Theorized and proven: "Zingers", an InDIYana 2019 8" 2-way design...

[dcibel]

Nice project Wolf, and a good result for an 8" 2-way with 5 components!

I hope you don't mind me adding this in here. I was looking at all the squiggly lines trying to figure out which was which, and getting confused that your far field right response looks too similar to the near field left plot. Anyway, I found the file names that you posted which cleared a few things up for me. This isn't really a criticism of your project, just something I decided to play around with because I can, and you had provided a lot of squiggly lines to play with.

It strikes me as odd that you took a nearfield measurement, but then didn't use it in your simulation, it appears that a far field "blended" plot was used in simulation, IMO this makes it hard to judge how much BSD you are adding in, but I'm sure you have your reasons and methods that work for you. I had a half hour to kill, so I traced your data with this neat online tool and loaded it into SoundEasy. I merged your nearfield data with a BSC sim with the far field data at 600Hz, so you can see the low end response a lot better. To be clear, I checked that the transfer function and impedance is a match, the only real difference here is the woofer frequency response data (I used the left speaker data).



I think what you have here is very good, especially the fact that there is a bit of a dip in the 3kHz region, this will help compensate for the "directivity bloom" from crossing a 8" woofer to a 1" tweeter. There is a bit of a lump in the response at 1kHz that doesn't show up really in your response simulation above, but it is evident in the filter transfer function. I think you made good compromises here, as you'd have to go over the 5 component rule to improve this further IMO.

For reference, this is the raw woofer "merged" response:

[Wolf]

Phil, yes, I got lucky on the plywood veneer again. This board had scuffs and marks on it. A little sanding is all it took to make it pretty again.
Eric, yes MWAF they will go. Under 200, I suspect.
Djg, I hope it is a hit! They really do sound much better than I anticipated.
Wogg, it is possible the other tracks would have bettered their standing. They do dig quite well.
Nick, we seem to agree more often than not.
Dcibel, yes, you hit the nail on the head as to what I did and why. This is my 3rd 8" 2-way, so I made some clever assumptions. The plots are however, gated responses without any blending. I use the time of flight phase from OM, and don't like the Blender program. Plots are on-tweeter-axis, 20deg H off-axis, and HD. The NF was only to get an idea of the bottom roll-off. Both channels have some dips in the right places, and off axis does not suffer one bit. Mind you, I do prefer the 6 part network, and I'm fine with your playing around. I would have given you the data files had you asked.

Thanks for looking!
Wolf

[Wolf]

Oh, and the 1k peak is in one channel and not the other. Like I said, the woofers measured differently. Even so, I feel the parts used reflect the best out of both responses. It's a good average compromise of the 2. Dcibel modeled the other channel than I did, but both look pretty good.

Later,
Wolf

[dcibel]

Dcibel, yes, you hit the nail on the head as to what I did and why. This is my 3rd 8" 2-way, so I made some clever assumptions. The plots are however, gated responses without any blending. I use the time of flight phase from OM, and don't like the Blender program. Plots are on-tweeter-axis, 20deg H off-axis, and HD. The NF was only to get an idea of the bottom roll-off. Both channels have some dips in the right places, and off axis does not suffer one bit. Mind you, I do prefer the 6 part network, and I'm fine with your playing around. I would have given you the data files had you asked.

What I meant is that you used the "blended" response from the Omnimic software, not the "only to" setting, which blends the gated response with the ungated response at low freq. You know you can splice the near and far responses within the Omnimic software right? Check the "how to" section of the help file, it's actually quite easy. You just load the 2 measurements into the "added curves" section of the software, then set the frequency limits so that the responses overlap just a little where you want them to be merged. Then offset the SPL to line then up, and add delay to the near field to line up the phase (optional, I would still recommend HBT for best results). When there are only 2 responses loaded to the "added curves", you will see a "save splice" button available on the curve settings dialog. Of course I merged the responses in SoundEasy and applied HBT (tails) to generate minimum phase, but the end result I think is not different enough to care about for this purpose.

(just nitpicking, but Omnimic doesn't do "time of flight", it does some "quasi minimum phase" using the peak of the impulse response as the reference. 2 channel measurement is required for any time of flight data)

Oh, and the 1k peak is in one channel and not the other. Like I said, the woofers measured differently. Even so, I feel the parts used reflect the best out of both responses. It's a good average compromise of the 2. Dcibel modeled the other channel than I did, but both look pretty good.

Yes, unfortunate that your woofers are so much different from each other. Looking at them closer, the left speaker that I modelled looks a lot like the original SD215-88 response, where the right speaker looks more like the SD215A-88 response, but even then the breakup is shifted lower in frequency.

[Wolf]

I don't know if my OM does all this. I have one of the original clear-case V1 units, so it's one of the first sold group. My impression of phase type was what I read from others, so I had no idea it was not ToF info.

For kicks, I just looked at the DS215-8, and it will not work in this design. I just wanted to make that distinction. The DS and SD are very different drivers.

Later,
Wolf

[4thtry]

Ben,

Your comments about the tweeter sound in post 6 is exactly what I heard at the event. If adding the 33 ohm resistor, as you suggest, solves this problem, then this should be one great sounding speaker at MWAF.

[Wolf]

It's kinda subtle unless you focus on it, but it was very apparent that it was required. Source material that highlights this issue make it stick out like a sore thumb. As you know, we listened to them with the resistor on Friday, and I flipped the switch off for the contest. 6 parts for the win on this project!

