Senior Project: Chimera, 3 headed beast

[Blazin]

You have to start looking at the phase to get a good blend. In a perfect world they would be in phase at all frequencies or at least through the crossover region. You might need to adjust the XO to compensate for acoustic center offsets -- the AC of a cone driver is behind the baffle. And that's assuming you measured phase right in the first place which is a BIG assumption. It's a delicate balancing act -- SPL, phase, impedance, how they all sum.....

Are we having fun yet?

Wait.. I measured using the same mic position... On Axis with Acoustic Center.. (well the tweeter) but I didn't move the microphone.. so shouldn't the acoustic offset be taken into account in my measurements?

"Note that this is generally not necessary if you did the acoustic measurements from the same point of reference as offset is incorporated into the data. Though using the Acoustic Offset command may “clean up” the phase of the individual driver, it will alter the relative phase measurements between drivers which is the more important aspect of phase measurement."

From the SW manual..

[Dennis H]

I don't have any experience with SW so maybe Chris can help you to get the phase right.

[cjd]

1. Don't work with smoothed data. That phase data looks like it has been smoothed - maybe not.

2. Why are there resistors on the woofer?! Shouldn't be. And be sure you're setting impedance on your inductors.

3. There's no good way to look at phase in SW when it's live data without a transform on it - it really tends to be al over the place. Sometimes. Go into your chart properties and enable phase display on the mid and tweeter, and disable it on the summed response (it's irrelevant). You'll probably need to zoom in on the transition region to see more clearly the phase relationships here. Check JonW's thread for more information on what kind of stuff you're looking for. Note that my example in that thread is data that has been hilbert transformed for phase with relative acoustic offsets set based on comparing the two data sets.

4. Try going to a 3rd order electric circuit on the tweeter. Bump up to a 12uF cap, drop the inductor value a little, start with a 25uF value on the 2nd cap.

5. What's the raw tweeter impedance look like compared to PE's published one? That network doesn't look like it should push impedance nearly that low.

C

[Blazin]

1. Don't work with smoothed data. That phase data looks like it has been smoothed - maybe not.

2. Why are there resistors on the woofer?! Shouldn't be. And be sure you're setting impedance on your inductors.

3. There's no good way to look at phase in SW when it's live data without a transform on it - it really tends to be al over the place. Sometimes. Go into your chart properties and enable phase display on the mid and tweeter, and disable it on the summed response (it's irrelevant). You'll probably need to zoom in on the transition region to see more clearly the phase relationships here. Check JonW's thread for more information on what kind of stuff you're looking for. Note that my example in that thread is data that has been hilbert transformed for phase with relative acoustic offsets set based on comparing the two data sets.

4. Try going to a 3rd order electric circuit on the tweeter. Bump up to a 12uF cap, drop the inductor value a little, start with a 25uF value on the 2nd cap.

5. What's the raw tweeter impedance look like compared to PE's published one? That network doesn't look like it should push impedance nearly that low.

C

1.) Oops may have smoothed data.. don't recall.. I'm making measures again so will not smoothe..

2.) Woofer's level is much higher than others? Wait.. setting the impedance.. what do you mean.... ohhh DCR.. yea oops

3.) EDIT found the post.. ok so I'm looking for a crossover point where the phase aligns..? clearly have to measure again.. first to get rid of that 10Khz peak.. and 2nd to unsmoothe my data.. Then Extract Minimum phase in the FRC.. then line up phase to see where they align.. then crossover.. then say delicious.

4.) Will try.. EDIT put that on there.. wow.. huge jump UP.. no impedance issues there.. 3ohms or greater.. but the FR is yuck.. this goes back to phase aligning.

5.) Here black is my measured.

[cjd]

On measuring and phase: DON'T do the minimum phase extraction step unless you're *really* having troubles. If you do, KEEP your original data. To make sure you get the correct offset on the mids and woofer you'll need to compare response with the minimum phase model and the live data to get this right. I don't recommend that as a first pass though. But, this is not a required step. If you work with the live data straight in SW it is quite possible and valid, it's just not always as clean looking. But, you can do other things to adjust response and get there without ever looking at the phase data.

