With drivers mounted in the enclosure, you dont have to take near field measurements. Far field (2-5 times of the baffle width) taken on the tweeters axis. Next is the mid without moving the mic. Next is the combined.
What Chuck said. Mount the speaker in a nice wide open area, equal distance from the ceiling and floor. I like to measure a narrow baffle about 28 to 30 inches. Whatever the distance write it down for PCD later.
Set the mic and test the tweeter volume then don't touch either. Measure the (1) tweeter, (2) mid, then (3) tweeter plus mid. Watch the polarity. You can break down the measurement gear then measure the mid/ woofer near field (4). I usually lower the volume by half for nearfield.
After you get some measurements the fun begins...Craig
I do near field last, or next to last if I'm measuring the port output too (rare). The next logical step in my experience is the most important step and unfortunately the trickiest to get right. You want to take your far field measurements next. 3 per speaker. This is where it is absolutely paramount that once you are set up you don't move the mic, not even a mm or change the volume control on your amp.
The shorter the mic distance you can do the better to strengthen first arrival vs reflections. But you need to be far enough away to capture all of the baffle step loss and diffraction effects. From Jeff Bagby's seminar I learned that you only need to be 2 times the baffle width away, or 3 times the largest driver's diameter away, which ever is larger. So a 10" wide baffle you can measure at 20" so long as the woofer is a probably an 8" or smaller (cone diameter + 1/2 surround, not frame diameter).
Set your speaker on a stand out in the middle if the room, at a height aproximately 1/2 way between the ceiling and the floor. Doesn't have to be exactly 1/2 way. Set up your mic at the tweeter level pointing directly at the tweeter. Make sure you are perfectly perpendicular to the baffle (left/right and top/bottom). Set the mic at an appropriate distance based on baffle width and driver diameter.
With a tape measure carefully measure and write down the mic distance. This is usually just measured from the baffle surface to the end of the mic. I add an extra 3/16" to this measurement because the actual mic captule is that far back in my OmniMic. I've never seen this talked about anywhere before. But I got frustrated figuring out the Z offset. I could make the overlays match pretty darn good, but never great. Based on experimentation I found that adjusting the mic dist + 3/16" made my overlays significantly better.
Time to actually measure. Use the short sine sweep (OmniMic track 2). You are going to take 3 different measurements. First the tweeter connected to your amp. Start with the volume all the way down. Start recording in OmniMic, start you CD and turn up the volume until your passband volume is approximately 90 dB. This is pretty loud and will swamp out most background noise. I am pretty anal and turn off my A/C or heat just to make the room as quiet as possible. Wife and kids go to grandma's . Set up OmniMic for blended, 24th octave smoothing, and 5 msec gating. Might need to change gating based on your room but 5 msec is a reasonable starting point. Let it woop woop woop about 10 times (it averages). Pause and save .frd file for the tweeter. Call it something good like Tweeter_1_raw.
Next measure the woofer. Same exact thing as for the tweeter. Then hook up your amp to the woofer and tweeter in parallel. This last measurement will be used to determine the relative Z offset between the drivers. Execute and report back. Also, I alway measure both speakers (6 measurements) because I've measured significant differences between drivers. So much that I had to make xo padding changes between the two speakers (RS28A!!!).
OK thanks guys, I'll take some measurements later today. So I will use 12 inches since that is 2x baffle width (6") and 3 times woofer diameter (4" driver but is actually 3.25")
Latest Bagby comment was 48th smoothing. As I was a big ARTA user, I check the impulse window and look for reflections. FWIW usually am something less than 5.0 ms.
Personally I think 12 inches is to short. I could make up something about more time for integration or seeing the baffle edge, but I've only got one data point where I thought my data looked weird.
Ok, I just shoot for 20 inches. Do I need a cap to protect this tweeter when I run the Tweeter and woofer simultaneously or is the volume level not enough to damage a robust tweeter and that is a ribbon type issue?
I thought that measuring " in box" included BSC and offset.
Bsc, yes. Offset no. It is capturing time of flight of the driver. This is why we take the 3 measurements. Make sure phase is turned on!!!!! Cant reiterate that enough.
Import each driver into pcd. Just FR. We dont need imp yet. Next use the import function for the combined data. Enter your x and y offsets, and mic distance. Now adjust the z offset to get the two lines to match up. With your rear mounted woof, i would probably start at -.05.
Stay at 36 inches, works for small and big speakers and 36 inches is easy to remember and is close to 1m. Center on tweeter axis and measuring tweeter, them mid and then woofer. Then wire everything in parallel and measure.
