It would make sense that a stopwatch be required here. A "fast" tone would not have been slowed down by stumbling (read "fat") into the 10db hole.
You should take into consideration that this is unsmoothed measurement and as such it is very good. Smoothing at low frequencies is necessary if we want to se what's happening.
That being said, smoothing of at least 1/6 octave at low frequencies should show situation at low frequencies that can be interpreted by everyone.
What was the smoothing of bass frequencies shown in this video clip ?
Earl Geddes on Mulitple Subwoofers in Small Rooms - YouTube
Last edited:
Where did I claimed that linked image - or any other single freq. response magnitude shows how this speaker compares to others or even sounds? What I'm slightly surprised that you confess to have close to zero self-confidence or actuals skills to interpret correlation between measurements and listening, or don't understand hifi terminology. Successful speaker engineering is not just looking simplified simulations in restricted (virtual) environment like your "small room" and calculation of standard deviations. Why to prioritize e.g low excess group delay, flat response or high dynamics over anything else if one is not able to identify sound properties from measurements. I've learned something within my 30 years diy career, but that's nothing compared to your impressive 50 years.
Yes, it is 1804 (lightweight and slightly resonating) prototype. Result is really impressive, but room & placement is very good for cardioid. Left ch. is not that impressive due to rear wall.
Dr. Geddes,
Perhaps we could slow down a bit and take it more rigorously. You know, just to make sure no one is stretching things to make his point 😉
So far so good...
Agreed. Detection time is far longer than reflection arrival time.
..but I think here you've jumped a bit to the conclusion. In the above, you are basically saying you expect steady state at 70Hz to happen after less than 60 ms (detection time). That's roughly 4-5 sound roundtrips in your room (and indeed a few times shorter than 224 ms)
Now looking at Linkwitz's simplified analysis in time domain, which looks sound to me when looking at only one direction, reaching steady-state depends on whether:
- you are exactly on a room mode peak. This takes long to reach steady state (far longer than your figure above and Linkwitz uses a fraction of RT60 as an estimate, which makes sense to me), as you need to pump a lot of energy into the system to build it up. It can actually take 150-200 ms and more, a few times ear detection time.
- you are on a dip or between two peaks (modes). A (kind of) steady state is reached much more quickly
So, at least at the room modes, reaching steady state (or actually 90% of it) _can_ actually take far longer than ear detection time or, conversely, the ear might have the time to detect low frequency sound before steady state is reached at specific frequencies.
This fact does not make steady-state less important (as the time domain behavior obviously largely depends on it !). But, for me, it might have some implications for what we call "articulate bass" (please note I've avoided "fast" and the like).
- considering one monopole source (or even two) at a modal peak frequency. Even if we equalize that peak, it will still take time to build up - longer than the ear detection time. What the ear would detect would be a slowly raising tone, at which point other psychoacoustic effects might chime-in (masking ? not sure !). Add the fact that the same mode will take just as long to decay - not a good thing at all ! Same Linkwitz uses burst tones to test bass: this should be tale telling.
- now looking at your multiple subs approach from this perspective. The mode is still there, but now you have at least an additional source radiating at the same frequency which does NOT (or does less) excite that mode. Consequently, the ear does not detect only the slowly raising mode, but also potentially the "steady-state" of the second source at the same frequency, which takes much less to build up because it does not excite the node (as per Linkwitz's analysis above). So, beyond reaching a much more uniform steady-state frequency response, you also reach a better time-domain behavior from the ear's point of view. But I think this is also what you ultimately claim 😉
- a dipole (or any kind of directional source) would not be fundamentally better than a monopole, but it is actually a "multiple-source" - actually mostly along the longitudinal axis (listening direction). Thus, the same consideration as for multiple sources might apply ().
To conclude: while the steady state response is still primordial, it just does not look as if we can just dismiss time domain behavior on the argument that the ear cannot detect it.
Perhaps we could slow down a bit and take it more rigorously. You know, just to make sure no one is stretching things to make his point 😉
You've ended up with 150 Hz, which is not that many times lower.
So far so good...
Agreed. Detection time is far longer than reflection arrival time.
