This is my first attempt at really designing a speaker. I know you're not an OB type of guy and you have your reasons, but your knowledge and understanding would be greatly appreciated by many I'm sure. I built a set of 2-way open baffle speakers and they sound very good to my ear and the measurement look OK to my eye, but I have little understanding of the finer details of these plots. I don't want you to pull your punches. I hope to know everywhere I went wrong with this build and right I guess. I'm not going to get upset and try to run you into the ground for your honesty. So I'll post 2 frequency response graphs I took. I ask that you analyze them and just tell me what's going on in them. Mic and soundcard are calibrated.
Both graphs are not smoothed at all as I'm sure you'll be able to tell. They are both taken in the listening room.
The first graph is at about 3.5ft away(I didn't measure) from the speaker and pointed between the 2 cones.
I'll tell you what I think I see: There's a boost at >=15k that is just part of the speakers response. It exists in every graph I've seen of the RS100s-8. There's a large dip at about 150Hz. I would love to know what's going on there esp after you see the next graph. It also looks like there is a regular and repeating peak and dip cycle in the MR/treble though I'm not sure if it's a problem with the speaker or a reflection issue in the room. The RS100-8 is mounted in a 3/4 inch hole with a round over placed in the back of the baffle hole. Might the FR indicate a cavity resonance? The balance knob was twisted all the way to the left, but the right speaker is still barely audible. Could it be some comb filter of some sort from the other speaker? Maybe you know..... Other than that it looks OK to me. I know this is just an on axis graph and a lot of plats need to be taken to really get an understanding of what's going on in the room.
In the second graph, the crossover region was equalized using an off the shelf RatShack meter. The treble was done with my wife sitting on the couch telling me where to twist the treble knob on the amp. This is with the stereo pair, mic roughly centered about 7ft from the speakers at the listening position.
What I see: The woofer is too loud, a 75Hz dip (1/2 the frequency of the dip at half the distance--kind of a cool phenomenon.
) I'm sure you can explain that, but I really can't directly and concretely, but I know it's simple.), A similar MR/treble ripple but looks less narrow--possibly half as many ripples, and the 15kHz spike is gone--probably eaten up just by the air.
Are my guesses correct? Are you able to explain any of the things I notice or do you not have enough information?
Again, thank you. Any knowledge you throw my way would be greatly appreciated. Any recommendations of where I should go from here?
Dan
Oh, here's a photo of the speaker(s) in question:
Both graphs are not smoothed at all as I'm sure you'll be able to tell. They are both taken in the listening room.
The first graph is at about 3.5ft away(I didn't measure) from the speaker and pointed between the 2 cones.
I'll tell you what I think I see: There's a boost at >=15k that is just part of the speakers response. It exists in every graph I've seen of the RS100s-8. There's a large dip at about 150Hz. I would love to know what's going on there esp after you see the next graph. It also looks like there is a regular and repeating peak and dip cycle in the MR/treble though I'm not sure if it's a problem with the speaker or a reflection issue in the room. The RS100-8 is mounted in a 3/4 inch hole with a round over placed in the back of the baffle hole. Might the FR indicate a cavity resonance? The balance knob was twisted all the way to the left, but the right speaker is still barely audible. Could it be some comb filter of some sort from the other speaker? Maybe you know..... Other than that it looks OK to me. I know this is just an on axis graph and a lot of plats need to be taken to really get an understanding of what's going on in the room.
In the second graph, the crossover region was equalized using an off the shelf RatShack meter. The treble was done with my wife sitting on the couch telling me where to twist the treble knob on the amp. This is with the stereo pair, mic roughly centered about 7ft from the speakers at the listening position.
An externally hosted image should be here but it was not working when we last tested it.
What I see: The woofer is too loud, a 75Hz dip (1/2 the frequency of the dip at half the distance--kind of a cool phenomenon.
) I'm sure you can explain that, but I really can't directly and concretely, but I know it's simple.), A similar MR/treble ripple but looks less narrow--possibly half as many ripples, and the 15kHz spike is gone--probably eaten up just by the air.Are my guesses correct? Are you able to explain any of the things I notice or do you not have enough information?