Woofer level circuit: right now the woofer output looks low to me. Get DCR on the inductors - if you really do need less level, go with a smaller gauge inductor or something. But DO make sure you have impedance data set on the inductors set! (DCR, yeah) VERY important on any series inductors in particular.

Impedance on the tweeter, looks fine. Just making sure. I didn't expect my random circuit suggestions to magically integrate. Will take some tweaking. Though, check it with the tweeter wired both ways.

[Blazin]

Ok I'll post what I got so far..

FRDs for Gate Measurements down below.. hmm peak at 2670Hz wonder what that is..




Alright thats to about 7.15 mS @ about 2 meters.. I can increase the distance but it appears the distance increase is messing with my gated measurements.. even though it shouldnt?


Now my phase is all yuck.. saw tooth like.. how am I supposed to line up the phase data?

[Blazin]

Oh I fail.. I've been doing this all wrong.

I must have read Claudio's site 50 times and the SW manual 10 times..

SW Manual:

Measure each individual driver Frequency/ON-AXIS -- I'd been mixing the two..

Then I see if you measure them all hooked up.. GATED you can see where they are in phase with one another potentially..

also..

I'm crossing over at 250Hz on the low end.. so I can use 1000/6mS = 166.67 Hz or a 6 mS gate to have accurate LF results... also I didn't realize that I should be changing my sample size/rates when measuring LF and HF.. ie


SW MANUAL, "Low frequency measurements require a larger sampling size and a slower sampling rate in order to get a reasonable amount of data per octave (resolution). Tweeter measurements do well with smaller sampling sizes and faster sampling rates as they have many more frequency points in each octave so, in order to provide the best noise immunity, you will want to raise the SAMPLE RATE and lower the SAMPLE SIZE for tweeter measurements. "


Wow ok.. this makes a difference lol

Be back in a few with better data.. I'm gonna shoot for a 6mS Gate and move my speaker/mic distance closer to promote noise immunity.. wow ok finally reading and understanding vs reading and saying, yea I think I'm doing this.

[cjd]

Do not move the mic - keep it on-axis with the tweeter for all measurements. It simplifies things significantly. 1.5M should be OK.

You've got a little peak below your stop gate that might be a reflection you need to avoid.

[Blazin]

CJD:

My confusion was that I did not realize all these 'gated' measurements were.. Driver\Measure\Frequency\On Axis..

I had been doing..
Driver\Measure\Frequency\Farfield Gated

Always been keeping mic on axis..

That explains why i was confused which freq response graph needed to be placed in each driver data slot.. because I wasn't measuring on axis response..

I have this data now.. its still a bit messy and phase is wrapping too much..

I guess I need someone to interpret my GATED FARFIELD measurements of total system then individual drivers.. to show me where I am in phase.. the example on Claudio's site and SW manual is kind of clear I think they made a typo tho.. so I'm a little confused..

CJD you willing?

[augerpro]

I just wanted to mention one thing, I hope it doesn't confuse you any more than you already are- the polarity of your impulse responses is negative. The first big peak should be pointed up not down. I suppose it's not a big deal as long as all the drivers are the same in this, but it is still convention to have the impulse displayed that way. Assuming you wiring is all correctly hooked up as far as polarity, there should be a flip the IR in the software since this is common.

Also the gate depends only on the mic distance and the distance to the first reflection (wall, floor, or ceiling) and nothing else. It's not a number you get to pick, it's determined by the room you measure in and the distances involved.

[Blazin]

I just wanted to mention one thing, I hope it doesn't confuse you any more than you already are- the polarity of your impulse responses is negative. The first big peak should be pointed up not down. I suppose it's not a big deal as long as all the drivers are the same in this, but it is still convention to have the impulse displayed that way. Assuming you wiring is all correctly hooked up as far as polarity, there should be a flip the IR in the software since this is common.

