Measuring/estimating driver acoustic center offsets

[Saurav]

I saw an online article recently where someone's procedure for estimating the acoustic center offset between his drivers was to measure the speaker, model the XO, then increment driver offset in the model until the predicted result matched the measured result. I was wondering if there was a different way to do this. Could we measure the impulse response of both drivers, plot them on the same timeline, and then measure the time difference between the two impulses? Then multiple that by the speed of sound to get the distance/offset between the two drivers?

Would that work? I think Speaker Workshop uses an MLS signal for the impulse response - does that mean it's computed? Would it be better to actually measure this with a real pulse input signal? Should one compare the point where the impulse starts, or the peaks of the impulses? Would the HF response/rise time of the drivers affect this?

Right now I'm using a rough estimate for the driver offset. I picked the midpoint of the magnet as the acoustic center for the cone midrange, and the ribbon's location as the acoustic center for the tweeter. This works OK, but the calculated response from my crossover is still somewhat different from the measured response. And it looks like it's mostly a phase/offset error, because it predicts peaks and dips in the XO region that don't exist in the measurements. Outside of the XO region, the two plots match quite well.

This was just something I was thinking about last night. I know people here design speakers with vertical front baffles. So as general guidelines/advice, how do you measure the offset in the drivers' acoustic centers? I think this would be an important step, because once you have this and have the driver measurements on the actual baffle, you could go off and work on the XO on the computer, and not actually build one until you had something that looked good on paper. Right?

Thanks,
Saurav

[dlr]

A few answers

Could we measure the impulse response of both drivers, plot them on the same timeline, and then measure the time difference between the two impulses? Then multiple that by the speed of sound to get the distance/offset between the two drivers?

This will be approximate, but varies depending upon the high frequency response of the drivers. The larger the driver diameter, the more error introduced.

I think Speaker Workshop uses an MLS signal for the impulse response - does that mean it's computed? Would it be better to actually measure this with a real pulse input signal?

The impulse you see is derived from the MLS signal that actually is better than a pure impulse because it provides a means to cross correlate the the driver response with the probe feedback that eliminates all influence (preamp, amp, cables) prior to the probe test point.

Should one compare the point where the impulse starts, or the peaks of the impulses?

Ideally the peaks would be sufficient.

Would the HF response/rise time of the drivers affect this?

Absolutely. The peak is dependent upon the high frequency response. The more extended the driver, the higher the Q (so to speak) of the peak.

Right now I'm using a rough estimate for the driver offset. I picked the midpoint of the magnet as the acoustic center for the cone midrange, and the ribbon's location as the acoustic center for the tweeter.

Having not worked with ribbons I can only assume that the method is correct for it. But the midrange (short of measuring) should be approximated by using roughly the point of attachment of the former to the diaphragm. My tests make me believe that the acoustic center is actually slightly in front of the point. I could explain the reasoning for this, but it would take more than I have time for now.

This works OK, but the calculated response from my crossover is still somewhat different from the measured response. And it looks like it's mostly a phase/offset error, because it predicts peaks and dips in the XO region that don't exist in the measurements. Outside of the XO region, the two plots match quite well.

It sounds like your assessment is correct.

This was just something I was thinking about last night. I know people here design speakers with vertical front baffles. So as general guidelines/advice, how do you measure the offset in the drivers' acoustic centers? I think this would be an important step, because once you have this and have the driver measurements on the actual baffle, you could go off and work on the XO on the computer, and not actually build one until you had something that looked good on paper. Right?

Yes, this is the goal.

Dave

[Saurav]

Thanks, that was very helpful.

But the midrange (short of measuring) should be approximated by using roughly the point of attachment of the former to the diaphragm.

I'm still learning the terminology here. Is the diaphragm the same thing as the cone? And what's the former? In other wods, where does the former connect with the diaphragm? Is this something I can see on the front of the driver? Sorry for the really newbie questions.

