I found a datasheet for what I guess is a re-issue of the ScanSpeak tweeter. I traced the response and overlayed it to your measurement. The difference is enough to challenge you to make a filter for it so it's important to find out what matters and what doesn't.
If it's a reflection then the answer is clear. If it's actual diffraction then it's incomplete without the other angles.
However, sometimes in this position it's better to work from the factory data in a temporary sort of way, because it may be a better average representation of what you have. If you do then I've shown this here scaled back by -16.5dB and -20.5dB.. I don't know exactly so you may have to play with the level a little if you do this.
If it's a reflection then the answer is clear. If it's actual diffraction then it's incomplete without the other angles.
However, sometimes in this position it's better to work from the factory data in a temporary sort of way, because it may be a better average representation of what you have. If you do then I've shown this here scaled back by -16.5dB and -20.5dB.. I don't know exactly so you may have to play with the level a little if you do this.
Yet another issue has surfaced and that's delay. I want to ask whether you're sure you set your woofer and tweeter measurements in synch with each other.
Be advised that sometimes people don't manage this with their measurements and reset their phase and ask vituixcad to do it for them.. so you may hear about that. It's useable in a pinch but it's less accurate, and it can be double dealed so you want to make a choice.
Apart from that, using the ScanSpeak factory data things fall into place a little better. (Later I'd want to see whether it can be done with a little less reduction in impedance)
Be advised that sometimes people don't manage this with their measurements and reset their phase and ask vituixcad to do it for them.. so you may hear about that. It's useable in a pinch but it's less accurate, and it can be double dealed so you want to make a choice.
Apart from that, using the ScanSpeak factory data things fall into place a little better. (Later I'd want to see whether it can be done with a little less reduction in impedance)
When doing measurements, I elevated the speakers so that tweeter level was approximately midway between floor and ceiling.
What I found a bit difficult, was achieving consistency in distance (difficult to get within +/- 2-3mm) when moving mic up or down on the different driver axis. So this may be a source of error.
If you can somehow obtain factory plots, that would be great!
Right, I understand what you are saying now.
I agree with you Allen, I am missing the point. Doing this measurement procedure and simulation first time, I find it's alot of information to take in.
Could you please explain how you would measure the two woofers together and how to use this measurement in the simulation?
Would you then import only one instance of the woofers, tweeter + woofer instead of tweeter + woofer + woofer?
This is all very useful information for me Allen, I'm eager to learn!
The poor measured response that I got from the D2608/91300 was the reason I wanted to try another tweeter, the Morel.
I have included +60 to -60 degree angles for the D2608 in the frd folder, do I have to go all the way out to 90 degrees?
I read this article by S.Linkwitz to get a better understanding of diffraction, and his examples look very much like the response anomalies I got in my measurements; https://www.linkwitzlab.com/diffraction.htm
I did measurements from many different distances and the tendency was consistent until I got very close to the tweeter, like closer than 5cm..
Can I just use more smoothing on my measurements, like 1/6 or even 1/3, to get a more usable result?
Last edited:
I used semi-dual connection measurement with loopback in REW, so the phase and delay should be ok. For this reason I didn't change the Z-position of drivers in Vituixcad. According to my on-axis measurements, the woofers lag behind the D2608 tweeter by approximately 20mm.
The only issue I can think of, is the possible inaccuracy in distance mentioned above.
Did you lock the woofers to the tweeters?
Yes you did, you changed the Y position and that changes the distance/delay/phase 😉
There are many compromises when measuring, and moving the mic is one I don't like to do unless I have a repeatable method of realigning it. Most of the time it is the speaker that you should be moving and not the mic.
Consider this single compromise as an example of prioritising what is important. If you measure the woofer after not moving the mic from the tweeter axis, it will be a few degrees off. However you aren't likely to operate them at the higher frequencies where that makes a difference in response. After all you are building an MTM where the goal should be to cross as low as possible. In any case preferrably below where the woofer is near the same on all axes. So in this case it hardly matters.
When I told you about measuring the woofers together I was hoping at the time to help you see the bigger picture. There are many different measurement and design techniques and some are rather involved.. but they are also different. So it's important to pin down how far you want to go this time around.. On one hand you could get the factory data, simulate the effect behind the baffle step, then find some filters for it.
