HOLMImpulse Measurements, Need step by step dummy guide.

[penngray]

Im using HOLMimpulse to measure my new DIY build and Im just confused on what to really do.

Generally, I use a DCX, I get some simple calcs for The CTC from experts, I plug in numbers, set crossovers. Measure with ARTA to get a smooth response in room and it ends there.

This time I want to do it better.

So I would love a list for what I should be measuring then what do I do with the measurements.

My build thread is over on AVS but a great deal of expertise is here.

Here are my first measurements, tweeter and woofer done independantly but show on one chart.

Image not available



The measurements are in room but I have 4" thick acoustical panels behind the speaker and on the floor infront of the speaker to help with reflections. I believe the measurements are gated too.

The speaker being measured is this speaker..

Image not available

Its the QSC Waveguide/BMS 4550 CD, TD12M woofer.

Edit by moderator to add link to HOLMImpulse users guide. Note it's a 18 meg pdf

http://www.holmacoustics.com/downloads/HOLMImpulse/HOLMImpulseUserGuide.pdf

[Saurav]

Are you planning on using the DCX, or are you looking for a passive crossover? Are those measurements of just the raw drivers, i.e. the DCX (if it's in the signal chain) set to 0 everything?

Also, with HOLM Impulse, the default setting moves the start gate for each measurement. You have to change that to lock down the 0-reference, and then measure both drivers using that setting, otherwise the relative phase will be wrong. The ARTA demo makes this a little easier, and you can export FR measurements using the demo.

I'm not one of the forum experts by any stretch, but I just bought LspCAD, and I can try to work up a crossover design for you if you like. No promises on how it'll sound though

[penngray]

Are you planning on using the DCX, or are you looking for a passive crossover? Are those measurements of just the raw drivers, i.e. the DCX (if it's in the signal chain) set to 0 everything?

Everything is off except the high pass and low pass filters for the tweeter and woofer. I ran one measurement with the woofer muted and then with the tweeter muted.

At First I will use the DCX but I would love to try and build some passive crossovers at some point!

Also, with HOLM Impulse, the default setting moves the start gate for each measurement. You have to change that to lock down the 0-reference, and then measure both drivers using that setting, otherwise the relative phase will be wrong. The ARTA demo makes this a little easier, and you can export FR measurements using the demo.

I will have to figure out what you just posted, DUMB would be my middle name right now. Everything is exponentially harder in the past couple of days, nothing is clicking.

I have ARTA too so I could go and use that although Im really liking the ease of use HOLM has. Its too easy so far and it has a decent support thread going over on DIYAudio. I have never had luck with finding great ARTA support .

I'm not one of the forum experts by any stretch, but I just bought LspCAD, and I can try to work up a crossover design for you if you like. No promises on how it'll sound though

Any help would be awesome!! I have Xover pro (from Parts express) I have been playing with it but I need to import the measurements. First I need to make sure Im doing all the right measurements.

[penngray]

I should mention that these are the drivers I sent to Augerpro for measuring so he has the accurate measurements of them. We could just take them and built the crossovers.

I still need to learn something and hopefully when Augerpro builds is passive crossovers I will build a set with his drivers/XOs too just to compare against my higher cost drivers.

His will probably still sound better because he is far better at this then I am

Next build is using the drivers/XOs found here

http://techtalk.parts-express.com/showthread.php?t=215536&page=5

[Saurav]

Everything is off except the high pass and low pass filters for the tweeter and woofer.

That makes more sense OK, let's take a shot at this. First, go through all the pages of all inputs and outputs and make sure everything's turned off. All crossovers, all EQ, all delays. Then, measure both drivers. Are you using a calibrated microphone? If you are, have you entered the mic calibration file into HOLM Impulse / ARTA?

If you want to use HOLM Impulse, here's how I have mine set up, I'm not sure if it's the best way or not. On the first tab - I've tried MME and ASIO drivers, not really sure what the differences are. Make sure the input and output are correctly set to your soundcard. Make sure the Measurement signal is set to Logarithmic sine sweep.