If the Q of the impulse peak is affected by the HF response of the driver, then the peak of the 'slower' driver will be delayed, right. Because the signal will get to both drivers at the same time, but the slower driver will take longer to reach the peak. So... I'll end up estimating an offset that is greater than what it really is, since I'm seeing the delay due to the offset plus the delay due to the slower rise time. Correct?

Also, you said 'short of measuring'. What is the correct way to measure this?

[JonMarsh]

Thanks, that was very helpful.



I'm still learning the terminology here. Is the diaphragm the same thing as the cone? And what's the former? In other wods, where does the former connect with the diaphragm? Is this something I can see on the front of the driver? Sorry for the really newbie questions.

If the Q of the impulse peak is affected by the HF response of the driver, then the peak of the 'slower' driver will be delayed, right. Because the signal will get to both drivers at the same time, but the slower driver will take longer to reach the peak. So... I'll end up estimating an offset that is greater than what it really is, since I'm seeing the delay due to the offset plus the delay due to the slower rise time. Correct?

Also, you said 'short of measuring'. What is the correct way to measure this?

VOICE Coil former to the diapgram- that's generally pretty easy to see from the side, and it's often at the juncture of the spider.

The peak of the slower driver will be delayed, AND modified in shape due to the rise time limitation inherent in the roll off of the frequency response.


Some suggested reading:

http://www.purespeakers.com/offsets.html

http://www.purespeakers.com/offsets2.html

http://stereophile.com/features/100/index3.html

[Saurav]

Thanks. I've gone through the Stereophile series of articles in the past, but I hadn't seen the other one before.

[morbo]

This is an interesting topic to me, since I just started using LSPCAD... so are most people using the method mentioned in the first post to determine acoustic centres (measure individually, model, measure system, adjust acoustic center offset in model until modelled=measured)?

[dlr]

This is an interesting topic to me, since I just started using LSPCAD... so are most people using the method mentioned in the first post to determine acoustic centres (measure individually, model, measure system, adjust acoustic center offset in model until modelled=measured)?

I'm not familiar with usin LspCAD, but that technique will yield IMO the most accurate results. This method, if done correctly, will eliminate all variables that can affect the result.

If only the impulse peaks are used, then there is error inherent due to the sampling rate, This translates into error in identifying the peak, since it's really a line through discrete time sample points. If the peak actually falls between sample points, you can't see it, so the accuracy is at best +/- one sample point for each driver. If each is in error in the opposite direction, then the error may reach 2 x sampling time.

As Jon pointed out, the larger drivers raise additional concern due to the more gradual rise. I suspect that it's closer to a point where a line is tangent to the rising side and has a slope not far from vertical.

Another error may be in the phase of each driver model. If you carefully create the model with the necessary highpass and lowpass slope and Fc so that the Hilbert phase of the model closely matches the measured phase with excess phase removed, then this part of the error can be minimized.

Then you have a value for offset that is at best that of the time sampling accuracy of the first paragraph above. Any error in phase due to not fully matching the model is additive.

However, the three-measurement method can be performed and doesn't even require an MLS or impulse mesaurement, only the actual measured SPL and phase is required. The phase of one or both drivers could even have a fair amount of error in them, but as long as the offset is adjusted experimentally in the program so that the summed SPL response in CAD matches the summed response of the measurement with close attention paid to the intended crossover region, then the model will yield extremely accurate results with a crossover design. Error is minimized for that particular model pairing.

Other ways may be close enough and the lower the crossover frequency the less sensitive the design requirements for offset. But for mid/tweeters, especially at higher Fc, it's important.

Dave

[Saurav]

Thanks, that makes sense. Morbo, look at the links that Jon posted, they describe teh process in more detail.

[Mark_W]

I tried to do it using the Impulse peaks once - it just doesn't work. The wavelength at the maximum frequency I could use, 22,000hz, is 15mm.