Also the gate depends only on the mic distance and the distance to the first reflection (wall, floor, or ceiling) and nothing else. It's not a number you get to pick, it's determined by the room you measure in and the distances involved.

hmm thats odd.. lemme check. yea they all go negative first.


I know, I can alter the mic-baffle distance to achieve a particular gate as I have nearly 25 ft of room to deal with.

[augerpro]

You have a 25 foot ceiling? If so than you can definitely use a long gate.

[Blazin]

You have a 25 foot ceiling? If so than you can definitely use a long gate.

no no Length wise.. prolly 10ft ceiling.. 10 feet wide

[augerpro]

Then that 10 feet is going to limit your gate. Try to center the speaker between teh floor and ceiling- about 5 feet off the floor. Now the closer you put the mic the longer the gate can be, but the more inaccurate phase difference between drivers. Pull the mic back and phase difference gets more accurate but gate length goes down. It's a balance. The important part: The difference between the path length from the speaker directly to the mic, and from the speaker to the floor and then to the mic is what determines the gate. Find this length and calculate how long it takes sound to travel that distance, this length of time is the maximum gate you can use. I'm just guessing but I would think with a 10 foot ceiling and the mic 1 meter away you could probably use a 5ms gate, maybe a bit less. Do the math and find out for sure.

[Dennis H]

Just adding to the confusion, Chris recommended leaving the start of the gate at 0 so you have a long flat spot before the impulse response. I think it's a peculiarity of SW that this WILL give you more precision -- a data point every X Hz and X is smaller if you do it this way. However, it won't give you any more accuracy. The lowest frequency that's actually any good is determined by the time from the impulse response to the end of the gate.

Bottom line, due to the nature of FFTs, you're unlikely to get good data below 300 Hz or so in the farfield unless you make the gate long enough to include reflections. See the other thread about splicing in nearfield data. It's just the nature of the beast called measuring indoors. If you want really good farfield data, go measure in the middle of the football field with the speaker hoisted 15' in the air.

Sorry I can't give any specific help with juggling all this in SW -- I've never used it. Chris is our resident SW guru so he can help you with the workarounds to get useable results.

[cjd]

The phase isn't "bad" it's just... difficult to work with. But that's not at all abnormal.

Feel free to email me some stuff. if chris at eldamar dot net doesn't work try cjdalessio at gmail dot com - no guarantee on how quickly I look at it but hey.

C

[Blazin]

Then that 10 feet is going to limit your gate. Try to center the speaker between teh floor and ceiling- about 5 feet off the floor. Now the closer you put the mic the longer the gate can be, but the more inaccurate phase difference between drivers. Pull the mic back and phase difference gets more accurate but gate length goes down. It's a balance. The important part: The difference between the path length from the speaker directly to the mic, and from the speaker to the floor and then to the mic is what determines the gate. Find this length and calculate how long it takes sound to travel that distance, this length of time is the maximum gate you can use. I'm just guessing but I would think with a 10 foot ceiling and the mic 1 meter away you could probably use a 5ms gate, maybe a bit less. Do the math and find out for sure.

Ok so whats optimal..

I have these dimensions: Width 11ft Height 7-8ft there are some differences in the height (theres a foot deep channel down the center It's a garage) Lenght well thats the longest dimension easily double 11ft

so.. 15" wide x 42" height x 16" depth speaker..

4.5ft on either side.. (Tweeter is currently 43" off ground with speaker on a milkcrate ) so that height.. Mic distance of 1 meter.. with a buffer of 10ft on either end of the length?

Hope someone can follow that..

[Blazin]

Just adding to the confusion, Chris recommended leaving the start of the gate at 0 so you have a long flat spot before the impulse response. I think it's a peculiarity of SW that this WILL give you more precision -- a data point every X Hz and X is smaller if you do it this way. However, it won't give you any more accuracy. The lowest frequency that's actually any good is determined by the time from the impulse response to the end of the gate.