Once done measuring, export to frd and import into response modeller or freq blender
Then extract min phase.
Now you are ready for PCD.
For gating i usually look for the first wiggle, that kind of ends up between 300 and 400hz.
I thought that measuring " in box" included BSC and offset.
Bsc, yes. Offset no. It is capturing time of flight of the driver. This is why we take the 3 measurements. Make sure phase is turned on!!!!! Cant reiterate that enough.
Import each driver into pcd. Just FR. We dont need imp yet. Next use the import function for the combined data. Enter your x and y offsets, and mic distance. Now adjust the z offset to get the two lines to match up. With your rear mounted woof, i would probably start at -.05.
Well... actually, yes - offset is captured in the measurements, at least in ARTA. I can't speak for Omnimic (not sure how it handles phase on a per-measurement basis), but actual time of flight is captured in individual measurements in ARTA. If you want to model from a variety of points, then yes - multiple measurements are required.
Couple nitpicky things, however....
1. Craig - excellent observation on the capsule placement. Seriously - good to know, and probably should be a published specification.
2. Because simulators use modeled pistonic behavior for off-axis calculations, woofers with breakups will be unpredictable in PCD/WinPCD etc.
3. Derive your offset from three points in space, and you will have three offsets - due to the nature of a triangle. Since we find the offset by isolating the time of flight difference between two sources, moving up or down from the initial axis will change that difference. This will change the model, in addition - see point 2. One thing I did to help minimize those differences was to lock my mic boom in place - permanent 90 degrees, lock the base swivel in place, mark my floor where each of the three feet sit, and established a consistent method for measuring height.
4. Have I mentioned that relying heavily on simulators to correctly simulate power response and off-axis response can lead you down a path you don't want to go? Learn how to take off-axis measurements and interpret them.
5. My personal opinion: a single set of measurements from a meter is usually sufficient. Limit involvement of simulators to establishing the critical "first network" and switch to pure measurements from there. Take your first measurement based on your first simulation, measure multiple off-axis points, learn what a listening window is, and if you need a sanity check - plug current values into your simulator.
6. Pay as much attention to impedance profile as you do response profile. A good amp will handle a nominal 4 ohm load, but a wildly swinging impedance with the corresponding wonky phase can lead to some serious head scratching on why things might not sound "right".
7. Do not use any Class D or Class T amp for measurements - you will have so many head scratching moments when your tweeter level is always off according to the simulator. A good, used, 15 year old AVR can be had for peanuts these days, and the amp sections are generally superior to most Tripath or similar - at least when it comes to maintaining a linear response independent of impedance.
8. In all honesty, I view simulators with a dose of skepticism - I try very hard not to let the simulators tell me what to do. I use them to try out ideas without dragging out measurement gear, and I use them to sanity check networks, but I still engage in extensive listening with a short (~25 songs) list of music in a listening position I am familiar with.
Excellent input guys! Mike and I had an off forum discussion about what amp to use for measuring, so I mistakenly omitted that from my posts. Thanks JR for adding that.
I am supposed to save "average curves" right? I turned phase on and used 1/24th smoothing. The first reflection in my room is at 4ms so I have it set to 3.9seconds.
hmmm, tweeter impulse response 4ms is the first blip with the woofer its at 3.5ms. I left it at 3.901ms, I checked with it was 3.5 and it didnt change the response at all so I just went back to 3.901ms
OK I got all 6 FRD's and 4 ZMA's. Time to go on to the next steps. I haven't used response modeler or the blender. So I could use some guidance on the next steps please.
I'm kind of confused now. The old blender white paper says you need NF and FF measurents but the 2.0 version Charlie's instructions it appears that you only need FF. I tried blending but my blend Range appears to be very high like around 1000hz when it meets up. What am I doing wrong here?
Now is the time to start sharing files and stepping through finding the offsets and woofer blending. I like to open a word document and drop screen shots into it.
You can skip the blending if you are crossing over high wrt the gating frequency. This is a two way, no need to blend, you should have good freq overlap in the crossover region.
OK will do, first off here is a comparison of my measurement FRD's for both tweeters and both woofers overlay for comparison purposes. They seem to match pretty well driver to driver.
Here are my settings for the blender. Should I use a steeper slope for the tails or is 12db correct? Let me know if I am matching this properly or not. I keep getting a constant error for a circular reference on this excel sheet as well. The first pic is just the raw files put in. The 2nd and 3rd pic are after blending. Im not sure if I did this correctly or not.
Ok I've re-read the papers a few times. It looks like frd's response blender does want a near field measurementservice which I didn't do. In field F10, it shows "near field measurement" on the frd's blender. Is this step necessary or do I move on to another spreadsheet at this point.
