Allan Posted April 11, 2024 Posted April 11, 2024 (edited) 9 hours ago, Keith_W said: I don't see it. Would you be able to point out what you see? EDIT I prefer looking at reflections in the Energy-Time Curve. I find it easier to see. It also tells you what the amplitude of the reflection is. But it does not tell you where it is on the FR. Have to look at the spectro for that. @Keith_W In the above plot, there's a swing between 1 and 2khz in the frequency response or red line. The phase or green line correlates. This could be a number of things and if its acoustics its most likely phase or a reflection issue, id have to run calcs to discount a room mode... but if it where me Id first eliminate or discount the mode. That leaves a reflection or phase issue. Using the impulse response window or as you call it an ETC, look for a spike that coincides distance wise with a potential reflection point. For example, if the IR shows a spike where its timing is 5 ft delayed then look for something 5ft or 2x2.5ft longer in distance than the speaker to mic. In the case of 2x2.5ft, this could be a front wall reflection where the speaker is 2.5ft out from the front wall, the reflection being in opposite phase to the direct speaker sound Edited April 11, 2024 by Allan 1
crtexcnndrm99 Posted February 27 Posted February 27 (edited) Hi all, Thought I’d post here my latest delving into speaker tuning. It’s a work in progress (of course), but I’m interested in feedback on the frequency response. Xo: sbAudience Bianco 0B350 - 700hz 12db BW slope JBL 2440 w titanium diaphragms in AH340 - 3200hz 6db BW slope Notch filter for peak ~2-6k - 3800 -5db 0.7 2 watt amp on HF and ~20 watt amp on LF, biamping with tubes (what I already have). Amps SPL-matched roughly as well. Here’s the FR plot with the above setup and minidsp for XO, measured approx 1m at seated head height: Current limitations - due to not yet having time to explore and optimise - are measuring time alignment using physical adjustment of HF driver placement relative to LF cabinet; playing with toe-in or angling on vertices and horizontal axis; working out the attenuation amount for the HF, to have the option of returning to one amp again instead of bi-amping; reducing volume and increasing rigidity of LF cabinet with sandbags. Edited February 27 by crtexcnndrm99 3
Keith_W Posted February 28 Posted February 28 Before we even look at measurements, we need to know that the measurements have been taken properly. I spend a good chunk of my time on other forums helping people and my first question is always - "are you using the mini tripod that came with your UMIK?". Throw that mini tripod away and get a proper mic tripod. If you are going to use your measurement for DSP, you need a meaningful measurement that reflects reality. This means taking a measurement which is uncontaminated by reflections. The distance between the mic and closest reflective surface affects the longest wavelength that can be meaningfully measured. You can examine reflections in your Energy-Time Curve (ETC): I pulled up an example from my collection of measurements. Here, we have a reflection arriving 2.35ms after the main impulse. Now we can do some maths. The distance travelled by the reflection can be calculated with the speed of sound, d = t/1000 * c (t = time in ms, c = speed of sound of 343m/s). So in this case, the distance is 0.8m. This means that the reflection travelled an extra 80cm before it reached the microphone, in comparison to the direct sound. The lowest reflection-free frequency can be determined by assuming that the time-lag of the reflection equals one period T of the frequency in question. Using the equation f = 1/T, the lowest frequency is 425Hz. This means all measurements below 425Hz are uninterpretable - and if you use this measurement for DSP, anything you measure below 425Hz will be unrepeatable. Before you go any further, it is imperative to take a proper measurement. Measurements taken to see what is going on are one thing, but measurements taken for DSP are another. The quality of your correction is directly impacted by the quality of your measurement. So ... no mini tripod please. No stacking cushions for proper height. Get a tripod with a mic boom similar to this. I can't vouch for the quality of that tripod I linked, I am only using it as an example. As you can see, these tripods are inexpensive. I suggest you go to a pro audio shop or a DJ store and ask them to show you mic tripods. Sturdiness and build quality are not super important, but maximum height is. It needs to be at least as tall as your speaker or even taller to cover other contingencies. 1
andyr Posted February 28 Posted February 28 (edited) 1 hour ago, Keith_W said: If you are going to use your measurement for DSP, you need a meaningful measurement that reflects reality. This means taking a measurement which is uncontaminated by reflections. Except that when I sit in my listening position, Keith ... my ears hear those reflections! So why wouldn't I want my DSP to include the effect of reflections? 1 hour ago, Keith_W said: The distance between the mic and closest reflective surface affects the longest wavelength that can be meaningfully measured. You can examine reflections in your Energy-Time Curve (ETC): In what way is your arrow pointing to a "reflection arriving 2.35ms after the main impulse? The 'X' scale of the peak appears to be at 0.1275ms. If we need to double or quadruple this number ... we still don't get anywhere near 2.35ms? Edited February 28 by andyr