This is just too coarse.

The other method listed -measure, model, correct, etc - does work but takes ages.

I found out that I need ~50mm offset. This was too much to recess a tweeter. So I just recessed it 10mm and just worked with the phase error in the crossover model.

You just have to pick your compromises

[dlr]

It doesn't have to take so long

I tried to do it using the Impulse peaks once - it just doesn't work. The wavelength at the maximum frequency I could use, 22,000hz, is 15mm.

This is just too coarse.

This wavelength limit isn't the issue. If both drivers had reasonable output even quite a bit less than 22K, it could work using the impulse. The issue is the sampling rate of the measurement system that results in a sample rate interval that is too large. For woofers, yes, they don't have enough higher frequency output as well. But woofers (except for 2-ways), aren't much of an issue, since the wavelength at crossover is rather long, which helps with this. Actually, woofers in a 3-way are pretty much a non-issue. The small distances in the offsets mean littlte phase rotation.

The other method listed -measure, model, correct, etc - does work but takes ages.

Yes it can, but it depends upon the software. You could run the SPL files through the FRC is your software doesn't have the options. Then the FRC output with its generated phase (all three files) can then be put into any standard software such LspCAD. This doesn't have to be too time consuming. And considering that one may make changes over the course of months in tweaking, it's time well spent.

I found out that I need ~50mm offset. This was too much to recess a tweeter. So I just recessed it 10mm and just worked with the phase error in the crossover model.

You just have to pick your compromises

Sometimes that's all you can do.

Dave

[morbo]

Saurav and Mark_W, is this the method you were attempting?

http://f5.infonet.ee/ergo/offset/offset.html

[Saurav]

Well kinda, but I don't have separate baffles so I was hoping to use that method to figure out the physical offset and then factor that into the crossover models. But like people here pointed out, that won't be a very accurate way of doing this.

I wonder if it will work for what he's trying to do, which looks like 'time aligning' the drivers.

[Paul H]

As was pointed out in one of the articles above for which Jon posted a link, the benefits of adjusting for acoustic centres are debated (not by me - I don't know enough yet to have an opinion ).

I'm curious - do our more experienced designers here typically make allowances for acoustical centres by 1) advanced, detailed microphone and simulator measurement 2) estimating based on physical driver dimensions or 3) ignoring it altogether?

Paul

[dlr]

It can get close, but not precise

Well kinda, but I don't have separate baffles so I was hoping to use that method to figure out the physical offset and then factor that into the crossover models. But like people here pointed out, that won't be a very accurate way of doing this.

I wonder if it will work for what he's trying to do, which looks like 'time aligning' the drivers.

Time aligning and relative acoustic offset are similar and may be close. The limit to time alignment still comes down to the time interval determined by the sampling rate, unless trial-and-error tests are made after the crossover is designed per section.

If the rate is 48Khz, then the time interval is 1 / 48000 = 0.0000208 sec. This cannot be improved.

Multiply this by the speed of sound: 0.0000208 x 345 m/sec = 0.007176 m, or roughly 7.2mm.

This means that the best accuracy is only +/- 7.2mm per driver. Add error due to larger drivers have more gradual slopes and you add more possible error.

The problem is also that if the crossover selected does not very closely match the target for probably at least 2 octaves on both sides of the Fc, then there will be error introduced by phase delays errors due to this, aside from the time interval problem.

For low crossover points at high order this may be of little consequence. For lower order crossovers that have broad overlap and/or those at higher Fc, it becomes increasingly important if close design to target is desired.

Unless a Transient-Perfect design is desired, then time alignment may not be as accurate as phase alignment, since it's the relative phase through the crossover region that is the more important in most cases.

dlr

[JonMarsh]

As was pointed out in one of the articles above for which Jon posted a link, the benefits of adjusting for acoustic centres are debated (not by me - I don't know enough yet to have an opinion ).

