I can consider some combination of automatic and manual adjustments for Ymax of charts. Few users have proposed manual Ymax and/or scale locking but that is out of question because for example SPL max could change >100 dB when project changes or scale of measurements or gain of buffers is radically adjusted.
Current method for SPL max is automatic with ~2 dB hysteresis preventing continuous jumping while small adjustments just below major grid. I have history with LspCAD having auto Ymax so quite difficult to find a single good reason for manual tweaking
Thank you for the consideration Kimmo.
Jay
Kimmo,
Just a few questions on your ‘Measurement Preparations’ document:
1. “Warning! Do not use B&O ICEpower or compatible amplifiers having balanced/bridged output and high D.C. voltage in both speaker terminals. Speaker minus terminal should be in the same potential with 0 V terminal of line input”. That is written in relation to measuring T/S with LIMP. I assume this is not an issue when measuring the IR of the speaker driver? I use a TPA3116 for measurements which has “balanced” speaker outputs (BTL mode) where the negative is not referenced to ground. I use this output from the amp into my balanced XLR input on my soundcard without any obvious problems so far. But should I worry there are or could be some.
2. “Measure far field responses of one woofer and one mid-range driver and tweeter at 1000 mm in horizontal plane around the speaker.” Shouldn’t the distance be something like 2-3 x the radiating distance, which would be the diagonal baffle length (upper left to lower left corner), to qualify as far-field?
3. “Elevation of mic is at the center point of driver under test i.e. mic and driver have the same Y-coordinate in mm”. Why do it this way instead of leaving the mic’s elevation at the design/listening height for all measurements? Repositioning also increases the risk of not getting the same exact measurement distance correct again and therefore also not the correct Z offset between the drivers, no?
4. If one is meant to reposition the mic to the same Y height as each driver, what about Z? If I have a stepped baffle (Troels style), should I also vary the mic’s depth (Z) so as to maintain the same measurement distance (e.g. 1000mm or otherwise)? That wouldn’t make any sense to me but neither does doing it for Y…
5. “Turn speaker back/front if front baffle is tilted.” Same again, how can we ensure that we incorporate the correct off-axis response and Z offsets in the listening position? In my world, if the baffle is titled, the reference “on-axis” Y-level is still at listening height (e.g. 96.5cm) even if drivers are pointing at the ceiling.
Would appreciate your elaboration on these concepts.
Just a few questions on your ‘Measurement Preparations’ document:
1. “Warning! Do not use B&O ICEpower or compatible amplifiers having balanced/bridged output and high D.C. voltage in both speaker terminals. Speaker minus terminal should be in the same potential with 0 V terminal of line input”. That is written in relation to measuring T/S with LIMP. I assume this is not an issue when measuring the IR of the speaker driver? I use a TPA3116 for measurements which has “balanced” speaker outputs (BTL mode) where the negative is not referenced to ground. I use this output from the amp into my balanced XLR input on my soundcard without any obvious problems so far. But should I worry there are or could be some.
2. “Measure far field responses of one woofer and one mid-range driver and tweeter at 1000 mm in horizontal plane around the speaker.” Shouldn’t the distance be something like 2-3 x the radiating distance, which would be the diagonal baffle length (upper left to lower left corner), to qualify as far-field?
3. “Elevation of mic is at the center point of driver under test i.e. mic and driver have the same Y-coordinate in mm”. Why do it this way instead of leaving the mic’s elevation at the design/listening height for all measurements? Repositioning also increases the risk of not getting the same exact measurement distance correct again and therefore also not the correct Z offset between the drivers, no?
4. If one is meant to reposition the mic to the same Y height as each driver, what about Z? If I have a stepped baffle (Troels style), should I also vary the mic’s depth (Z) so as to maintain the same measurement distance (e.g. 1000mm or otherwise)? That wouldn’t make any sense to me but neither does doing it for Y…
5. “Turn speaker back/front if front baffle is tilted.” Same again, how can we ensure that we incorporate the correct off-axis response and Z offsets in the listening position? In my world, if the baffle is titled, the reference “on-axis” Y-level is still at listening height (e.g. 96.5cm) even if drivers are pointing at the ceiling.
Would appreciate your elaboration on these concepts.
Yes. Connection example copied from LIMP manual is for single-ended power amplifier. No for BTL. Using of BTL is possible with some restrictions but usually there cannot be shortcut connection from speaker output minus to line input minus due to quite high D.C. voltage at both output terminals. For example icePower would explode with example connection. Transformer is one possibility for isolation. Some BTL ready powers amps have also negative supply rail and average/center voltage of both output terminals is 0V. That power amp type allows ZR measurements, but the driver should be connected to single output, between positive output and 0V (negative output disconnected).