On the 2nd tab (data analysis), make sure 'detect time zero' is selected. Then, go to the 'measurements' tab and measure the tweeter (make sure the DCX XO is disabled too). After taking the measurement, come back to the 'data analysis' tab, and select 'time zero locked', then click on 'use' next to 'last detected'. Then go back and measure the tweeter again, and make sure nothing looks weird.

If that works, then measure the woofer. Hopefully that will work too, and will look OK. If it looks weird, then repeat the process, but this time start with the woofer. So set it back to 'detect time zero', measure the woofer, switch to 'time zero locked', 'use' the last detected value, then measure woofer and tweeter, and hopefully both look good.

I don't fully understand how HOLM handles this, so if you're comfortable trying ARTA, I have a better handle on how to do this using ARTA.

[penngray]

Thanks, I will try it in the morning when Im back home.

When measuring I do not want to run HOLM full sweep for the waveguide do I?

Yes, I have calibrated mic (professionally done with file).

[Saurav]

You can pick the starting frequency for the sweep. And yes, don't start at 10Hz for the tweeter.

[Dennis H]

On the 2nd tab (data analysis), make sure 'detect time zero' is selected.

I was going to mention this in your thread but this may be a better place. I haven't tried Holm so take this with a grain of salt. It looks to me like the time zero is too late in the measurement you posted. It's cutting off the first part of the impulse and might explain why the frequency response looks different than ARTA after you do the FFT. Can you manually set it before the first wiggles in the impulse, say maybe 1/3 msec earlier, with Holm?

[Saurav]

Thanks for chiming in. I'm hoping I'll learn more from this exercise than I actually contribute.

Yes, you can set it manually. The easiest way is to do what I described - let it detect the impulse, then switch to setting it manually, and then enter a number that's a little smaller than what it detected. The number you enter is in samples, and you can set the time axis to show ms / cm / samples, so it's easy to pick the sample value you want. However, I tried this once, and got weird measurements, and then didn't mess with it much after that. I probably moved the 0 point in the wrong direction.

You can also set it to detect the first peak, largest peak, first positive, largest positive... or something called 'causal impulse (analog / IIR filters)'. I think the manual said something about pre-ringing for that option, so that might make it pick up the wiggles. I'll try that next time and see if it picks 0 to be before the wiggles. penngray, you might want to experiment with these options as well.

Oh, one more thing - on the main tab, make sure 'keep in / out stream active' is checked. That does something so the timing stays synchronized from one measurement to the next.

For my measurements, I think most of the problems were from HOLM setting the end gate too early, which was causing response differences in the 1.5kHz region. If I push the end gate out to just before the wiggle at 3.5ms, it looks pretty close to ARTA. And for the measurements I showed here, I wasn't trying to get a phase match between the two, so both were set to auto detect.

But with all that said - I prefer ARTA for this, it's much easier to see exactly where to set the gates. I usually try both, and Room EQ Wizard and sometimes Speaker Workshop as well. If most of them agree on the measurements, I figure I'm OK

[Dennis H]

About the phase with ARTA, I learned that it ignores the phase in the mic cal file. I emailed Ivo about that and he said it's because nobody measures mic phase right and it doesn't matter anyway because the mic phase is pretty close to zero in the XO region. I'm not so sure about that because phase at low frequencies affects group delay and that's down where the mic isn't flat anymore.

Soooooo.... I'm thinking measuring real phase for XO design isn't such a great idea. I'm thinking the best way is to use the ARTA measurement to determine the acoustic center offset. Plot excess phase (measured phase minus minimum phase) and adjust the delay until it's flat in the XO region. Compare the delay you dialed in to the measured mic distance and get the AC location (usually somewhere behind the baffle). When you save the FRD file of the measurement, save it with minimum phase. When you load the driver in LspCAD, enter the Z-axis location (positive is behind the baffle). Just trying to get around mic phase errors.

Edit: sorry for the technobabble, Penngray.

Phase is just time expressed in degrees out of 360 at the frequency of interest. Take a 1 msec (1/1000 sec) time period. At 100 Hz, that would be 1/10 of a cycle or 36 degrees. At 10 kHz, that would be 10 cycles or 3600 degrees. Phase is usually plotted on a +/- 180 degree scale where it wraps to the other side when it passes the limits.