Bottom line, due to the nature of FFTs, you're unlikely to get good data below 300 Hz or so in the farfield unless you make the gate long enough to include reflections. See the other thread about splicing in nearfield data. It's just the nature of the beast called measuring indoors. If you want really good farfield data, go measure in the middle of the football field with the speaker hoisted 15' in the air.

Sorry I can't give any specific help with juggling all this in SW -- I've never used it. Chris is our resident SW guru so he can help you with the workarounds to get useable results.

Hahah and he said I don't need nearfield data.

[Blazin]

The phase isn't "bad" it's just... difficult to work with. But that's not at all abnormal.

Feel free to email me some stuff. if chris at eldamar dot net doesn't work try cjdalessio at gmail dot com - no guarantee on how quickly I look at it but hey.

C

Sent on over..

[Blazin]

I sent you crap CJD..

I didn't have the latency set properly on my computer at home.. explains all the garbage in my measurements.. today's 1M on axis/gated produced good results.


I think I found the crossover point for my Tweeter/Mid



Relative phase at 2.1K = 0* this is from the Farfield Gate response of the two drivers.


And if I'm reading this properly Sub/Mid at 250Hz is ~90* out of phase..

[Blazin]

I think I got it!

Yea!!! I appear to be in phase at both crossover regions.. I had to pad the tweeter because it was higher than LF to start.. (was about 6dB higher.. Baffle?)

Someone tell me I'm good so I can order this stuff..








NOTE: Freq Response is limited due to the fact that I didn't merge the nearfield for a better look.. but its ok, where the gate becomes worthless, the nearfield lowend shows steady amplitude til a roll off begins at 40ish Hz..

[cjd]

Phase tracking does seem better, but... it won't sound too good. Big hump at 500Hz to a drop trough - that's all going to be quite audible.

I just want to confirm, you are including DCR on your inductors now? If so, you do have an issue with too much energy on the mid still, even in series. You'll either need to consider padding them down, or finding higher DCR inductors to use on the mid-woofers (I recommend the inductor route first). That'll force a change in slope somewhat, though.

Actually, looking over the data again: Let's see if you can eliminate the second large inductor value on the woofers (I hope you're using low DCR inductors here - go cored if you don't want to spring for an expensive air-core for this project). Also, consider raising the crossover frequency slightly. The goal here is to bring that level up so you can NOT pad the tweeter down as much and bring the overall level up more.

If you can, I'd also like to see the tweeter level up a little on its low end. What happens if you drop the second cap to 20uF?

I did see files show up at my gmail address at least, didn't get a chance to look at them last night.

C

[Blazin]

Phase tracking does seem better, but... it won't sound too good. Big hump at 500Hz to a drop trough - that's all going to be quite audible.

I just want to confirm, you are including DCR on your inductors now? If so, you do have an issue with too much energy on the mid still, even in series. You'll either need to consider padding them down, or finding higher DCR inductors to use on the mid-woofers (I recommend the inductor route first). That'll force a change in slope somewhat, though.

Actually, looking over the data again: Let's see if you can eliminate the second large inductor value on the woofers (I hope you're using low DCR inductors here - go cored if you don't want to spring for an expensive air-core for this project). Also, consider raising the crossover frequency slightly. The goal here is to bring that level up so you can NOT pad the tweeter down as much and bring the overall level up more.

If you can, I'd also like to see the tweeter level up a little on its low end. What happens if you drop the second cap to 20uF?

I did see files show up at my gmail address at least, didn't get a chance to look at them last night.

C

Yea that stuff I sent you was with the latency issues... it was pretty much garbage.. I could send you new files, I'm getting the phase tracking.. basically you want the crossover to be in phase.. a 4th-3rd sub-mid slope was 180 out of phase.. so I changed the mid to a 2nd and voila closer a little tweaking and I think I got them in phase.

Hmm I am using DCR values now, I tried to use consisten practices (all 18awg) btw I'm loading DCR values from PE.. is that what I'm supposed to be doing?