I'm curious - do our more experienced designers here typically make allowances for acoustical centres by 1) advanced, detailed microphone and simulator measurement 2) estimating based on physical driver dimensions or 3) ignoring it altogether?

Paul

I go with option 1 when at all possible. OTOH, you could say I"ve been cheating a lot lately, becuase these relatively low crossover frequency steep emulated LR-8 networks are pretty easy to tweak to get the phase to track in the crossover region. And that region isn't very wide, as far as the required driver overlap.

There's issues of phase which you have due to the time delay/natural response of the driver, and due to the physical positioning; then there's phase through the crossover region, for which the desired target is a function of the crossover type. With some crososver types, like Butterworth 3rd order, you have phase effects which are not simple, becuase flat amplitude repsonse on axis ignores that you have an off axis lobe for non-conincident which summs the power response differently than the on axis amplitude response. That's not necessarily a bad thing, but you have to be careful about how the two interact.

For a crossover like an acoustical 3rd order Butterworth, you have to pay attention to the amplitude and phase for up to two octaves each side of Fx, as DLR mentions, in order to get the phase "right" in the crossover region. This itself can get you into a bit of trouble; for example, with drives like the SEAS W18 in an MTM, Joe D'Apollito had to implement a clean controlled roll off above the crossover frequency (about 2 kHz) for two octaves, which means the cone peak, though attnauated with a trap filter, maybe isn't as attenuated as one might like. You also have to play some games with crossover phase and slope in the immediate crossover region to compensate for the difference in acoustic center/phase as he did, which affects the LF driver roll off, also.

[Mark_W]

Saurav and Mark_W, is this the method you were attempting?

http://f5.infonet.ee/ergo/offset/offset.html

This is the method that DOESN'T WORK. What you will find is the impulse peaks will only sit at a few coarsely spaced points - due to the sampling rate (yes, I phrased it wrong DLR, it's the sampling rate, you are correct). With a 96khz soundcard it will be better, but not good enough still IMO.


The method mentioned by Saurav in his very first paragraph did seem to work to me.

[Paul H]

I go with option 1 when at all possible. OTOH, you could say I"ve been cheating a lot lately, becuase these relatively low crossover frequency steep emulated LR-8 networks are pretty easy to tweak to get the phase to track in the crossover region. And that region isn't very wide, as far as the required driver overlap.

When the acoustic offset has been calculated, where specifically do you use it in LSPCad? In the dZ field in the driver parameters screen, which would indicate the distance ahead or behind of the baffle face?

Thanks,

Paul

[morbo]

So I tried the measure/sum/model method, and the meausred value matched the model with -2mm in the dz field for the 'tweeter'... seems a little strange, but since both my 'tweeter' and woofer are cone drivers with a phase plug, I suppose its possible. Does this seem radically wrong to anyone? On a side note, it was pretty easy and took all of 20 minutes to do. I don't know why I was so apprehensive about it.

[dlr]

Don't be concerned with the absolute offset result

So I tried the measure/sum/model method, and the meausred value matched the model with -2mm in the dz field for the 'tweeter'... seems a little strange, but since both my 'tweeter' and woofer are cone drivers with a phase plug, I suppose its possible. Does this seem radically wrong to anyone? On a side note, it was pretty easy and took all of 20 minutes to do. I don't know why I was so apprehensive about it.

The short answer, no, it's not necessarily radically wrong.

It would at first seem odd, but you have to look into what determines phase. SPL and phase for minimum-phase devices are related such that one can be determined from the other. The Hilbert-Bode transform defines the method used, though the implementation may be different. Essentially, though, the phase is a function of SPL from minus infinity to plus infinity. Fortunately we don't have to have measurements to infinity to closely approximate this.

What we do need for accurate phase is to know the slope and knee (Fc) of the highpass and lowpass of the driver. Every change in the selection of these changes the resulting phase calculation. This makes it somewhat arbitrary. However, for our purposes this isn't an insurmountable problem.