Thanks, Bill,
Wolf

[4thtry]

I did not hear them on Friday, someone mentioned that they played earlier in the morning before I got there. Look forward to hearing them with the 6th part!!!

[dcibel]

I don't know if my OM does all this. I have one of the original clear-case V1 units, so it's one of the first sold group. My impression of phase type was what I read from others, so I had no idea it was not ToF info.

For kicks, I just looked at the DS215-8, and it will not work in this design. I just wanted to make that distinction. The DS and SD are very different drivers.

Later,
Wolf

I have a V1 Omnimic as well, just update to the latest software...unlike the DATS, Omnimic software is free. Realistically, anyone can use it to view and manipulate FRD files, such as merging data and generating polar plots.

http://www.daytonaudio.com/downloads...ll_v5_2019.zip

Remember that OM is just a USB mic, because it's a "single channel" measurement, there is no point of reference of when the electrical signal arrived at the speaker. In the case of the Omnimic software, the start of the FFT is automatically set by the peak of the impulse, which gives you minimum phase-ish response. A 2 channel measurement system is required for any accurate time of flight data. It is specifically this distinction, that is the reason why you have to complete some extra measurements to determine acoustic offset between drivers when using an Omnimic. With a 2 channel measurement, you can simply look at the timing differences between 2 measurements, and for design purpose, simply keep the FFT start at the same time reference, so one measurement will include the excess phase of one driver being further away than the other. Since I have both Omnimic and a EMM-6, I use the OM for room setup, EQ and those sorts of tasks, but for the speaker design itself I prefer the 2-channel setup.

[4thtry]

I have a V1 Omnimic as well, just update to the latest software...unlike the DATS, Omnimic software is free. Realistically, anyone can use it to view and manipulate FRD files, such as merging data and generating polar plots.

http://www.daytonaudio.com/downloads...ll_v5_2019.zip

Remember that OM is just a USB mic, because it's a "single channel" measurement, there is no point of reference of when the electrical signal arrived at the speaker. In the case of the Omnimic software, the start of the FFT is automatically set by the peak of the impulse, which gives you minimum phase-ish response. A 2 channel measurement system is required for any accurate time of flight data. It is specifically this distinction, that is the reason why you have to complete some extra measurements to determine acoustic offset between drivers when using an Omnimic. With a 2 channel measurement, you can simply look at the timing differences between 2 measurements, and for design purpose, simply keep the FFT start at the same time reference, so one measurement will include the excess phase of one driver being further away than the other. Since I have both Omnimic and a EMM-6, I use the OM for room setup, EQ and those sorts of tasks, but for the speaker design itself I prefer the 2-channel setup.

Isn't it true that OmniMic raw phase data, even though it is single channel and not a true minimum phase or time of flight calculation, that it is nonetheless consistent from measurement to measurement. And because it is consistent from measurement to measurement, it can be used to calculate the "mod delay" accurately in XSim using the "get file" procedure. When using raw phase data to develop my crossovers in XSim, I always get a very good match between my model and my final measurements.

[djg]

This is a very odd sized cab, at least in the world of Craigslist junk. I'm going to make a "faux vintage" cab with some Sapele PSA veneer I have over particle board. The slightly recessed baffle and the back will be flat black. The picture frame front veneer edges will be about 1/8" proud, hopefully enough to look like an old 70s cab. I ordered parts with a $10 discount offer.

[Wolf]

The main reason for the shape I used was determined by the size I could cut out with a 2' x 4' handipanel, and how everything fit/looked in the cab. HWD 18.5" x 11.5" x 12" just worked well. Are you going for the full 1.5ft^3, or just the 1ft^3 like I built?

Later,
Wolf

[djg]

One cube. I'm going to see if these work on a desktop, if I enlarge my desk and get some distance. My other 4" and 5" two ways have a 4 1/2" to 5 1/2" driver center to center. Imagine that illusive desktop system that doesn't need a sub. Maybe a tad shorter and deeper to get the cab up off the desktop a bit.

I haven't found any junk cabs with a desired baffle width that had even one cubic foot of volume.

[dcibel]

Isn't it true that OmniMic raw phase data, even though it is single channel and not a true minimum phase or time of flight calculation, that it is nonetheless consistent from measurement to measurement. And because it is consistent from measurement to measurement, it can be used to calculate the "mod delay" accurately in XSim using the "get file" procedure. When using raw phase data to develop my crossovers in XSim, I always get a very good match between my model and my final measurements.

You're right, it's not technically minimum phase, my understanding is that it uses the peak of the impulse to determine the timing reference. It may be "good enough" in many cases, just keep in mind that it is a 48kHz measuring system, being off by a single sample (1/48000th of a second) for the start of the FFT represents +/- 38 degrees phase error at 5kHz, or +/- 15 degrees at 2kHz. My process would always be to apply HBT (tails) and extract minimum phase for maximum accuracy, provided the HBT tails are selected well.

I'm not very familiar with Xsim,so I don't know this "get file mod delay procedure", but with a single channel measurement system I would complete the 3-measurement process that has been put forth by Jeff B and others to determine the acoustic offset necessary. This method removes any error in sampling, FFT window start, etc as you end up adjusting an offset to reproduce a measured acoustic interaction between the drivers, however also has a possibility of human error in the correct steps to complete the process.

[jhollander]

The "get file" is the overlay in Xsim, which is the same as the three measurements, (1) Tweeter, (2) Woofer, and (3) Tweeter plus Woofer, where the Tweeter plus Woofer is the overlay.