Minimum phase is the theoretically perfect phase that can be calculated from the frequency response with a Hilbert transform. Every dip or peak in the frequency response has a corresponding change in the phase.

Measured phase is just that. If you don't subtract the time of flight, it will go nuts, wrapping through +/- 180 many times as you go higher in frequency. 1 msec time of flight is many more degrees of phase at 10 kHz than it is at 100 Hz.

Excess phase is the difference between measured phase and minimum phase. With a perfect measurement and a perfect driver, if you subtract exactly the right time of flight, excess phase would be a flat line at zero no matter what the frequency response does.

[augerpro]

It might help to see Doug's cal file plot and see how far off the phase is in the crossover region. Reason I say that Dennis, is I found it was much easier-and more accurate- to dial in delay AFTER the driver responses have been EQ'ed to their final slopes, for obvious reasons.

I agree with the others, shut off all filters. Also you need to set an appropriate gate manually, the software won't do it for you. Unless you want completely ungated, but that is much harder to work with in the crossover designer and will include room effects. It is however very handy to plot the response with ungated measurements as a check on the modeling you did with the gated responses. Are you using LSPcad for crossover design? That should be helpful as it has an optimizer which makes it much easier and quicker to attain target slopes/EQ. Also the start and end of the gate must be exactly the same for each driver, or the phase computed for each will not be correct, relative to each other. Biggest advice I can give you though is make sure the measurements are absolutely competent. If you mess this part up, all the rest of your work will be worthless.

BTW was working on teh passive NQ today. Here is the result that will be the prototype to be built:



Horizontal response is quite good, nearly as good as the active version. Vertical is a bit tighter, about +/- 10 degrees, with nulls developing after that. Although by about 50 degrees response is getting good again. Reading some papers from Devantier and Olive the vertical response around this angle becomes quite important due to it being reflected back to the listening position from ceiling and floor. More so than say 30 degrees vertically.

[Saurav]

I'm not sure how much flexibility we have with the demo version of ARTA as far as saving data. I know it can export a basic FR + phase dataset, but I don't know if that can be the output of a Hilbert transform or any other function. Though I guess the processing could be done in LspCAD?

Brandon, I just bought LspCAD and offered to help with the crossover. I'm hoping the rest of you will help me learn how this is done

I have to admit that I've never really understood all the issues surrounding phase ops: I thought mic phase wasn't an issue if you measured all the drivers with the mic in a fixed location, because then the same mic phase error got applied to every measurement.

Also, with say an LR4 XO, what happens if I ignore phase completely, EQ both drivers to get as close to the LR4 targets as I can, and then have penngray dial that in and then adjust driver delay by hand while looking at the reverse null? This isn't practical with an analog XO, but with a DCX at his disposal, this would be quite easy. So theoretically, what would be wrong with the results achieved this way?

Also the start and end of the gate must be exactly the same for each driver, or the phase computed for each will not be correct, relative to each other.

I know the start gate affects phase, I thought the end gate controlled how low the calculated FR is valid. I didn't realize it affects phase as well. I always keep the gates the same when taking measurements to put into the software, but I'll move the end gate around to see what the effects look like.

So the next question is, what measurements does penngray need to take, so I can work with them in LspCAD? If he's using HOLM Impulse, and we can get a fixed time-0 that's to the left of the pre-ringing, and he measures both drivers with the same gates - can I do the rest of the math in LspCAD?

Also, off-axis measurements would be useful too, unless there's enough tribal knowledge around these drivers that someone can just tell me what XO frequency I should aim for.

[Dennis H]

I'm not sure how much flexibility we have with the demo version of ARTA as far as saving data. I know it can export a basic FR + phase dataset, but I don't know if that can be the output of a Hilbert transform or any other function.

It will export whatever you have showing on the graph -- measured phase, minimum phase, excess phase, etc.

[Dennis H]

So the next question is, what measurements does penngray need to take, so I can work with them in LspCAD?

The easiest way is probably to measure both drivers with the mic at about 1m, height about the top of the woofer. If he can send you the impulse files, you can figure out the gating, etc. Worst case, he'd have to slightly adjust the delays you come up with to steer the lobe.