Document is written for indoor measurements at home or compatible environment where room height is 2-3 m. First reflections arrive in 3.6-5.4 ms if measurement distance equals to shortest distance from driver to floor or ceiling. We can only select better from two evils; reflections or some loss of diffraction data. Most significant diffraction data has quite short delay which fits into time window so shortish distance and time windowing is obvious selection.
Situation changes if we can measure in anechoic or other very damped room or room with height > 4m or outdoors with lifted speaker & mic. Then it's possible to increase both measurement distance and time window to capture more diffraction data and enable lower transition frequency while merging - or even measure full range and skip merging for good.
Lifting DUT to mic elevation typically room height / 2 has few advantages:
* No need to compensate different distances from driver to mic with Delay[us] and Scaling[dB] parameters of measurement data in Drivers tab.
* Possible cone break up is captured to measurement data also with drivers without tilting towards listening spot. Without adequate data we would need different measurement and session to simulate notch filter for cone break-up.
Tip: Mic positioning is easiest when speaker is rotated 90 deg. Just align mic with front baffe and measure & adjust 1000 mm from mic to vertical center line of the driver. Usually piece of cake, especially with phase plug or baffles with sharp/bevel edges.
Moving mic so that distance between mic and front baffle is constant and Z-coordinate of rotation center is baffle surface of each driver is no less than mandatory if horizontal measurement data is mirrored to vertical plane which is default setting.
If "origin of measurement data" rotates along different circular path in horizontal and vertical planes, simulation result in vertical plane (with horizontal measurement data) will not be correct anymore because driver would "jump" in the simulation causing phase difference variation between adjacent bands/ways.
Simulator calculates exit off-axis angle and distance from driver to mic/listening point by drivers X,Y,Z,R,T parameters and Listening distance setting in Options window. Requirements are that measurement data is normalized:
- Rotation center while off-axis sequence is at the center point of DUT on baffle surface.
- DUT if not rotated or tilted i.e. mic orbit is clean horizontal (and vertical if both planes measured) including axial 0/0 deg response.
- Measurements are dual channel with common Reference time to include differences and rotation of acoustic centers in measurement data. Single channel and USB mics without output and internal loop back are banned.
- Distance from rotation center to mic is constant (1000 mm) or known so that measurement data can be scaled in dB and delay difference compensated with Reference time parameter while converting IR to FR.
Oops. This was not correct answer so lets replace it with something else:
If Z-coordinate of rotation center is far from baffle surface (at the center point of driver), variation in distance between acoustic center and mic increases while 0-180 deg off-axis rotation sequence. Distance variation causes changes to delay of impulse response i.e. IR peak travels more back and forth within time window. If acoustic center rotates too much compared to start and end points of time window, some HF or LF data could flow out of time window and damage exported FR response. So it is advantageous to maintain position of acoustic center as constant as possible during 0-180 deg sequence by measuring all drivers at constant distance from mic to baffle of each driver also with stepped baffle.
In addition, if measurement data travels along circular path due to non-zero Z-coordinate of rotation center, magnitude and phase calculation by simulator does not match with reality anymore if user accidentally enters actual physical coordinates for each driver instance. 'Measurement preparations' document says that rotation center is center point of DUT on baffle surface. User manual says that X,Y,Z is center point of driver instance compared to common origin of speaker, and design origin is (typically) perpendicular endpoint of listening axis on baffle surface. So this structure is quite well pre-designed and hopefully clearly enough documented to be easy for users.
Automatically adapting time window and possible minimum phase extraction eliminates effects of previous IR drifting, but detection of IR peak is quite difficult >90 deg in reflecting environment, and minimum phase extraction is otherwise bad concept because it cannot detect increasing sound path around enclose edges >90 deg or horns and polarity changes which are obvious with dipole radiators. Especially random polarity effects are impossible to include in geometry simulation with partly dipolic radiators such as resistance enclosure. Constant position of time window and Reference time ensure stable indoor measurement result also with directive radiators.
Deep horns may require longer time window and more frequency points to handle rotation in case 90-180 deg is also simulated, but stronger directivity helps with that.
Got you (I think).
So to be clear, for something like the speaker below, you would keep adjusting the depth (Z position) of the DUT on the turntable so that, in order, the tweeter baffle, then the midrange baffle, then the midbass baffle etc, in each case is at the center of rotation (i.e. each driver baffle (sub-baffle) always at 1000mm from mic).