crtexcnndrm99 Posted February 28 Posted February 28 Thank you Keith, I’ve enjoyed reading your contributions on threads here on this forum before today. In my initial post, I failed to mention that the intention of measuring is to tune and find a relatively straightforward XO I can implement passively… hence the experimenting with different slopes and the notch filter, rather than full blown EQ. Point taken on the microphone stand front, though. Something that would benefit this experimentation and any future digital based systems I may assemble. It seems that retaking the measurements with a proper stand could be one step towards more accurate speaker tuning, prior to spending money on pricey passive XO components
Keith_W Posted February 28 Posted February 28 10 minutes ago, andyr said: Except that when I sit in my listening position, Keith ... my ears hear those reflections! So why wouldn't I want my DSP to include the effect of reflections? The answer is very complex, Andy. In a nutshell, we deal with long and short wavelengths differently. For long wavelengths, the speaker is corrected together with the room. So we want to include reflections in your DSP. For short wavelengths, psychoacoustics says that the direct sound and reflections are heard as separate events depending on the timing and the Haas fusion window. DSP can correct the direct sound of the loudspeaker only, and not the reflections. Ideally, the reflections should be spectrally correct (i.e. good loudspeaker design), delayed (i.e. larger room), attenuated (absorptive room treatment, furniture), and reverberant (diffusers or more furniture). Note that all the factors that influence reflections have nothing to do with DSP. If you include reflections in your high-frequency correction, the reflections are extremely specific to the microphone position since the wavelengths are so short. Remember that the shortest wavelength (20kHz, 17.2mm) is much smaller than the distance between your ears (typically 15cm-17cm). In other words, including reflections in DSP corrections for high frequencies is correcting for a single and extremely specific point in space which is smaller than your head. And furthermore, the measurement is unrepeatable. Moving the mic only a few cm will change the pattern of reflections. In summary: for long wavelengths, we correct the speaker and room together. For short wavelengths, we correct the direct sound of the speaker ONLY. If we can't take an accurate measurement of the direct sound (and there are all sorts of limitations) it may be wiser to leave it alone. 10 minutes ago, andyr said: In what way is your arrow pointing to a "reflection arriving 2.35ms after the main impulse? The main impulse is that big spike at 125ms (0.125 on the horizontal scale). This is Acourate, so the scales are not so intuitive. The horizontal scale is time in seconds. Acourate always centres the main impulse at sample 6000. With a sampling rate of 48kHz, sample 6000 equals 125ms (6000/48000). Acourate says that the reflection arrives at sample 6113. So the time-lag is 113 samples, or (113/48000) = 2.35ms. 3 minutes ago, crtexcnndrm99 said: In my initial post, I failed to mention that the intention of measuring is to tune and find a relatively straightforward XO I can implement passively… hence the experimenting with different slopes and the notch filter, rather than full blown EQ. FWIW, if I were you, I would stick to a DSP crossover. It is vastly superior to a passive crossover for many many many reasons. The only advantage a passive XO has is its ease of use. Sell the loudspeaker to a consumer, and it is idiot proof. All the other supposed advantages of passive XO's is grounded in audiophile myth. But if your intention is to use the speaker yourself, there is no need to make it idiot proof. This post is already getting too long so I won't expand on this point. If this is what you want to do, then you will need accurate nearfield driver measurements. This is going to be your problem: if you measure too close, you will not account for the baffle step. The baffle step is a volume loss of about 6dB that occurs over four octaves. Its centre frequency (f3) can be calculated with the equation f3 = 115824/W with W being the width of the baffle in millimetres. If your bandpass is within the baffle step, then you need to measure a minimum of 2 baffle widths away. If you measure too close, you won't account for the baffle step. If you measure too far, your measurement will be contaminated by reflections. IOW you have the choice of one meaningless measurement (no baffle step) or another meaningless measurement (contaminated by reflections). This is why people take their loudspeakers outside and elevate it. However, I have recently learnt a new method. First, take a nearfield measurement (mic almost touching the driver) to obtain the pure driver response. Use a baffle step simulator like this one or this one to simulate a measurement from the main listening position. Convolve the simulated baffle step with your actual measurement. Then take a measurement of that driver alone from the MLP to compare. Here we have the result of my simulation. In faint red is the nearfield measurement of my woofer. The upper graphs is the amplitude, the lower graph is phase. In pink is the convolved measurement of the woofer measurement and the BDS at MLP. And in black is the actual measurement at MLP. This experiment proves that this simulation works. But let me show you what is possible with a DSP crossover: It can take a nonlinear amplitude and phase response and completely flatten it. You can see the lower graph (wrapped phase) has the typical rising group delay at lower frequencies. I can completely flatten it so that it is linear phase. This is only possible with linear phase FIR filters. Which unfortunately, your MiniDSP can not do.