..but I think here you've jumped a bit to the conclusion. In the above, you are basically saying you expect steady state at 70Hz to happen after less than 60 ms (detection time). That's roughly 4-5 sound roundtrips in your room (and indeed a few times shorter than 224 ms)
Now looking at Linkwitz's simplified analysis in time domain, which looks sound to me when looking at only one direction, reaching steady-state depends on whether:
- you are exactly on a room mode peak. This takes long to reach steady state (far longer than your figure above and Linkwitz uses a fraction of RT60 as an estimate, which makes sense to me), as you need to pump a lot of energy into the system to build it up. It can actually take 150-200 ms and more, a few times ear detection time.
- you are on a dip or between two peaks (modes). A (kind of) steady state is reached much more quickly
So, at least at the room modes, reaching steady state (or actually 90% of it) _can_ actually take far longer than ear detection time or, conversely, the ear might have the time to detect low frequency sound before steady state is reached at specific frequencies.
This fact does not make steady-state less important (as the time domain behavior obviously largely depends on it !). But, for me, it might have some implications for what we call "articulate bass" (please note I've avoided "fast" and the like).
- considering one monopole source (or even two) at a modal peak frequency. Even if we equalize that peak, it will still take time to build up - longer than the ear detection time. What the ear would detect would be a slowly raising tone, at which point other psychoacoustic effects might chime-in (masking ? not sure !). Add the fact that the same mode will take just as long to decay - not a good thing at all ! Same Linkwitz uses burst tones to test bass: this should be tale telling.
- now looking at your multiple subs approach from this perspective. The mode is still there, but now you have at least an additional source radiating at the same frequency which does NOT (or does less) excite that mode. Consequently, the ear does not detect only the slowly raising mode, but also potentially the "steady-state" of the second source at the same frequency, which takes much less to build up because it does not excite the node (as per Linkwitz's analysis above). So, beyond reaching a much more uniform steady-state frequency response, you also reach a better time-domain behavior from the ear's point of view. But I think this is also what you ultimately claim 😉
- a dipole (or any kind of directional source) would not be fundamentally better than a monopole, but it is actually a "multiple-source" - actually mostly along the longitudinal axis (listening direction). Thus, the same consideration as for multiple sources might apply ().
To conclude: while the steady state response is still primordial, it just does not look as if we can just dismiss time domain behavior on the argument that the ear cannot detect it.
Last edited:
If I'm doing this FAST, it's probably not going to be horn-loaded 🙂 Still, thank you for sharing your experiences. If I may ask, what was the order of crossover between mid-bass and TB1772? This wideranger is certainly good for one octave below 250 Hz.
The main question for me is the topology of crossover. Basically, I have no experience in DIY audio, so I will have to keep it simple and cheap. But I also want to make it close to optimal (because otherwise I could just use my Microlab Solo 6c, which are cheap but not bad 2way speakers). This means I need to achieve:
1) Point-source
2) Phase/impulse-coherent
3) Flat frequency respose design (but I will EQ anyway).
This is pretty doable with a modest FAST design, but this leaves me with several options for crossover network:
1) Simple first order XO, maybe passive line level. Easiest to do, but I'm not sure about this rule of at least two octaves flat for each driver. If I use a 8inch woofer and a 3inch fullranger, than a fullranger is a weak link since it will play from maybe 150 Hz. This pushes XO point to 600 Hz, which will probably violate the point-source quality of a speaker, since the drivers will no longer be 1/4 wavelength from each other.
2) Kreskovsky's Subtractive Delay Constant Voltage Linear Phase XO are 2nd order, which is good, but are too complex for me and I'm not sure that I could even emulate them with MiniDSP.
3) 2nd or 4th order Linkwitz + time delay on woofer, probably with MiniDSP. I remember a paper from Kreskovsky that the time-alignment allows to approximate the time domain qualities of 1st order XO.
4) FIR filtering. That seems to me like a brute force solution for achieving a 'perfect' impulse response, but is way more expensive than a 1st order XO. Basically, I'd need a miniSHARC, two DACS, digital input boards. I'd really rather use a cap and a coil.
What do you think?
You are correct, that is ultimately the point - a smooth response will also have a good time response. The Fourier transform guarantees that.
That there are exceptions to the rule that steady state behavior is dominate at LF and that the situation is not simple are both quite true, but I have not seen anything that says that time domain behavior is a significant effect at LFs, except perhaps in the exception case of a very resonant undamped mode, which we would all agree is a very detrimental thing to have. In a well LF damped room like mine and all that I design this situation would never occur and the steady state is all that matters.