The only thing that really matters is the relative phase between the drivers. This is what allows us to properly design a crossover. You may have selected highpass and lowpass of both drivers in such as way that the phase isn't exactly accurate of one or both drivers (it never is, BTW, we approximate), but when you take the resultant modeled phase of each driver and adjust the offset so that the summed model response matches the measured model response, you're done. The single biggest issue is generally the tweeter lowpass, since they often extend out of the measurment system range. Thus the reason why we select some estimate of the lowpass to use in a model if we want to be a bit more accurate (say 2nd order lowpass at 30KHz).

A point to keep in mind. The resultant model is only good for this instance. That is, you can't share it with someone else, since their SPL model will likely not match yours, so every driver pairing should be measured and modeled on a case-by-case basis. It's all relative.

If you're interested, go to my web site and look at the page on the Hilbert transform. It's short, but demonstrates the issue with graphs. Also, check out the link of the first reference at the bottom of the page. I also have some information relating to driver model accuracy on another page.

Dave's Speaker Pages

dlr

[Mark K]

So I tried the measure/sum/model method, and the meausred value matched the model with -2mm in the dz field for the 'tweeter'... seems a little strange, but since both my 'tweeter' and woofer are cone drivers with a phase plug, I suppose its possible. Does this seem radically wrong to anyone? On a side note, it was pretty easy and took all of 20 minutes to do. I don't know why I was so apprehensive about it.

This might be obvious, but one of the things you have to remember is that, if you measure ACTUAL experimental phase at an arbitrary distance and you model the summed response, the z offset in lspCAD should be around zero. Why? Because you've already accounted for it in your experimental measurements.

That is, if you measure the woofer and tweeter's actual phase at, say, 1m and then the combination, lspCAD's summed response at 1m is the same with no z offset (since any relative difference in acoustic centers IS ALREADY IN your measurement.) Sorry to yell. Alot of folks get way overconfused about this. John K's SE tutorial is pretty good at explaining this (even if you don't use SE.)

If you use a different measurement system and measure spl but DERIVE minimum phase, then your number for z will be larger, since using the minimum phase method doesn't account for either acoustic center offset or distance.

Now one thing I've been meaning to ask Ingemar is about how exactly the 3d acoustic model is generated. (i.e. off axis plots etc.) This has always confused me with lspCAD. I've alway just measured and modeled at a particular distance. If your axis chosen isn't very good, there may be off axis problems.

The other way to measure in described in detail in John K's SE tutorial. This seems to me to make more sense in terms of generating an arbitrary modeled measurement point and off axis plots. But there is more room for model error, I think, in this method.

Of course, you can just measure and model for the same point, as for lspCAD.

[dlr]

That is, if you measure the woofer and tweeter's actual phase at, say, 1m and then the combination, lspCAD's summed response at 1m is the same with no z offset (since any relative difference in acoustic centers IS ALREADY IN your measurement.) Sorry to yell. Alot of folks get way overconfused about this. John K's SE tutorial is pretty good at explaining this (even if you don't use SE.)

If you use a different measurement system and measure spl but DERIVE minimum phase, then your number for z will be larger, since using the minimum phase method doesn't account for either acoustic center offset or distance.

This is a good point. I responded on the assumption that H-B generated (derived) minimum-phase phase data was used. If the measurement program has the ability to use the phase imported in a measurement, then the Z offset would remain zero. However, this will also invalidate any other off-axis responses generated by the program.

So if one does as you pointed out, measure and design to a single point, this will work fine. If off-axis (random) points are to be evaluated, then the H-B generated phase and relative offset are both needed.

dlr

[morbo]

Thanks for the extended discussion on this gentlemen, very informative. I have started going through the mentioned links, there's a lot to digest there. To clarify my method a little, I beleive I did it as recommended in the JustMLS manual. That is;

1. set microphone on 'tweeter' axis at IIRC 40cm (I have a small room and want to avoid reflections).