[Saurav]

Thanks, that's what I was thinking too. I'll have to play with HOLM Impulse to see how to export impulse responses. I just ran a test and the re-imported response had different phase, and the time offset / delay was different.

Can the ARTA demo export impulse responses?

[penngray]

Edit: sorry for the technobabble, Penngray

No worries Dennis, it will take me the whole week to figure out what you guys are talking about!

[Saurav]

OK, let's try this.

* Set up the speaker and microphone the way Dennis described

* Set up HOLM Impulse. 1st tab - 'keep in/out stream active', 'log sweep' (increase the start frequency when measuring the tweeter), 2nd tab - 'detect time zero', and 'causal impulse', and at the bottom, make sure 'auto apply time window' is checked. On the Options menu, in the 'Impulse Axis' section, make sure 'Impulse as samples' is selected

* Measure the tweeter, ignore the frequency response for now. Then go back to the 2nd tab, select 'time zero locked', and then click 'use'. Now un-check 'auto apply time window' and 'highpass result to start frequency'

* Then go back to the measurements tab, measure the tweeter again, and measure the woofer. Click 'auto zoom' on the impulse response graph, then post an image of the impulse responses and let's see what they look like

[Dennis H]

BTW was working on teh passive NQ today. Here is the result that will be the prototype to be built:

That looks nice, Brandon!

[penngray]

Thanks for the great input. I should have new results by Tuesday. I have full kid duties the next couple of days with my wife being out of town so I limited time to get back into the room.

[Saurav]

Soooooo.... I'm thinking measuring real phase for XO design isn't such a great idea. I'm thinking the best way is to use the ARTA measurement to determine the acoustic center offset. Plot excess phase (measured phase minus minimum phase) and adjust the delay until it's flat in the XO region. Compare the delay you dialed in to the measured mic distance and get the AC location (usually somewhere behind the baffle). When you save the FRD file of the measurement, save it with minimum phase. When you load the driver in LspCAD, enter the Z-axis location (positive is behind the baffle). Just trying to get around mic phase errors.

Trying to understand this better. So the end goal is to use minimum phase measurements in the crossover calculations, and have the Z offset entered into the driver parameters, correct. And the above is a procedure for correctly determining the AC offsets. With the demo version of ARTA, I can't import anything, so penngray would have to do all this. I can use Jeff Bagby's spreadsheet to extract minimum phase. Is there anything else that can help me figure out excess phase? What if I use SW or HOLM to subtract one measurement from the other... I guess most of those calculations would handle the level, and I have no idea how they would handle phase.

Also, do I need to check the measured distance from mic to baffle? All I need is the difference between the two drivers, I don't need to know the absolute AC distances behind the baffles, right, so I should be able to calculate a difference from the 2 measurements? I didn't really understand the part about 'plot excess phase and adjust delay until it's flat in the XO region', hopefully that'll make more sense when I see what an excess phase plot looks like.

I tried a simpler variation of this with my own measurements. The results were confusing.

* What I've been doing so far - using the phase in the measurements unchanged, leaving driver offsets at 0. With this, LspCAD shows the deepest reverse null with 0.04ms delay on the tweeter. Measurements always show the deepest null at around 0.09ms.

* I used SW's 'remove excess delay' on the measurements. I used to think this generated minimum phase, but the phase ended up different from Jeff B's spreadsheet (and that one looked more 'minimum'), so now I don't know what SW does. Remove time of flight maybe? Anyway, I took these measurements and put them into LspCAD, and used the impulse responses to eyeball the AC offset by measuring the time between the peaks. With this, the XO calculation for the deepest null switched to 0.11ms, which is closer to reality.

* If I repeat the experiment with minimum phase FRD files generated from the spreadsheet, the null point moves to 0.02ms.

Of course, none of this is conclusive because I just eyeballed the AC offset. But it would be nice to understand what exactly is going on here.

[Saurav]

For example, here's my tweeter measurement:



Blue is measured phase, red is minimum phase as calculated by the spreadsheet. These are gated measurements from ARTA, so HOLM is set to 'raw data', i.e. no additional gating applied.

If I do C = A - B, I get:



Does that look correct for excess phase? Next, turning off the amplitude display and adjusting the delay, I get this:



From the bottom, that's 7, 8, and 9 samples of delay. The sample rate for this measurement was 44.1kHz, so that corresponds to .16, .18, and .2ms. I can go fractional samples, of course, but I'd like to make sure this is what you meant by flattening the response.