crtexcnndrm99 Posted February 28 Posted February 28 1 hour ago, Keith_W said: However, I have recently learnt a new method. First, take a nearfield measurement (mic almost touching the driver) to obtain the pure driver response. Looking at one section at a time - how does this nearfield method of measurement change, if at all, for a horn or waveguide?
Keith_W Posted February 28 Posted February 28 1 hour ago, crtexcnndrm99 said: Looking at one section at a time - how does this nearfield method of measurement change, if at all, for a horn or waveguide? Good question. Each driver type needs a different measurement strategy. If you measure too close to a horn/waveguide, internal reflections may arrive out-of-phase if the mic is placed at the mouth. The exact distance varies depending on the design of the horn, the flare, etc. but as a rule of thumb the mic should be placed two horn diameters away. Some experimentation will be necessary. If I don't know where to measure, I take measurements at 0cm, 20cm, 40cm ... up to 100-150cm and then overlay and compare them. Observe the curves for two things (1) the shape of the curve will change, and (2) at some point reflections will creep in making the measurement look wavy. Usually for horns, very close measurements look wavy (internal reflections), then they straighten out, then they become wavy again (room reflections). At some point the shape of the curve will stop changing - hopefully this will be before the measurement becomes wavy. This will be where you should measure. I hope that makes sense. 1
andyr Posted February 28 Posted February 28 Thanks, Keith. 3 hours ago, Keith_W said: The answer is very complex, Andy. In a nutshell, we deal with long and short wavelengths differently. If you include reflections in your high-frequency correction, the reflections are extremely specific to the microphone position since the wavelengths are so short. Remember that the shortest wavelength (20kHz, 17.2mm) is much smaller than the distance between your ears (typically 15cm-17cm). In other words, including reflections in DSP corrections for high frequencies is correcting for a single and extremely specific point in space which is smaller than your head. And furthermore, the measurement is unrepeatable. Moving the mic only a few cm will change the pattern of reflections. But if, when setting up the DSP, you average, say, 9 readings with the mic set: slightly above the ears at the ears and slightly below. Plus: centred 30mm to the right and 30mm to the left ... then I would've thought you have "got over" the fact that "Moving the mic only a few cm will change the pattern of reflections"? 3 hours ago, Keith_W said: The main impulse is that big spike at 125ms (0.125 on the horizontal scale). This is Acourate, so the scales are not so intuitive. The horizontal scale is time in seconds. Acourate always centres the main impulse at sample 6000. With a sampling rate of 48kHz, sample 6000 equals 125ms (6000/48000). Acourate says that the reflection arrives at sample 6113. So the time-lag is 113 samples, or (113/48000) = 2.35ms. Aah, OK.