But then, if the physical changes in Z are not recorded, wouldn’t the simulator then be oblivious to this and think that all drivers are on a flat baffle?
So to be clear, for something like the speaker below, you would keep adjusting the depth (Z position) of the DUT on the turntable so that, in order, the tweeter baffle, then the midrange baffle, then the midbass baffle etc, in each case is at the center of rotation (i.e. each driver baffle (sub-baffle) always at 1000mm from mic).
But then, if the physical changes in Z are not recorded, wouldn’t the simulator then be oblivious to this and think that all drivers are on a flat baffle?
Simulator does not think - it just calculates by measurement data you load and parameters you enter
Each driver instance in crossover schematic has X[mm], Y[mm], Z[mm], R[deg] and T[deg] parameters you need to enter to match simulation with reality. As already mentioned XYZ is location of rotation center point of each driver compared origin on baffle surface common for all drivers.In that particular case all drivers have common X=0mm and R=0deg but close to everything else is individual per driver.
Tweeter: Z and T are small negative values.
Small mid 12M: Could be origin (=endpoint of design axis) so Z=0mm and T=0deg.
Big mid 18W: Z is small negative and T is small positive.
Woofers have common negative Z (bigger than previous Zs) and T=0deg.
Y of each driver is relative to small mid which was selected as origin.
This is not very simple but enables mechanical adjustments with X,Y,Z,R,T parameters with simulator as well as combining whole construction from different prototypes for each driver (accepting some error in inter-diffraction/scattering/reflection).
Sorry didn’t see this before your latest reply. It was exactly that quote that motivated my post.
However my question still stands that how does the simulator keep track of Z-offsets between drivers if we keep moving the mic on the Z-axis without manually recording it (not that a manually praxis would be preferable obviously, but for the arguments sake)?
Also, related to your last reply, would it be worthwhile finding the acoustical center on the Z-axis and use that as the rotational center instead of the baffle (which otherwise will inevitably lead to the deviations and errors that you are talking about, not only for horns)? Otherwise I would have thought the simulator had some kind of IR peak detection to time sync the full set of off-axis measurements.
On a general note, since VituixCAD seems to take a strict do-it-100%-right stance (no single-channel software or hardware etc), I’m not sure optimising for making measurements indoors in domestic living rooms at 1.2m height is consistent with that approach. I suspect that most users fanatic enough to delve into VituixCAD are also OK with measuring outdoors or in an auditorium with much larger time windows.
EDIT: Cross-posting again, sorry. I’ll take a break here to digest all you’ve said. Thanks for your support.
Last edited:
Sorry but I don't take this very well.
Already told that you don't have to measure at 1x1.2 meters if better environment is available. For example I don't have auditorium or anechoic and weather is rainy, windy or temperature freezing. Still no bad conscience or results when designing with this system, though it's not 100% perfect for sure.
Ability to capture timing differences without cheap tricks such as hanging acoustic reference radiator to mic boom is basic feature of decent measurement gear and software. Why to question or even smile to that? It has been self evident for more than decade until USB mics screwed up things.
^No problem. Freeware does not give total immunity, but sometimes I may lose patience. Especially if I have to repeat justifications and public instructions more than few times for the same person.
Quite opposite in fact. Rotation center on front baffle or horn mouth produces smaller timing variation while off-axis (0-180 deg) measurement sequence than rotating around driver's acoustic center or horn throat. Smallest variation exists when rotation center is in front of (unidirectional) driver or mouth, but that is too abstract as rotation center (imo).
Relative amplitude varies with shortish measurement distances too but also that should be better when rotation center is at Z of mouth instead of throat.
Quite opposite in fact. Rotation center on front baffle or horn mouth produces smaller timing variation while off-axis (0-180 deg) measurement sequence than rotating around driver's acoustic center or horn throat. Smallest variation exists when rotation center is in front of (unidirectional) driver or mouth, but that is too abstract as rotation center (imo).
Relative amplitude varies with shortish measurement distances too but also that should be better when rotation center is at Z of mouth instead of throat.
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
Simplified simulated study about measurement distance and radiating distance. Here is quite slim 220x1000mm baffle with 10mm edge rounding. 3" driver 800mm from bottom. Includes 1st order diffractions only (72 rays). Thin lines are mic at 2500mm and thick lines mic at 1000mm.
Difference in power response is visually undetectable. Two ~0.5 dB difference peaks in axial response which are more visible in DI due to narrower span. Error with this magnitude would not affect to crossover design. Biggest difference exists below merging frequency which is irrelevant.
