofer. For the final test, the microphone was 1 meter away, still centered on the tweeter and woofer. 

Six Inches centered on Tweeter & Woofer
Frequency Response Chart from B&W 602 from Six Inches centered on Tweeter & Woofer


One Meter centered on Tweeter & Woofer
Frequency Response Chart from B&W 602 from One Meter centered on Tweeter & Woofer


  Since I don't have a proper sound room to perform testing, the in room reflections are worse as the microphone is further away from the speaker. The measurement from 6" is pretty good, but since the port is below the woofer the microphone isn't able to measure the deep bass (as seen in the 3rd chart) that comes from the port. 

  I don't see anything in any of these tests that would warrant any additional modifications to the crossover. Given this is an off-the-shelf speaker that is to be expected. Here are some things to look for in speaker testing: 

  • A large spike (~20dB) at a crossover point would indicate that the drivers are cancelling each other out. Try reversing polarity (swap + & -) on one of the drivers to see if the spike goes away.
  • Any other narrow spike (~10dB) (Ex: at the resonant frequency of a tweeter) might need to be fixed using a Series Notch Filter.
  • A wide dB gain in the response that corresponds to a single driver's output could be caused by that driver having a higher sensitivity than the others. This can be resolved with a Driver Attenuation Circuit / L-Pad.
  • Other dB gains over a wide area that are not attributed to different driver sensitivities can be resolved using a Parallel Notch Filter.
  • A subwoofer with a declining response curve (caused by rising impedance of the large coil in the crossover at higher frequencies) can be fixed with an Impedance Equalization Circuit.
  • A driver with a rising response curve can be resolved with a Contour Network.