Keith_W Posted February 28 Posted February 28 59 minutes ago, andyr said: But if, when setting up the DSP, you average, say, 9 readings with the mic set: slightly above the ears at the ears and slightly below. Plus: centred 30mm to the right and 30mm to the left ... then I would've thought you have "got over" the fact that "Moving the mic only a few cm will change the pattern of reflections"? Yeah you can do that. Or you can do an MMM. It's a different correction philosophy. I don't like doing that because IMO it gives an "overcorrected sound". In the past I did tend towards overzealous DSP, but these days I tend more towards doing what I can to do proper nearfield or quasi-anechoic correction and then let the sound do what it wants in the room - you know, like a normal speaker. I think it sounds better. But you can of course do what you want.
crtexcnndrm99 Posted February 28 Posted February 28 2 hours ago, Keith_W said: Good question. Each driver type needs a different measurement strategy. If you measure too close to a horn/waveguide, internal reflections may arrive out-of-phase if the mic is placed at the mouth. The exact distance varies depending on the design of the horn, the flare, etc. but as a rule of thumb the mic should be placed two horn diameters away. Some experimentation will be necessary. If I don't know where to measure, I take measurements at 0cm, 20cm, 40cm ... up to 100-150cm and then overlay and compare them. Observe the curves for two things (1) the shape of the curve will change, and (2) at some point reflections will creep in making the measurement look wavy. Usually for horns, very close measurements look wavy (internal reflections), then they straighten out, then they become wavy again (room reflections). At some point the shape of the curve will stop changing - hopefully this will be before the measurement becomes wavy. This will be where you should measure. I hope that makes sense. That does make sense, thanks. I will need to try this and report the results. 1
RoHo Posted March 1 Posted March 1 (edited) On 28/2/2025 at 10:58 PM, crtexcnndrm99 said: That does make sense, thanks. I will need to try this and report the results. To add a highlight to Keith’s comprehensive information: To do XO development you need to exclude all the room reflection stuff and look at what is only coming directly from the speaker. In REW this is straightforward using the Impulse Response function and the technique Keith describe. Look for the 1st reflection and set a “window” that ends just before it. This is called gating. In practice in a typical room the length of this window will be a max of 5ms. This corresponds to a frequency of 200hz. With my room it’s 4.2ms and 240hz. This is the lower limit of your measurement but also the minimum resolution between measurements. So you will have data points at 200, 400,600,800,1000 etc. Below about 500hz the curve won’t make any sense. Measuring for a low frequency XO is difficult! I’ll dig up some of my REW curves to illustrate. (1 hr later after toast and tea ) My speaker are a simple 2-way, XO at 3.5kHz. I measure at 1m distance, on the tweeter axis. Here is the un-gated response with the impulse response above and the conventional frequency response below: On the IR trace note the large spike at Time 0, that's the direct sound from the speaker. Follow along the trace and you will see the reflected information arriving at various times after the initial spike. These reflections give all the ripples in the FR making it impossible to determine what the actual speaker is doing. We need to eliminate all these reflections so set a "window" just before the first. To do this use the IR window box which I've shown in the pic. The only parameter of real interest is the Right Window which defaults to 200ms, so large it is effectively off. The first blip on the IR is at about 4.4 ms so that's where the window will go: The data under the purple shading is all excluded. Note that in the IR window pane the frequency limitations of the window are stated. In this case there will be no data below 227Hz and this is also the increment between data points. This is why the lower frequencies in the FR are smoothed. I'm really interested to see what the data from the horn looks like. I've lusted after Azurahorns for years Edited March 2 by RoHo Extra information 1 1
awayward Posted March 6 Posted March 6 I’m just about to begin my journey with REW, mic stand arrived yesterday and a UMIK-1 arrives tomorrow. I thought I’d create a pink noise wav file using REW and add it to my Roon library, is this a suitable option? Are there any specific options I should use to create the wav file, appreciate any advice from the gurus