2. Used JustMLS in 2 channel mode, 48000 sampling rate, 10ms window with the correct mic-baffle distance inputted in the offset window to measure the tweeter's response - 'tweeter' was run fullrange.

3. With the mic still on the tweeter axis did the exact same procedure on the woofer, again run fullrange. Was unsure whether to use the diagonal distance to the woofer, or still use the previous distance from the mic to tweeter, so kept the distance from mic - tweeter in the offset field.

4. Connected both drivers up together and did a measurement with both drivers running on tweeter axis. This generated a combined spl/phase plot which I left up on the screen.

5. Opened LSPCad, and started an advanced 2 way crossover. Imported the drivers, and set the Dy for the tweeter to 210mm, the vertical distance between centres of both drivers on my baffle. Dz and Dx were both 0 at this point. Opened the summed frequency response, and tweeter properties, and started adjusting Dz on the tweeter until the phase and spl matched the measured one I still had open in justmls.

I don't quite understand the concept of minimum phase, and how/why one would use a hilbert transform, but I'm hoping since its not mentioned anywhere in the JustMLS manual, I don't have to worry about it 8) Thats probably wishful thinking though.

[ergo]

morbo,

You write that you have set the Dy of tweeter to 210mm....

You should do it the other way around, meaning you should keep the coordinates of the tweeter at 0,0,0 and offset the woofer to 0,-210,z (where z is the Dz that you start to adjust)

The other two important points in LspCad are

*To set the driver radius to 0 mm (to disable the offaxis simulation as you have it already in your measured response)

*To set the mic distance to 0.4m in your case (in LspCad 5 it's infinity by default if I remember correct)

Regards,
Ergo

[morbo]

Thanks for the tips Ergo! I will redo the measurement with the changes you describe.

One thing though, once you've determined your offset and want to start modelling your XO, do you reset the distance to infinite, set it to your intended listening distance, or keep it at the distance from the mic to the front baffle?

I know to keep the offset distance in JustMLS at the distance from baffle-mic, but I believe the distance you are talking about is the one in general options where you set the graph scales etc

[dlr]

How does LspCAD get it's phase data?

It's still not clear to me how LspCAD gets its phase data. If it's directly from the measurements and all were made at one mic position without moving the mic, then it's unclear to me what should be in the offsets. Direct use of imported phase should mean all offsets set to zero, since all delays are inherent in the measurements. In this case, there will only be one accurate design point possible.

If generated phase is used, then the Dx, Dy and Dz should all be set referenced to the origin, usually the listening/measurement axis (I like to use the tweeter), with the others set accordingly.

In the case of generated phase, if all measurements are made at one mic point (for determination of the relative offset), the measurements to use for the actual design won't necessarily be the same as those for offset determination. This is what allows use of the driver diameter settings. Direct on-axis (or close) measurements are needed.

This means that after determining the offset, use direct on-axis (say 1m) measurements of each driver as a start. Then with the Dx, Dy and Dz settings coupled with the driver diameter settings, the designer is free to move the design point freely and maintain the proper phase relationship at all times. This can't be done if measured phase is used, since the delay changes will add error into the phase at essentially all points other than the measurement point.

dlr

[Dennis H]

Looking at Ingemar's Ugly Duckling tutorial, he measures each driver, on the driver's axis, with the mic the same distance from the baffle. Then, when setting up the crossover for a flat baffle, he sets Dx and Dy to their actual values relative to the tweeter and Dz to zero because the reference point was the baffle when they were measured.

[dlr]

Looking at Ingemar's Ugly Duckling tutorial, he measures each driver, on the driver's axis, with the mic the same distance from the baffle. Then, when setting up the crossover for a flat baffle, he sets Dx and Dy to their actual values relative to the tweeter and Dz to zero because the reference point was the baffle when they were measured.