How interested are you in knowing how much the third dimension affects those filters? I see you are cloning a design so the acoustic design aspect afforded by going deeper into vituixcad is going to go underutilised in this case.
Allen, I see that alot of my questions get lost and not replied to, possibly because we make several posts at a time 🙂
If by lock the woofers to the tweeter you mean did I use the same mic distance, then yes I tried to do this.
Or do you mean something else?
You are right of course. So if I measure all drivers on their axis, I should leave all X,Y,Z positions to 0 right?
It makes alot of sense to make all measurements on tweeter axis. But, would it make even more sense to make the measurements at listening height?
The LV tweeter is 83cm above floor level, meaning about 10cm below typical listening height..
Ok, I'll try to catch up 😉
Lock your measurement software or otherwise ensure the gate starts at the same for both.
It depends on whether you are doing a measurement scheme that uses them or not. In post #27 I'm trying to pin you down on whether you want to simplify the process or continue at the more intermediate level. I'll can't answer these questions until I know whether you want to do the factory data version or a partial sim version or something else.
Yes, I'm expecting you'll either discover why the measurement went that way or do something else. What you say about the close measuring distance is interesting because outside influences would be likely to cause a variation as you change distance.
Linkwitz is a trusted name. Don't forget he's talking about dipoles. Some similarities still exist.
Smoothing is a maybe.. but not better than the other options.
Okay so going back to post #27, I understand how measuring both woofers together can be beneficial. This will capture their interaction better than measuring them separately and leaving the simulation software to predict the interaction.
So to start from scratch, you would place the mic permanently on the tweeter axis (which is offset relative to the woofers) and make one farfield measurement for the tweeter and one for both woofers connected in parallell? I have to ask again about this from post #29;
It makes alot of sense to make all measurements on tweeter axis. But, would it make even more sense to make the measurements at listening height?
The LV tweeter is 83cm above floor level, meaning about 10cm below typical listening height..
I thought I had discovered this with my reference to the diffraction article, do you think there are other factors at play?
This would also explain why the measurement smooths out when moving closer to the tweeter I guess.
The Linkwitz measurements were done with a tweeter on different baffle sizes and shapes, so no connection to dipoles in this case.
With the ScanSpeak + woofer, try working with the 35 degree angle to begin with. This averages better.
Are you suggesting I should listen to the speakers 35 degrees off axis, or do you mean that this represents in room response better?
So I should disregard the 6-7khz peak, or does this require a notch filter?
I've found that listening to the speakers (with active crossover for the time being) more than 5 degrees off listening axis sounds dull and unfocused.
So I should disregard the 6-7khz peak, or does this require a notch filter?
I've found that listening to the speakers (with active crossover for the time being) more than 5 degrees off listening axis sounds dull and unfocused.
It helps you find what you should do with them.. without the diffraction making it difficult to see. You did have problems earlier in this thread because of the shape of the response.
I'm not saying you should listen at 35 degrees. Maybe you could but you will probably choose less. It doesn't matter if you find it dull and unfocussed because you haven't crossed it and equalised it properly yet so you don't know.
Should you ignore the peak? No, but I would not fully notch it.
I'm not saying you should listen at 35 degrees. Maybe you could but you will probably choose less. It doesn't matter if you find it dull and unfocussed because you haven't crossed it and equalised it properly yet so you don't know.
Should you ignore the peak? No, but I would not fully notch it.
Got it Allen 🙂
Let's try to nail down the basics of my measurement scheme and driver positions in Vituixcad before we move on, as discussed in my pm to you 🙂
Let's try to nail down the basics of my measurement scheme and driver positions in Vituixcad before we move on, as discussed in my pm to you 🙂
What is important is that you understand this and use it when you develop your measurement scheme.
Lament, I dont believe should take it literally when Dcibel says it's not good practice, I think this was said in haste to show you that you were missing something. Understand that the delay is always something for you to watch, and you can make it work any way you choose to measure. For example, the problem is not fixed even if you measure at each driver's axis because thats not the listening distance either. Unless you plan to simulate the distances. One way is to use virtual driver placement options. Therefore Dcibel's way works for his method and not necessarily for all methods.
I'm not sure this method is what you are looking for today. I simply tried to help you see the bigger picture because I have seen when people read step by step instructions they still have many questions. So when it becomes clear to you then you will branch out and find better ways, but until then you might want to take a simpler approach.