crtexcnndrm99 Posted March 7 Posted March 7 (edited) Perhaps some progress today. I measured using a mic on a stand (not pillow stack) at several distances from the horn/waveguide mouth. Given a diameter of 60cm or more at termination, I tried measuring at a distance of 100-130cm (separated and smoothed 1/6). Not quite sure how to interpret the impulse graph yet for the gating. Did I need to set a timing reference or something such as that for the frequency sweep? Lastly, the Azura horn measured without any PEQ in minidsp, and a simple 1st order butter worth at 800hz instead of the 3200hz one. Interesting… is the PEQ of a couple dB around 2-6k working… Azura horn measured below or off-axis about 20-30 degrees. edit: these measurement were taken with the dsp xo in place - I realise now that some raw measurement may be needed? How might I do that? Edited March 7 by crtexcnndrm99 1
Keith_W Posted March 8 Posted March 8 (edited) On 06/03/2025 at 4:29 PM, awayward said: I’m just about to begin my journey with REW, mic stand arrived yesterday and a UMIK-1 arrives tomorrow. I thought I’d create a pink noise wav file using REW and add it to my Roon library, is this a suitable option? Are there any specific options I should use to create the wav file, appreciate any advice from the gurus Congratulations. There are a few types of pink noise which are better for different reasons. I have to admit I don't know all the different settings for REW - some are better for subwoofers, some are better for speakers, and so on. I just use full range pink noise and that's it. Maybe someone else can weigh in. 5 hours ago, crtexcnndrm99 said: Perhaps some progress today. I measured using a mic on a stand (not pillow stack) at several distances from the horn/waveguide mouth. Given a diameter of 60cm or more at termination, I tried measuring at a distance of 100-130cm (separated and smoothed 1/6). Not quite sure how to interpret the impulse graph yet for the gating. Did I need to set a timing reference or something such as that for the frequency sweep? edit: these measurement were taken with the dsp xo in place - I realise now that some raw measurement may be needed? How might I do that? 1. Don't separate the curves when you are comparing them. Use "Align SPL" and align all the curves and overlay them on to of each other. Then look to see where they deviate. 2. Timing reference is not necessary for these types of sweep, but it is a good idea to always have a timing reference. 3. Some people use impulse response for gating, but I prefer the Energy-Time Curve. If you post the ETC I will show you how to read the time for gating. Or you can upload your .MDAT file somewhere and put a link on SNA. 4. Connect your amplifier directly to the driver without the DSP XO in place. Set REW to -60dB and gradually increase the volume until you get a good measurement. Edited March 8 by Keith_W 2
crtexcnndrm99 Posted March 8 Posted March 8 17 hours ago, Keith_W said: Congratulations. There are a few types of pink noise which are better for different reasons. I have to admit I don't know all the different settings for REW - some are better for subwoofers, some are better for speakers, and so on. I just use full range pink noise and that's it. Maybe someone else can weigh in. 1. Don't separate the curves when you are comparing them. Use "Align SPL" and align all the curves and overlay them on to of each other. Then look to see where they deviate. 2. Timing reference is not necessary for these types of sweep, but it is a good idea to always have a timing reference. 3. Some people use impulse response for gating, but I prefer the Energy-Time Curve. If you post the ETC I will show you how to read the time for gating. Or you can upload your .MDAT file somewhere and put a link on SNA. 4. Connect your amplifier directly to the driver without the DSP XO in place. Set REW to -60dB and gradually increase the volume until you get a good measurement. Hi Keith, re numbers 1 and 3, here are some ETC curves I pulled from the above measurements and an all SPL graph. The setup for measurement is current computer > minidsp > two amplifiers > drivers. 1
playdough Posted March 9 Posted March 9 2 hours ago, crtexcnndrm99 said: Hi Keith, re numbers 1 and 3, here are some ETC curves I pulled from the above measurements and an all SPL graph. The setup for measurement is current computer > minidsp > two amplifiers > drivers. Gday Very interesting series of measurements. Especially off axis (is excellent) What model Azura horn and driver are you using ? . I ask as I’m about to start measuring a B&C 464 coax in an AH 120. Watching on ,,,,, have fun
crtexcnndrm99 Posted March 9 Posted March 9 7 minutes ago, playdough said: 2 hours ago, crtexcnndrm99 said: Hi Keith, re numbers 1 and 3, here are some ETC curves I pulled from the above measurements and an all SPL graph. The setup for measurement is current computer > minidsp > two amplifiers > drivers. Gday Very interesting series of measurements. Especially off axis (is excellent) What model Azura horn and driver are you using ? . I ask as I’m about to start measuring a B&C 464 coax in an AH 120. Watching on ,,,,, have fun Thanks mate. It’s the Azura 340 with a JBL 2440, titanium diaphragms and some extra felt behind the diaphragm. Ive been considering those Eastern European paper and light copper diaphragms by the ebay seller who does them for most any CD… Interesting project with the larger horn and coax you have there - I’ve considered the full range 8” in horn on bass bin before as an alternative to this. You may find interesting this article on the Le cleach style horn - Foam in the mouth