So then I repeat the procedure with the midrange, and figure out the delay to get it flat through the crossover. Take the difference between the two, convert from samples to time to distance, and that's what I enter as the Z value for the midrange.

Is that correct?

Edit: WTH, I just realized that my tweeter measurement is at 44.1 and the midrange is at 88.2. That can't be good. OK, time to try this again with an older set of measurements. Regardless, I'd like to know if I understood the procedure correctly. I'm curious to see what the calculated AC offset ends up as, and how that affects the LspCAD simulation.

Edit 2: Well, eyeballing the impulse peaks gave me 24mm difference, this calculation gave me 34mm, which maps back to 0.04ms for the deepest null. So it looks like this does more or less the same thing as using the measurements with no phase manipulation? Or maybe that's a coincidence, or I did something wrong.

Anyway, I'm not sure if a 10mm difference is in splitting hairs territory or not (doesn't seem like it should be, but I really have no idea). Especially since penngray can adjust the delay on the final measurements and line up the phase that way.

Edit 3: I just realized that I didn't enter any Y offset into the driver parameters. I expect that would make a difference too?

[Dennis H]

All that minimum phase stuff was just me thinking out loud. I don't have a clue how to do it other than with ARTA and it's probably no big deal anyway. If Penngray measures both drivers from the same mic position, that should work fine with real phase.

Can HOLM save the raw impulse, no gating and no flight time subtracted, and later open the file and apply the gating and other post-processing? That's how ARTA works and it seems like the most flexible way to go.

[Saurav]

Thanks. I just remembered another technique I've read for determining AC offset - measure both drivers together (no XO), then use the minimum phase measurements and adjust delay until the combined output looks like the measured value. They all seem to be different ways of getting to the same end-point.

Can HOLM save the raw impulse, no gating and no flight time subtracted, and later open the file and apply the gating and other post-processing?

I know it can export and import impulse responses. The steps I outlined in post 17 are my best guess for what needs to be done to make the 2 impulse measurements have the correct relationship. Hopefully we'll be able to tell if it did the right thing by looking at the impulse screenshots.

[penngray]

OK, let's try this.

* Set up the speaker and microphone the way Dennis described

Done

* Set up HOLM Impulse. 1st tab - 'keep in/out stream active', 'log sweep' (increase the start frequency when measuring the tweeter), 2nd tab - 'detect time zero', and 'causal impulse', and at the bottom, make sure 'auto apply time window' is checked. On the Options menu, in the 'Impulse Axis' section, make sure 'Impulse as samples' is selected

* Measure the tweeter, ignore the frequency response for now. Then go back to the 2nd tab, select 'time zero locked', and then click 'use'. Now un-check 'auto apply time window' and 'highpass result to start frequency'

Did it

Image not available

Image not available

I know you said ignore the measurement but something is wrong

Image not available

* Then go back to the measurements tab, measure the tweeter again, and measure the woofer. Click 'auto zoom' on the impulse response graph, then post an image of the impulse responses and let's see what they look like

Image not available

Im trying to figure out what I did wrong.

[Saurav]

On the first tab, make sure the input and output devices have the correct soundcard selected, and the correct left/right channel selected. You have input on the left channel, and output on the left channel as well. Make sure that's right. If the output is wrong, you wouldn't have heard the sweep on the tweeter, so I'm assuming that's right. Also, for API, if MME doesn't work, try one of the other options if you have any. I don't remember what I did here, I think I might have tried one or two until I found one that worked.

On the 2nd tab, at the bottom, only leave 'auto apply time window' selected for now, and un-select the other 2.

I know you said ignore the measurement but something is wrong

Agreed

You posted measurements earlier, so clearly you had it working at one point. Which setting(s) did you change to match my instructions? That's probably the one where my instructions were wrong

[penngray]

On the first tab, make sure the input and output devices have the correct soundcard selected, and the correct left/right channel selected. You have input on the left channel, and output on the left channel as well. Make sure that's right. If the output is wrong, you wouldn't have heard the sweep on the tweeter, so I'm assuming that's right. Also, for API, if MME doesn't work, try one of the other options if you have any. I don't remember what I did here, I think I might have tried one or two until I found one that worked.