I always do the near field splice (blended), then use the blender software to add the appropriate tails to the lower and upper roll offs, then extract minimum phase. It is the good way to get accurate phase data. And this is important in the final step, WinPCD. WinPCD actually has instructions on the page where you calculate Z offset and it says the woofer and tweeter files should be pre-processed.
Min Phase is necessary - but you can do that with the far field data. Blending of the near field and the far field per se is not necessary, but is good if you want accurate bass response down to 20 hz.
Comments
Set the mic and test the tweeter volume then don't touch either. Measure the (1) tweeter, (2) mid, then (3) tweeter plus mid. Watch the polarity. You can break down the measurement gear then measure the mid/ woofer near field (4). I usually lower the volume by half for nearfield.
After you get some measurements the fun begins...Craig
The shorter the mic distance you can do the better to strengthen first arrival vs reflections. But you need to be far enough away to capture all of the baffle step loss and diffraction effects. From Jeff Bagby's seminar I learned that you only need to be 2 times the baffle width away, or 3 times the largest driver's diameter away, which ever is larger. So a 10" wide baffle you can measure at 20" so long as the woofer is a probably an 8" or smaller (cone diameter + 1/2 surround, not frame diameter).
Set your speaker on a stand out in the middle if the room, at a height aproximately 1/2 way between the ceiling and the floor. Doesn't have to be exactly 1/2 way. Set up your mic at the tweeter level pointing directly at the tweeter. Make sure you are perfectly perpendicular to the baffle (left/right and top/bottom). Set the mic at an appropriate distance based on baffle width and driver diameter.
With a tape measure carefully measure and write down the mic distance. This is usually just measured from the baffle surface to the end of the mic. I add an extra 3/16" to this measurement because the actual mic captule is that far back in my OmniMic. I've never seen this talked about anywhere before. But I got frustrated figuring out the Z offset. I could make the overlays match pretty darn good, but never great. Based on experimentation I found that adjusting the mic dist + 3/16" made my overlays significantly better.
Personally I think 12 inches is to short. I could make up something about more time for integration or seeing the baffle edge, but I've only got one data point where I thought my data looked weird.
Import each driver into pcd. Just FR. We dont need imp yet. Next use the import function for the combined data. Enter your x and y offsets, and mic distance. Now adjust the z offset to get the two lines to match up. With your rear mounted woof, i would probably start at -.05.
Once done measuring, export to frd and import into response modeller or freq blender
Then extract min phase.
Now you are ready for PCD.
For gating i usually look for the first wiggle, that kind of ends up between 300 and 400hz.
Sorry forgot this is a two way....
Couple nitpicky things, however....
1. Craig - excellent observation on the capsule placement. Seriously - good to know, and probably should be a published specification.
2. Because simulators use modeled pistonic behavior for off-axis calculations, woofers with breakups will be unpredictable in PCD/WinPCD etc.
3. Derive your offset from three points in space, and you will have three offsets - due to the nature of a triangle. Since we find the offset by isolating the time of flight difference between two sources, moving up or down from the initial axis will change that difference. This will change the model, in addition - see point 2. One thing I did to help minimize those differences was to lock my mic boom in place - permanent 90 degrees, lock the base swivel in place, mark my floor where each of the three feet sit, and established a consistent method for measuring height.
4. Have I mentioned that relying heavily on simulators to correctly simulate power response and off-axis response can lead you down a path you don't want to go? Learn how to take off-axis measurements and interpret them.
5. My personal opinion: a single set of measurements from a meter is usually sufficient. Limit involvement of simulators to establishing the critical "first network" and switch to pure measurements from there. Take your first measurement based on your first simulation, measure multiple off-axis points, learn what a listening window is, and if you need a sanity check - plug current values into your simulator.
6. Pay as much attention to impedance profile as you do response profile. A good amp will handle a nominal 4 ohm load, but a wildly swinging impedance with the corresponding wonky phase can lead to some serious head scratching on why things might not sound "right".
7. Do not use any Class D or Class T amp for measurements - you will have so many head scratching moments when your tweeter level is always off according to the simulator. A good, used, 15 year old AVR can be had for peanuts these days, and the amp sections are generally superior to most Tripath or similar - at least when it comes to maintaining a linear response independent of impedance.
8. In all honesty, I view simulators with a dose of skepticism - I try very hard not to let the simulators tell me what to do. I use them to try out ideas without dragging out measurement gear, and I use them to sanity check networks, but I still engage in extensive listening with a short (~25 songs) list of music in a listening position I am familiar with.
Just my $0.07