Keith_W Posted March 9 Posted March 9 I should have been more specific and said "zoom in to the first 20ms of the ETC". Things are a little too compressed for me to see in much detail, but never mind. I normally also view in dBFS rather than %. I have annotated your ETC so we know what we are looking at. After the main impulse, reflections start arriving at the mic. I have marked two reflections 1 and 2. Hover your crosshairs over either reflection, and REW will tell you the time-lag to the main impulse. If you look closely at your tweeter chirp, you will also see the same reflections, except that the second reflection is higher in amplitude than the first. Let us assume that 1 is 1.5ms. This means that the reflection had to travel an extra 1.5ms to reach the mic. The distance can be worked out with the speed of sound, d = t/1000 * c (t = time in ms, c = speed of sound 343m/s). The distance is 0.51m. Now look around your room and decide what could have contributed to an extra 0.51m distance - maybe you are 25cm from the wall, maybe there is a sofa 25cm away, or a side wall, and so on. The time-lag also tells you the lowest frequency you can measure. Let us assume that 1.5ms is the period (T) of that frequency, so f = 1000/T. So all frequencies below 667Hz will be contaminated by reflections. Many people use @RoHo's method using the impulse response to view reflections as described in the post upthread. But I find the ETC to be much easier to read since the reflections are more obvious. It's up to you, it's the same measurement, just a different view. All these views of the impulse response (impulse, step, ETC) are like choosing whether to wear sunglasses. It depends on the conditions and what you want to see. If you wear sunglasses for the wrong conditions, it worsens your vision but what you are looking at is exactly the same. To find optimum measurement distance from your horn, I would go 0cm, 50cm, 100cm, 150cm (instead of 100, 110, 120, 130 which is what you did). Then apply 1/12 smoothing to all graphs, align SPL, and overlay them. Look carefully to see where the graphs start to deviate. Although theory says you should measure from 2 horn diameters away, sometimes you can get away with measuring much closer. Closer measurement = more time-lag between impulse and reflections = able to measure lower frequencies without contamination by reflections. 2
playdough Posted March 9 Posted March 9 4 hours ago, crtexcnndrm99 said: Thanks mate. It’s the Azura 340 with a JBL 2440, titanium diaphragms and some extra felt behind the diaphragm. Ive been considering those Eastern European paper and light copper diaphragms by the ebay seller who does them for most any CD… Interesting project with the larger horn and coax you have there - I’ve considered the full range 8” in horn on bass bin before as an alternative to this. You may find interesting this article on the Le cleach style horn - Foam in the mouth Yes interesting reading thank you I’ve done similar with a wool blanket laying on the PSE144 as a coincidence. Sorry about the bike in the photo I’m a long way away from that system for a better photo. Results from doing that were a slightly narrower propagation pattern and better frequency response plot at 320hz the transition between the 21” below and the horn and just sounded smoother and very slightly sweeter. I’ve had the AH120 made with an adapter for 8” driver in case the B&C is a failure and at that point order the 340. Do want a horn loading/transition below 300 ish hz though or to omnidirectional is the goal. Trying different CD diaphragms will be fun, that and modifying the driver. Not something I’ve played with yet. cheers from TAS where today is 30 degrees and the roads are melting. 1
La scala Posted March 18 Posted March 18 Satisfactory Room measurement correlation to actual desired sound quality seems a rather limited guarantee. Sure, its a brilliant tuning tool to see how ones room combines with ones Speakers and to id discrepancies, enabling to test out for positive remedies, wish things were that simple but ultimately ones ears are the final decider. Little article may be of interest to us System Tuners.. https://www.soundstageultra.com/opinion/20100501.htm 2 1