Again, thank you. Any knowledge you throw my way would be greatly appreciated. Any recommendations of where I should go from here?
Dan
Oh, here's a photo of the speaker(s) in question:
An externally hosted image should be here but it was not working when we last tested it.
Hey dantherman, in regards to the dip you're seeing at 75Hz there are a few reasons but the one I'm most drawn to is that's nodal in origin (a function of room acoustics). you can also see that this dip was really already there a bit higher up the spectrum and when you moved back it appeared to move with you. You can see another "moving dip" at about 310Hz that falls back on the other graph aswell. try taking the same measurement further back. I'll bet those nulls continue to move down in frequency.
Hi Dan
First - go get HolmImpulse and retake all your data. From what I see its not very useful.
Try and get the speakers up off the ground and as far away from anything else as possible. Then measure them without any room reflections and post that, or send me the data.
If you want to do things right then put the speaker on a rotating platform and take data every 7.5 degrees or so.
The bottom line here is that I'm no better at looking at bad and insufficienct data than anybody else. The key is to make decisions on good data. Getting "good data" is the first thing that you have to do or nothing positive can follow.
HolmImpulse is the answer. I am still amazed at how good it is for free. Its all I use anymore.
First - go get HolmImpulse and retake all your data. From what I see its not very useful.
Try and get the speakers up off the ground and as far away from anything else as possible. Then measure them without any room reflections and post that, or send me the data.
If you want to do things right then put the speaker on a rotating platform and take data every 7.5 degrees or so.
The bottom line here is that I'm no better at looking at bad and insufficienct data than anybody else. The key is to make decisions on good data. Getting "good data" is the first thing that you have to do or nothing positive can follow.
HolmImpulse is the answer. I am still amazed at how good it is for free. Its all I use anymore.
Yes, a $37 Parts Express buyout deal originally built for Genelec.
Not that it really means anything of course.Thanks "Q" and Dr. Geddes! Input much appreciated.
So maybe I should build something like a free standing lazy susan (I have a cutout from a 15" speaker that should work for the base of such a device) and mark it in 7.5 degree increments, set it on top of a tower (I have a 3ft bar stool--good enough?) of some sort and take a portable rig out to like a soccer field, park myself in the middle and start over? Sounds like fun! Too bad my only portable computer is now a Mac. HolmImpulse is not supported as far as I know, and my mic pre requires a wall outlet w/o making or buying some sort of batter power alternative. My Toshiba just died after 10yrs of faithful service. Maybe the street or driveway would be good "good enough" since I could reach it with my home PC and some extension cords/ long mic cable? My neighbors should have fun with that one.
They already think I'm crazy with all my projects.Is there a good thread anywhere on how to take good measurements? I hate to pester with newbie questions if they've been covered. I don't even have enough basic knowledge to ask good questions.
Thanks for getting me started in the right direction,
Dan
You could use Audacity or similar to play/record the signal.
Then import it into HOLMImpulse.
A suitable signal can be exported directly from HOLMImpulse.
You could search for HOLMImpulse and get the two threads the author have started, one for the program itself and one for measuring with it.
Then import it into HOLMImpulse.
A suitable signal can be exported directly from HOLMImpulse.
You could search for HOLMImpulse and get the two threads the author have started, one for the program itself and one for measuring with it.
Thats a good idea, worth some time.
The key to doing good measurements is to understand the reflection issue. Reflections will mess up the measurement in ways that are not very predictable and as such, once there is a significant reflection from anywhere, the the data is not reliable any more. Thus, one only need be concerned with the first reflection. A good thing to do is to draw a measurement setup on a piece of paper and trace out the reflections. This will give you an idea of the issues. With a 3 foot high stool, your first reflection is going to be the floor. That means that any surface that is further away from the speaker than 3 feet won't matter. Hence, almost any inside room will suffice.