On the 2nd tab, at the bottom, only leave 'auto apply time window' selected for now, and un-select the other 2.


Agreed

You posted measurements earlier, so clearly you had it working at one point. Which setting(s) did you change to match my instructions? That's probably the one where my instructions were wrong

Thank you for the quick reply!
I will try to get back at it tonight when the kids are a sleep, have to get them from daycare right now.

I will set everything back (if I can remember) and start over. The input/output driver is correct, it never change (mobilepre)

I know I changed
'Impulse as samples' in the menu portion.
'casual impulse' on the 2nd tab
'keep in/out stream active' on the 1st tab.

[Saurav]

Hmm, those 3 settings shouldn't have caused the weirdness you're seeing. You could change 'causal impulse' back to the option that says 'recommended' or 'safest' for now (I think it's 'largest peak'). I think we'll need it back to 'causal impulse' eventually, but right now we're nowhere close to that point.

Also, let's try just one step. Set up HOLM, measure the tweeter, click 'auto zoom' so the frequency response and impulse response are both visible, then see what that looks like. You should see a clear impulse like you had in your first post.

[penngray]

Sorry I haven had much time to do more. I spent the last few days figuring out how to build a LPAD to tame waveguide/CD about -9dB (less hiss).

I also had to figure out the "noise issue" with my soundcard/notebook. I found out that I just have to unplug my notebook from the wall. No other solution worked including plugging my notebook into my power conditioners.

Last night I re-installed HOLM from scratch so that I could have default measurements again and I got something working but nothing that was consistent so Im very fustrated with it right now. Im having a hard time with getting the levels setup properly. Clipping or weak signal kept on happening.

I will try again tonight.

[Saurav]

That sounds frustrating. Have you considered trying the ARTA demo? You should be able to get up and running with the default settings. Click on the red triangle to record, go to the 'sweep' tab, make sure the input and output are right, then click 'record'. You'll be able to see the dB level on both inputs so you can tell if it's too low or clipping, and if it works, you should get a clean impulse. Up to this point, I don't think it's much harder than HOLM. where HOLM gets easier is in the processing that comes after this.

[kendomusic]

Saurav,
I have been testing Holm and ARTA as well. How do you setup ARTA to reflect true amplitude levels so I can have consistent data when doing multiple measurements. I guess I am looking more of a technique. Appreciate your input.

[Saurav]

I haven't tried that. I always measure all the drivers in one session, and I've never tried to correlate data between different measurements. I think I saw a calibration section somewhere that let you specify the SPL reading, that ARTA then matches up to whatever voltage it was seeing at the soundcard input. But I could be way off on this.

[penngray]

That sounds frustrating. Have you considered trying the ARTA demo? You should be able to get up and running with the default settings. Click on the red triangle to record, go to the 'sweep' tab, make sure the input and output are right, then click 'record'. You'll be able to see the dB level on both inputs so you can tell if it's too low or clipping, and if it works, you should get a clean impulse. Up to this point, I don't think it's much harder than HOLM. where HOLM gets easier is in the processing that comes after this.

I have used ARTA for gated measurements already so I have done that much with it along with tons of in room measurements (I love it for that) but with ARTA I never figured how to get all the calibrations set (All that voltage stuff).

HOLM just seems very easy but in my limited time I just can not get a consistent measurement to even be confident with using it. I will have more time hopefully this weekend.

[Dennis H]

I have used ARTA for gated measurements already so I have done that much with it along with tons of in room measurements (I love it for that) but with ARTA I never figured how to get all the calibrations set (All that voltage stuff).

It doesn't have to be calibrated if you measure both drivers in the same session from the same mic position. You just need a clean measurement with no clipping. Do you have an ARTA license so you can save the impulse responses? If so, I can help with the gating and send the FR files to Saurav for the crossover work. Make sure you have the mic cal file loaded.

[Saurav]

Right. Just don't adjust any gain knobs between the measurements. All we need is the relative differences between the drivers, the crossover doesn't care about the absolute SPL level you're listening at.