This does sound as though it could be either Hilbert phase or measured phase. In this case I have a question on the measurement part. When using the three measurement technique, the start marker for measurements has to be identical for all three measurements to be valid for measured phase.

If Hilbert phase is used, the Z offset does have to be determined and should be done using the three measurement technique if best accuracy is desired. I've been able to get precision (for the models I've created in CALSOD, I should say) showing that precision to 5 or 6 decimal points of a meter is possible (but hardly necessary I should also say).

dlr

[ergo]

Thanks for the tips Ergo! I will redo the measurement with the changes you describe.

One thing though, once you've determined your offset and want to start modelling your XO, do you reset the distance to infinite, set it to your intended listening distance, or keep it at the distance from the mic to the front baffle?

I know to keep the offset distance in JustMLS at the distance from baffle-mic, but I believe the distance you are talking about is the one in general options where you set the graph scales etc

Yes I meant the distance in general options. For later crossover building/simulations I usually put the distance of intended listening point in there (2-3m usually) but I do check what happens when this is changed to 1m or infinity occasionally. Some filter types give the best result at few meters or farther and not as good at 1m. Also Do not forget to put the radius of speaker units in later also as otherwise you'll get wrong offaxis simulation.

***

dlr, my head is not at the moment cracking the problms too well as I'm still recovering from some virus I managed to pick up... but still I will try to explain the method with JustMLS and LspCad that I have been using.

JustMls uses the 2 channel method where one channel has a feedback impulse right after the amp on speaker contacts and the other for measured impulse through mic. Additionally there is a setting to set the mic distance from front baffle. This will eliminate the "time of flight" from measured phase.

Take a look at

http://www.hifiklubi.ee/fat_margaret_project/cross.html

There the second and third picture show the imported woofer and tweeter response (without moving the mic) and also the two speaker units parallel response (Target). The third picture shows the adjusted Dz and the match precision to measured response with both speakers in parallel.

The below picture shows the match of simulated crossover compared to actual measurement of the built protoype. The acoustic offset for this project was measured the same way as I described.

[dlr]

Looks like it works well

dlr, my head is not at the moment cracking the problms too well as I'm still recovering from some virus I managed to pick up... but still I will try to explain the method with JustMLS and LspCad that I have been using.

Same here.

JustMls uses the 2 channel method where one channel has a feedback impulse right after the amp on speaker contacts and the other for measured impulse through mic. Additionally there is a setting to set the mic distance from front baffle. This will eliminate the "time of flight" from measured phase.

Then LspCAD is using measured phase. That makes the process much easier. The only issue might still be accuracy if an off-axis or other distance is used. In that case, the distances have to be adjusted accordingly to keep relative offset correct. But for single point design (how I requently do it), it's not an issue.

The close correlation between design and measurement is the ultimate indicator. It seems to work well.

Does LspCAD show the time domain data? That is, does it allow adjustment of the start and ending time markers or do you simply put in a time delta? I see reference to number of msec for an input, sounds like stop time marker only.

dlr

[ergo]

Does LspCAD show the time domain data? That is, does it allow adjustment of the start and ending time markers or do you simply put in a time delta? I see reference to number of msec for an input, sounds like stop time marker only.
dlr

Yes, JustMLS has the ability to use a two channel mode where one channel is reference signal and the other one is measured signal. Then one can set the mic distance roughly by measuring with a measuring tape and type it in - then make one measurement and check that the impulse responses in time domain start at the same position. If not it is possible to fine adjust the mic distance.

Ergo

[morbo]

So just to update I followed Ergo's reccomendations and ran the procedure again, this time came out with -61mm as the offset. My thanks again for the tips, I am new to this and there are so many options that I am never sure I'm doing it right.

[ergo]

At least I have always felt that LspCad is one of the best programs in speaker design, there are options but one gets to know them quite fast and after that it does not seem that complicated anymore

Ergo