I do all of my measurements in-doors, in my living room in fact. But I am lucky to have a high ceiling in that room which helps alot because floor to ceiling is usually the biggest limitation. So find your largest room and envision fitting a giant football into it. Orient the football such that you can get the biggest one possible in. Draw this out if you have to. Now for the best possible measurement in that room the speaker should be at one foci and the mic at the other. Almost always this will place the speaker midway in the room from ceiling to floor and out from a corner by the same distance. The mic will go in the other corner at about the same point. Its often advantagious to move the mic a little closer but never less than about 1 meter since then you won't get very good data. You will need to move the furniture from this room while you take the measurements.
One thing that I do can improve things a bit. My floor bounce is always the closest reflection, so I make a big pile of blankets and pillows on the floor between the speaker and the mic. This then makes the ceiling the closest "significant" reflection, but there is, of course, nothing that I can do about that.
The turntable is precisely what I do. Just use two flat 3/4" boards with smooth surfaces (I use melamine) and put grease between them. Put a pin at the center of rotation through them as an axis. Generally I make the top board smaller than the bottom one so that I can mark the bottom one with the 7.5 degree lines. This works nice as long as the speaker is not too heavy. A purchased bearing is not going to be stable enough and vibration can be a problem in a measurement.
With this setup and Holm you should be able to get good data above 200-300 Hz. For data below this frequency you need to do near-field - an inch or two away from the woofer. For a dipole this is not going to yield accurate results, but its better than nothing. Fortunatly IMO, the LF stuff down this low is not critical at all and is best set with EQ after the speakers are actually in the real room at their final locations.
Using Holm is well documented elsewhere as stated.
Last edited:
Thanks David. I will do that today. I do seem to remember a thread started not long ago about HOLM. I just didn't have a clue to what it was saying at the time. I should be much better off now.
Dr. Geddes, that is good info to know. After your first post I thought I would need either an anechoic chamber or to go way outdoors in a field to get anything useful. I've got high ceilings as well, so the floor should be my biggest issue--but I've got lots of pillows and blankets. This should be like a FR measurement 101 cliff note and something all beginners should read. I thought it was going to be way more complicated--not that it's all that simple or obvious to the uninitiated, but that will change once I get to the other side of simplicity.
Thank you!
Dan
Ask (the author of HolmImpulse) should be praised
as seldom have I used such a great piece of software, and never for free. He has added everything that was ask for. Thanks Ask!
Thank you Ask! You must have written an excellent program to get such praise from so many.
Thank you again Dr. Geddes. I was hungry and returned to camp with the fish and you just taught me how to prepare them properly. Now if I could only figure out how you cook. Sorry for the cheesy analogy.
It's interesting that so many people judge internet personas by their tact, but in the end a person's actions are a far better measure of their character.
IOW, actions speak louder than words,
Dan
Hopefully I'll have better measurements next week.
Thank you again Dr. Geddes. I was hungry and returned to camp with the fish and you just taught me how to prepare them properly. Now if I could only figure out how you cook.
Sorry for the cheesy analogy.It's interesting that so many people judge internet personas by their tact, but in the end a person's actions are a far better measure of their character.
IOW, actions speak louder than words,
Dan
Hopefully I'll have better measurements next week.
I'm too old to take all that time being tactful. I'd rather be blunt and get more done.
Dr. Geddes has given a perfect description of good measurement protocol. You must have clean time-domain data before doing anything else.
The floor bounce (typically the first reflection) may sound minor, but it's not. When viewing your data in the time domain, getting rid of that first reflection will turn out to be harder than you expect.
1~2" of audiophile miracle absorber (made of a proprietary blend of unobtanium and melodium) will have dismayingly little effect - a few dB of absorption at most. So you add more. And more. The reflection hardly goes down at all. In frustration, you throw away the audiophile stuff, run around the house collecting anything soft, and start piling up coats, pillows, blankets, and comforters. When the mountain (halfway between the speaker and the microphone) gets to be more than a foot high, you'll start seeing 10~20 dB of absorption. You don't want the mountain to actually intercept the direct wave, but you do want 20 dB or more of absorption for that first reflection.