If ARTA can save / export an impulse response as a WAV file, I think I can import that in HOLM. If not, help with generating gated measurements would be great. Or penngray could post screenshots of the impulse responses, and we could tell him where to set the gates. I know ARTA can export frequency response data.

Hmm, looks like the free ARTA won't export an impulse response. We could probably still do the gating visually.

penngray, if you get a clean impulse response using any software, I think we can figure it out from there

[penngray]

It doesn't have to be calibrated if you measure both drivers in the same session from the same mic position. You just need a clean measurement with no clipping. Do you have an ARTA license so you can save the impulse responses? If so, I can help with the gating and send the FR files to Saurav for the crossover work. Make sure you have the mic cal file loaded.

Nope, just the demo version and that is why I really want to get HOLM working properly. I might have some success this weekend Im feeling confident about my abilities....atleast for now

[penngray]

Right. Just don't adjust any gain knobs between the measurements. All we need is the relative differences between the drivers, the crossover doesn't care about the absolute SPL level you're listening at.

If ARTA can save / export an impulse response as a WAV file, I think I can import that in HOLM. If not, help with generating gated measurements would be great. Or penngray could post screenshots of the impulse responses, and we could tell him where to set the gates. I know ARTA can export frequency response data.

Hmm, looks like the free ARTA won't export an impulse response. We could probably still do the gating visually.

penngray, if you get a clean impulse response using any software, I think we can figure it out from there

We always have the measurements Augerpro did. I wonder if he has the response data.

My TD12M woofer

http://sites.google.com/site/drivervault/driver-measurements/tang-band-75-1558se/ae-speakers-td12m

My waveguide/CD (Celestion 1425)

http://sites.google.com/site/drivervault/driver-measurements/acoustic-elegance-td12m/celestion-cdx1-1425

My waveguide/CD (BMS 4550)

http://sites.google.com/site/drivervault/driver-measurements/acoustic-elegance-td12m/bms-4550

[Saurav]

It would work much better with measurements from your build. I can get the frequency response from Brandon, but that's only part of the problem. Your build has a specific distance / depth between the drivers, which affects their phase relationship, and the crossover needs to handle that. I cannot get that information from Brandon's measurements.

This is also why Dennis suggested measuring both drivers with the mic in the same position - that way the measurements contain information about how far away the drivers are from each other. If you move the mic around between measurements, then that information is lost. Hope that made sense.

I could use Brandon's measurements and get roughly in the ballpark, but it would be a lot better if we had measurements from your speakers.

You had it working once, so take it one step at a time and I'm sure you'll get it working again

[penngray]

Thats cool and that is what I figured.

I have something working. I believe my problem was not understand the gated stuff in HOLM. It would gate very small window for some reason (400Hz to 800Hz) and I would not get any measurements past that. I had to reset settings, set to "raw response" then set my gated response again.

Anyways here are my measurements.

0 deg on axis

Image not available

30 deg off axis

Image not available

[Saurav]

Much better. Yes, I've had that problem before, where HOLM will calculate some really strange alues for the gates, and the response will be completely wacky. When you switched to raw response and then back to gated in the options, did you change any of the options to specify the gate?

Can you post a screenshot of what your 'data analysis' tab looks like? Is the 'time zero' set to auto-detect, or locked? Or, can you describe what you did? Did you do what I suggested the last time - set it to 'detect', measure the tweeter, then change it to 'locked', clicked 'use', and then measured the woofer? Your measurements kinda make me think that you might have followed that procedure, but I'd like to make sure.

Here's how you export an impulse response:

* On the 'measurements' tab, under 'A', click Export
* Go to the 'Impulse' tab
* Click 'browse', select a directory and give it a filename
* Make sure it says file format - wave file, from sample -10000 to sample 10000, format = 32 bit float (these should be the defaults). Make sure 'Normalize to magnitude = 1' is NOT selected.
* Click 'export impulse response' to save the impulse response to the file you specified

Then repeat that for B. Make sure you give it a different filename.

Then post the files here, or if they're too big, feel free to email them to me.

[penngray]

Okay, I have the files....