EDIT:

Oh now that I re-read and looked at the freq again, I see what you're saying.. leave the LF alone, try to raise HFMid response, unpad tweeter some.. should then remove that trough thats occuring after 500Hz. Good advice was a little groggy this morning.. (I still don't want to cross the tweeter too high though correct? aka 2.5K)


The tweeter was 6db+ higher than the mids/sub on axis response

I just got to work so I'll need to wait on these til later.. but I'm pretty sure I'm understanding the phase relationship and how to choose the cross frequency.. I'm seeing a lot less TOTAL under the individual responses.. :lol:

[cjd]

180 degrees out of phase, simply reverse polarity on one of the drivers! Magic, it's fixed.

[Blazin]

180 degrees out of phase, simply reverse polarity on one of the drivers! Magic, it's fixed.

Yea yea.. lol that doesn't fix the high end issues :B

[Blazin]

Well I got the mid/tweeter SET for sure..

Totally in phase at 1.781KHz hook up out of phase -40dB sink

awesome..

Now I gotta do mid/sub.. thats hard the sub phase is not exactly super.. I tried merging in nearfield to get an idea.. I can get something that looks good, go back to farfield only and its terrible (because of the roll off soo soon in farfield measurement)

Should I be using the gate.splice? If so.. what do I do about phase.. its not that accurate with a splice..

[cjd]

The question is whether the relative phase is 100% accurate in the farfield measurements such that you can ignore the appearance of a roll-off, if your nearfield data proves that the roll-off won't actually be there and a problem.

[Blazin]

The question is whether the relative phase is 100% accurate in the farfield measurements such that you can ignore the appearance of a roll-off, if your nearfield data proves that the roll-off won't actually be there and a problem.

Yea nearfield shows it wont be there.. so cross farfield data without splice.. use phase data, ignore the roll off?

[Blazin]

Alright beat me with this one.










I left the phase on there so you can see they are all pretty well matched... within crossover regions.. whaddya think?

[cjd]

Yeah, you're getting it.

So, I'm still a little concerned you may have too much midrange energy, and a little much on the top of the tweeter. But it's also at a point where your'e getting close - if you want to keep at this stuff as a hobby, it's time to build and keep tweaking, but to HEAR what some of the changes do. IMO.

C

[Blazin]

Yeah, you're getting it.

So, I'm still a little concerned you may have too much midrange energy, and a little much on the top of the tweeter. But it's also at a point where your'e getting close - if you want to keep at this stuff as a hobby, it's time to build and keep tweaking, but to HEAR what some of the changes do. IMO.

C

I don't have the luxury of really building and tweaking on this set.. $$$ I'm a poor college student- haha..

what's the return policy on components haha

[Blazin]

I don't have the luxury of really building and tweaking on this set.. $$$ I'm a poor college student- haha..

what's the return policy on components haha

900Hz is Baffle Step..? I have 15" Baffle The wavelength of those Mids.. and they are centered on the baffle..

BSC?

[cjd]

If these are in-room gated measurements, they should automatically include baffle step.

If you're not wanting to put out the cash, keep fiddling. Experiment with topology some. You've got data and are picking up the working knowledge to make this practical. You're already looking at something that is +/-3dB which many consider perfect. So, it depends how much of a perfectionist you are, and how much more experimentation you want to do as far as fiddling and learning from that.

C

[Dennis H]

I don't have the luxury of really building and tweaking on this set.. $$$ I'm a poor college student- haha..

what's the return policy on components haha

I hear you but you really can't finalize the design until you build it and see if it measures and sounds like your sims predict. (It would be a miracle if you nailed it on the first try but stranger things have happened. ) You can do things like build the caps out of several smaller ones with a few extras for trimming on your first proto. The Madisound surplus GE caps are a good buy at $.75 for a 5uF and $.30 for a .68uF.

http://www.madisound.com/manufacturers/ge/ge_caps.php

Once you get the values finalized and see what parts you have left you can order parts to add to what you have to get the final values for the second speaker. There shouldn't be many wasted components that way. Do you have access to an LC meter to check values so you get both speakers the same?

[Blazin]

If these are in-room gated measurements, they should automatically include baffle step.