A mixture of heavy wool winter coats and down-filled comforters and pillows works pretty good. The reaction of domestic horror to the spectacular mess you've made in the middle of the living room, is, regrettably, one of the little prices we must pay. Think of it this way - you will find out for yourself just how worthless audiophile miracle absorbers really are, and your negotiation skills will improve. It's a win-win all around.
Last edited:
Here's an example of the lengths I go to remove time-domain artifacts. The curvature you see is an artifact of barrel distortion in the digicam zoom lens, not the physical steel rod, which has no visible curvature.
The 1/2" steel rod has a plastic sleeve so the mike-stand can grip it, and there's a thin slotted plastic shell between the 1/2" Aco Pacific 7012 condenser mike and the 1/2" steel rod. If you look closely, you'll see the front edge of the shell has a smoothly rounded edge between the microphone and the shell. The counterweight is about 6 feet behind the front of the microphone, and commercial mike-stand is most of the way towards the rear.
The point of all this foolishness is to minimize (and delay) reflections from the mike-stand itself, which would contaminate all measurements. When I started making time-domain measurements back in the early Nineties, I just mounted the mike directly on the commercial mike-stand, and it took a while to realize the rough-looking HF measurements were actually artifacts, not problems with the tweeter. Things started really smoothing out after Mike Spurlock built this mike stand, and I started using good-sized pillow piles to reduce the floor bounce.
As mentioned above by Dr. Geddes, contamination in the time domain makes it impossible to tell what the speaker is doing - no matter what kind of speaker it is, or how much you spend on measuring equipment.
The 1/2" steel rod has a plastic sleeve so the mike-stand can grip it, and there's a thin slotted plastic shell between the 1/2" Aco Pacific 7012 condenser mike and the 1/2" steel rod. If you look closely, you'll see the front edge of the shell has a smoothly rounded edge between the microphone and the shell. The counterweight is about 6 feet behind the front of the microphone, and commercial mike-stand is most of the way towards the rear.
The point of all this foolishness is to minimize (and delay) reflections from the mike-stand itself, which would contaminate all measurements. When I started making time-domain measurements back in the early Nineties, I just mounted the mike directly on the commercial mike-stand, and it took a while to realize the rough-looking HF measurements were actually artifacts, not problems with the tweeter. Things started really smoothing out after Mike Spurlock built this mike stand, and I started using good-sized pillow piles to reduce the floor bounce.
As mentioned above by Dr. Geddes, contamination in the time domain makes it impossible to tell what the speaker is doing - no matter what kind of speaker it is, or how much you spend on measuring equipment.
Attachments
Last edited:
Hi Lynne
Could you estimate at what frequency you began to see reflection problems from the stand? Seems to me that below 8 kHz this is not likely since the physical dimensions aren't large enough. But above 8 kHz no contest. Maybe another reason why I don't pay much attention to these frequencies - they aren't reliable.
But I like your design and might try that myself.
Have any comparison data per chance?
Could you estimate at what frequency you began to see reflection problems from the stand? Seems to me that below 8 kHz this is not likely since the physical dimensions aren't large enough. But above 8 kHz no contest. Maybe another reason why I don't pay much attention to these frequencies - they aren't reliable.
But I like your design and might try that myself.
Have any comparison data per chance?
hi Dan, I recently posted some pics of how I do measurements.
It takes a lot of effort but the result is very reliable.
outdoor setup:
http://www.diyaudio.com/forums/multi-way/155632-s12-ob-returning-true-ob-3.html#post2002420
results:
http://www.diyaudio.com/forums/multi-way/155632-s12-ob-returning-true-ob-4.html
For OB, very important are the off-axis measurements.
With OB I found with gating not enough resolution can be had for lower frequency. There is a reliable workaround though as provided by john k.
I'm trying to understand at what distance the tweeter should be measured for realistic reproduction? I found out that when flat at 1m the sound is still too bright. What distance is the summa/nathan measured at?
I'm trying to understand at what distance the tweeter should be measured for realistic reproduction? I found out that when flat at 1m the sound is still too bright. What distance is the summa/nathan measured at?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.