That will not solve the issue of phase linearity, besides I'm not really interested in off-axis response of speakers as I'm going to listen from maybe 0,5-1 meter away from them.
"If I'm doing this FAST, it's probably not going to be horn-loaded Still, thank you for sharing your experiences. If I may ask, what was the order of crossover between mid-bass and TB1772? This wideranger is certainly good for one octave below 250 Hz."
48 db/oct Linkwitz-Reilly. Not much happening off axis, since the ribbon tweeter is horn load as well. Besides, my listening room is very heavily treated with acoustic absorbent. I don't much care for any reflections, particularly those from open baffle di-poles.
Agreed about the TB 1772. I have been entertaining buying another pair and make a waveguide even larger than the horn I am using. That way I could go down to 150Hz and capture a larger portion of the male voice.
The Fourier Transform has no sense at all since it's established that it works only in an invariant system. I guess that everybody assigns the listening environment to be invariant, well...
Earl, you can EQ any response to smooth and flat (with a given amount of smoothing, as with any measurement), even if it has several clear impulse peaks (for example strong reflexions, or multiple sources...).
A smooth response does not necessarily imply a good time response, and the Fourier transform certainly does not guarantee anything like that.
A smooth response does not necessarily imply a good time response, and the Fourier transform certainly does not guarantee anything like that.
So you chose an 8th order filter? That's rather steep. Is that an electronic network order, or a resultant acoustic slope of the horn? Have you time-aligned the outputs of woofer and TB 1772?
I wonder if I am going to need time-alignment anyway, even if I go with first-order crossover. If we take a woofer and a small fullrange driver, their acoustic centers are not going to lie in the same plane, which is probably going to screw with the goal of achieving transient-perfect response.
As a matter of fact, things seem to be again more complex than "LF room response is minimum phase". Here's a good (IMHO) description of what happens at LF in terms of minimum phase:
Minimum Phase
While a single mode is a minimum phase system, the superposition of several may not necessarily be (because adding the responses of several minimum-phase systems is not necessarily minimum phase).
In particular, dips are not minimum phase (a good reason for not equalizing them 😉). But room mode peaks are MP, and equalizing them out does result in a corresponding improvement in time domain as previously discussed. Thus, multiple/directional sources do actually improve time domain by 1. equalizing MP peaks out and 2. filling in the dips. You cannot achieve the same result by merely equalizing one single source flat.
I'm not Earl, but no, you actually can't. Think about a LF deep modal dip: you can't equalize it. No matter how much energy you put into it, it will suck it out - see my previous post on minimum phase.
At LF, where things are mostly minimum-phase except for dips that you can't EQ, but only "fill-in" (with additional sources) the Fourier transform does imply a good (=as good as it gets) time response if the frequency response is uniform.
Last edited:
That is why I stated "with a given amount of smoothing".
Earl uses critical band averaging in his measurement, that would easily smooth out any sharp dip in LF (ie make them "EQable" for the given measurement settings)
True:
That is why i told Mr Earl that measurements made by Mr Kimmosto is unsmoothed (which he knows for sure but failed to mention) and dip that can be seen at 110Hz that Earl adressed to as a bad thing is too narrow to show up on smoothed response that Earl uses when measuring his own speakers. I'd like to see how unsmoothed response looks like for any of his loudspeakers and compare it to Kimmostos. For any loudspeaker anybody made for that matter.
Last edited:
Bzfcocon - its good to know that someone understands. You have it absolutely correct.
Others - I do not use smoothing when I setup the EQ for a rooms LFs. I only do that when I show the frequency response of a loudspeaker.
Others - I do not use smoothing when I setup the EQ for a rooms LFs. I only do that when I show the frequency response of a loudspeaker.
Wouldn't be very informative. Different speaker types should be compared in the same room in their best and still acceptable positions i.e. possible WAF and all other sound qualities evaluated.
Flexibility/adaptability should be tested in a same way, but in several different rooms (acoustics, dimensions, ..) and listening setups.
For example, if I compare super-cardioid prototype and studio monitor with 10" woofer in our living room, and primary listening position would be only possibility due to WAF, this result would be quite valid (red=super-cardioid+horns, blue=3-way monopole):
An externally hosted image should be here but it was not working when we last tested it.
Otherwise it would be very smart thing to search better position for monopole speaker.
Last edited:
- Status
- Not open for further replies.