Can you post a screenshot of what your 'data analysis' tab looks like? Is the 'time zero' set to auto-detect, or locked? Or, can you describe what you did? Did you do what I suggested the last time - set it to 'detect', measure the tweeter, then change it to 'locked', clicked 'use', and then measured the woofer? Your measurements kinda make me think that you might have followed that procedure, but I'd like to make sure.

Yes, I did the "Detect" then I changed it to locked and clicked "use" then measure the CD and woofer again. I tried it all over just to make sure I works again too...
Here is the screenshot, after I clicked locked and clicked "use"

Image not available


Here is the response. Showing Gated is checked

Image not available

[Saurav]

Thanks. I can work with that. These are the on-axis measurements, right? If you have the 30-degree measurements, please export those impulse responses as well.

HOLM keeps crashing every time I try to import these But I think I'll be able to figure something out.

Can you do me one more favor when you get a chance - un-mute both drivers, measure that (so both drivers playing together, but no XO/EQ/ etc.), and post the results here as well. That'll help me cross-check that the simulations match reality. Try to get the mic at the same distance / position you had for the individual measurements, but it's not critical if it's slightly different (I assume you've torn down the measurement setup).

[penngray]

Thanks, I will try to post those later today. I have to measure the 30 degrees again, I screwed them up.

I will also get the measurements with both drivers un-muted.

My mic and speaker have not moved at all. No movies in the HT room until Im done meauring the new speakers

[Saurav]

That's better than my setup Also, I think you need to set 'time zero' back to 'detect' each time you shut down HOLM and restart.



That's the result of some initial processing. I moved the 0-point back to about 5 samples before the start of the woofer impulse. Both measurements have the same 0 offset, and end gate (287 samples, which is about 6.5ms as the image shows).

Also, does anyone understand the differences between half vs. full windows? I don't know if this is Hann or Hamming etc., I think the documentation would probably tell me. Anyway, I can leave the left edge of the window locked at 0, or drag it to the left and it then forms a curve that's a mirror of the right side. I can move this end of the window independently, i.e. the window doesn't have to be symmetrical around 0. Doing this doesn't change the phase at all, so I suspect for this kind of measurement, it doesn't make a difference. But if anyone understands this stuff and can explain it, that would be great

[Saurav]

Here are Brandon's FR and distortion plots for these drivers:









Based on comments I've read about how you shouldn't use a horn / CD close to it's cutoff, and the fact that in the 1-2kHz range the TD12M seems to have lower distortion than the Celestion tweeter - should I push the crossover up higher? 2.5kHz? Brandon, what XO frequency are you using for the NQ?

The other question is off-axis response - I don't want to go so high that the 12" woofer has narrower dispersion than the horn. I think Dennis posted some sims recently (though that was with a 15" driver IIRC), and if I understood those right, an XO in the 2-2.5kHz range would be OK?

I guess I could try both - one in the 1.5kHz range and one in the 2.5kHz range, and see which one sounds better to penngray

[penngray]

Based on comments I've read about how you shouldn't use a horn / CD close to it's cutoff, and the fact that in the 1-2kHz range the TD12M seems to have lower distortion than the Celestion tweeter - should I push the crossover up higher? 2.5kHz? Brandon, what XO frequency are you using for the NQ?

The other question is off-axis response - I don't want to go so high that the 12" woofer has narrower dispersion than the horn. I think Dennis posted some sims recently (though that was with a 15" driver IIRC), and if I understood those right, an XO in the 2-2.5kHz range would be OK?

I guess I could try both - one in the 1.5kHz range and one in the 2.5kHz range, and see which one sounds better to penngray

You can easily run the crossover down to 1200Hz if you choose too. Just today Geddes posted this ""Studies have shown that distortion in a compression driver is not perceivable, so the difference is not relavent."

Here

http://www.diyaudio.com/forums/multi-way/139921-hi-end-hi-efficency-loudspeakers-horn-onken-20.html#post2111295

He crosses is DE250/OS waveguide at around 900Hz.


In the end its the least of the worst, I wouldnt run the woofers up to 2KHz at all. I do not like their sound even past 2000Hz. I can live 2KHz with but I will still be testing different crossovers to findout which I like the best.

A 1500Hz and a 2000Hz deisgn would be awesome for me to try on the DCX.