If you're not wanting to put out the cash, keep fiddling. Experiment with topology some. You've got data and are picking up the working knowledge to make this practical. You're already looking at something that is +/-3dB which many consider perfect. So, it depends how much of a perfectionist you are, and how much more experimentation you want to do as far as fiddling and learning from that.

C

Heh I think I figured it out!












So what do you call that.. nominal 8 ohms?


This is with the merged nearfield spliced in..

[augerpro]

Why does your impedance ahve such bad resolution and only extends to 100Hz? you should be able to use a very large gate (>200ms) and get a high res plot down to single digit frequencies. My concern here is that the XO CAD may be inaccurately modelling the response because the resolution of the impedance is too coarse. This might result in a frequency response quite different than the model predicts. Just a thought.

[Blazin]

Why does your impedance ahve such bad resolution and only extends to 100Hz? you should be able to use a very large gate (>200ms) and get a high res plot down to single digit frequencies. My concern here is that the XO CAD may be inaccurately modelling the response because the resolution of the impedance is too coarse. This might result in a frequency response quite different than the model predicts. Just a thought.

It's based on the resolution of the On Axis response.. since the resolution isn't great at low frequencies it 'guesstimates' the impedance at those low frequencies.. ie it does some sort of best fit line I believe..

I agree the chart missing that stuff is kind of an issue, so spliced far field adds in some resolution or at least bandwidth.. and goes down to 20Hz.. its not totally accurate as there should be 2 humps due to the vented enclosure.. but showing it shows ~4 ohms down to 20 Hz.. I just need it to not be below 3ohms.. as my amp can do 3 ohms

[augerpro]

Hmmm...Gating for Impedance measurements has nothing to do with frequency response measurements. Impedance measurements don't care about room reflections so you can make the gate as big as you want.

[Blazin]

Hmmm...Gating for Impedance measurements has nothing to do with frequency response measurements. Impedance measurements don't care about room reflections so you can make the gate as big as you want.

I understand this, I'm saying.. I'm not gating the impedance.. thats how SW works.. if someone knows how to change it so it had better resolution, let me know.. The resolution for the crossovers and the frequency bandwidth is deteremined by the gate of the farfield measurement.. it doesn't display impedance data with any more data points than the FResponse..

The individual measured impedance doesn't look anything like what I've posted, it has lots of resolution and is 20-20k.. I don't know why SW re-samples the impedance from the measured impedance charts..

[Blazin]

So -project presentation question..

What do you think of me constructing the crossovers and mounting them on the back of the enclosure.. So they are visible during presentation and other events I have to show them at.. then when all done, mounting them inside for finished look..

Obviously if I'm going to mount them externally I'd try to make them a little neater and presentable.. maybe actually paint the mdf/pegboard they are mounted on-hah-

[cjd]

Lotsa people make a box they put underneath the speaker.

[Blazin]

Lotsa people make a box they put underneath the speaker.

A Plexi box?

[cjd]

Like a foot or base. Plexi top perhaps, but wood sides to the whole tower permanently sits on top of it.

C

[Blazin]

Like a foot or base. Plexi top perhaps, but wood sides to the whole tower permanently sits on top of it.

C

I'm confused how does that allow people to see the crossover components..

Maybe I'm thinking about it wrong..

[Dave Bullet]

The individual measured impedance doesn't look anything like what I've posted, it has lots of resolution and is 20-20k.. I don't know why SW re-samples the impedance from the measured impedance charts..

I believe when doing crossover modeling, SW will reduce the impedance resolution and range to match the farfield FR measurement of the driver.

One thing I'm not comfortable with is your measurement resolution below 500Hz for your woofer/mid crossover design. It looks far to coarse to me to adequately get good phase and FR data for good lower midrange phase alignment. You want at least 1 if not 2 octaves either side of "good enough" resolution. If you can't get your gate large enough, you might need to resort to nearfield measurents and hilbert transforms with relative acoustic offset entered in SW for woofer and mid. Chris knows better than I here so take his advice.

David.