BugPowderDust Posted March 18 Posted March 18 20 hours ago, La scala said: Satisfactory Room measurement correlation to actual desired sound quality seems a rather limited guarantee. Well, if all you are looking at is frequency response, you're leaving a lot of data on the floor. Looking at phase and frequency is important, as is the impulse response, the decay in the room and waterfall and the spectrogram. Each gives you a different dimension on your listening and the in room response. In the right hands, looking at these measurements WILL tell you how measured response correlates to listening experience, much as looking at the spinorama of speakers is a very high level indicator of what will sound good in room. 2
almikel Posted March 19 Posted March 19 11 hours ago, BugPowderDust said: Well, if all you are looking at is frequency response, you're leaving a lot of data on the floor. Not quite. 11 hours ago, BugPowderDust said: Looking at phase and frequency is important, as is the impulse response, the decay in the room and waterfall and the spectrogram. Each gives you a different dimension on your listening and the in room response. Agreed - each measurement gives you a different view - but of the same data - with no data left on the floor, just different ways of looking at the same data - all important. One of the smartest and best contributors here on SNA - @davewantsmoore, taught me a long time ago that the amplitude/phase responses in the frequency domain are inextricably tied to the time domain responses (eg impulse response). In simple terms - a wiggle in one measurement always shows corresponding wiggles in the other measurements. Of course the hard part is interpreting the measurements, and implementing changes to improve the sound! On 18/03/2025 at 1:38 PM, La scala said: Little article may be of interest to us System Tuners.. https://www.soundstageultra.com/opinion/20100501.htm Reading the article, I agree with the author that measurements can't well determine resolution, transparency, soundstaging, imaging, etc., but I disagree completely with the authors opinion that the FR responses of the 2 speakers are perfect overlays. Based on the FR of each I'm not at all surprised that the Rockport sounds better than the Dynaudio - the Rockport's FR is smoother. For some bizarre reason the author dismisses the differences in FR because he interprets them as room effects? The author misses the point. It doesn't matter what causes the lumpiness in the FR, a smoother "in room" FR sounds better, based on good science documented by Toole, Olive etc. And a smoother FR will show less wiggles in the corresponding phase and time domain responses. Mike 3
playdough Posted March 19 Posted March 19 2 hours ago, almikel said: In simple terms - a wiggle in one measurement always shows corresponding wiggles in the other measurements. True, even at the start of measurements before a driver goes into/on to a baffle, the Impedance and phase of the electrical small parameters, wiggles in the driver bandpass will correlate to wiggles in the actual response at the same frequency, very handy to know when in design. This speaker placement guide popped up a couple of days ago,,,,,,rather interesting demonstration, starts of a little slow, gets cracking at Demo #2, boundary effect = smearing of the soundstage 1
almikel Posted March 23 Posted March 23 On 19/03/2025 at 10:21 PM, almikel said: One of the smartest and best contributors here on SNA - @davewantsmoore, taught me a long time ago that the amplitude/phase responses in the frequency domain are inextricably tied to the time domain responses (eg impulse response). In simple terms - a wiggle in one measurement always shows corresponding wiggles in the other measurements. This is a bit off topic, but relevant... ...I best make a clarification before @davewantsmoore or @Keith_W (another very clever contributor here on SNA) correct me... ...This is absolutely true for real world environments (eg speakers, rooms, reflections) and Infinite Impulse Response (IIR)/minimum phase filters (eg passive and electronic analog Xovers, DSP minimum phase EQ etc). With Digital Signal Processing (DSP), you can also implement Finite Impulse Response (FIR)/linear phase filters. With DSP FIR/linear phase filters, you can independently adjust the amplitude and phase responses in the frequency domain, which translate to the time domain response via the Fourier transform. Both Dave and Keith will have use cases for when "adjusting amplitude independently from phase" is a good idea, and when it's a bad idea. I'm a bit of a Luddite when it comes to EQ, remembering the bad old days of graphic EQs, and how much evil poorly applied EQ can do. I prefer to keep things simple, and only apply minimum phase EQ corrections to regions of the FR that I'm reasonably confident are actually minimum phase, and avoid EQ elsewhere. I like the fact that with minimum phase filters, "a wiggle in one measurement always shows corresponding wiggles in the other measurements"... ...my brain melts when I consider EQ/filters that don't do this (ie using linear phase filters to muck with amplitude independently from phase). cheers, Mike 1
Recommended Posts