According this kind of simplified study "too short" measurement distance is by far less significant than length of time window and window function => shorter measurement distance with longer time window is preferred.
For example shortish Hanning window could cause really significant magnitude error above merging frequency, though it is more stable for indicating directivity at bass.
Large drivers such as 18" woofers produce near field effects to 1 m. They are usually crossed below 400 Hz so moving mic from 1 to 1.5 m is not mandatory to get nominal far field with 6xRadius distance.
So instructions should be valid and consistent also from this point of view, though it could be mentioned that user has more freedoms in larger free space or anechoic.
Last edited:
Yes I have. Typically total response to selected reference angle or listening window average is straight horizontal or slightly tilting up/down. High pass slope of bass range can be optimized with driver's target curve which could be either textbook response or loaded response file. Not enough motivated to make more complex.
Total sound balance designed with Optimizer should be weighted average of axial and power responses. Special shape for power response would be more valuable than for axial response because concepts with e.g. directive bass (should) have flat horizontal power below low treble and possible straight tilting down from lower treble to top octave. I have designed those concepts so that weighted average of straight horizontal axial within ~100Hz-15kHz and tilting power response within ~1-15kHz is optimized first, and then fine tuned manually. Bass slope manually avoiding too high excursion. This method has been quite adequate imo.
Try the scaling option in dB. You can add or deduct dBs to each driver. Make sure you add/deduct the same dB amount for each driver to keep the same relative level.
Thank you for you prompt reply, but I am looking for a permanent solution. But I have this problem only with one driver, only of one project.
So the questions actually is: Why does it happen? And how to return to normal operation?
If I understand correctly, the problem is with only one driver, but the project is 2 or 3 (or more) way?
If you have measured the drivers involved in this project with the same method (let's assume a bass-mid and a tweeter) then the dB levels of both drivers should be visible because they should be within 5 to 10 dB relative level (unless the tweeter is high eff pro driver in a horn, which may be off scale).
The default range in Vituix is 95 dB to, say, 55 dB so there is easily at least 40 od 50 dB visible range. Please check what absolute level your measuring is showing for each driver. Some SW are normalizing at around 0 dB so you have to add dBs to lift up in the Vituix graph.
The most important is that both (or all three, four) drivers have the same measurement origin - measured with the same SW, same level. If that is so, they will all have the same problem (solvable, as shown) and not just only one. Sort of apples with apples situation.
Just my best prediction, based on your limited information input.
Edit: also this - you can set the target level in the optimizer as well, so it can also go up or down (in dB)
If you have measured the drivers involved in this project with the same method (let's assume a bass-mid and a tweeter) then the dB levels of both drivers should be visible because they should be within 5 to 10 dB relative level (unless the tweeter is high eff pro driver in a horn, which may be off scale).
The default range in Vituix is 95 dB to, say, 55 dB so there is easily at least 40 od 50 dB visible range. Please check what absolute level your measuring is showing for each driver. Some SW are normalizing at around 0 dB so you have to add dBs to lift up in the Vituix graph.
The most important is that both (or all three, four) drivers have the same measurement origin - measured with the same SW, same level. If that is so, they will all have the same problem (solvable, as shown) and not just only one. Sort of apples with apples situation.
Just my best prediction, based on your limited information input.
Edit: also this - you can set the target level in the optimizer as well, so it can also go up or down (in dB)
Last edited:
I'm still waiting problematic file(s) to my e-mail... Or is this already solved?
Anyway, automatic Y max works above -200 dB SPL so curve will be visible if response is loaded to new project with single driver data in Drivers tab and single driver instance in crossover schematic connected to generator and associated to data in Drivers tab, and SPL is at least about -240 dB in the file. Scaling parameter in Drivers tab has +/-200 dB range.
Few possible reasons if response will not show up:
- file is bad
- some extrapolation problem
- impedance response is zero or impedance Scaling is 0 i.e. driver is electrically close to short circuit (~10 mOhms)
- file naming rules are not followed -> response to initially selected reference angle does not exist -> response is not updated to chart
Anyway, automatic Y max works above -200 dB SPL so curve will be visible if response is loaded to new project with single driver data in Drivers tab and single driver instance in crossover schematic connected to generator and associated to data in Drivers tab, and SPL is at least about -240 dB in the file. Scaling parameter in Drivers tab has +/-200 dB range.
Few possible reasons if response will not show up:
- file is bad
- some extrapolation problem
- impedance response is zero or impedance Scaling is 0 i.e. driver is electrically close to short circuit (~10 mOhms)
- file naming rules are not followed -> response to initially selected reference angle does not exist -> response is not updated to chart