So I was reading about a bleeding-edge Manger speaker where they favour 96dB/8ave slopes. So I thought I'd see if my tweeters should come in at L-R48 instead of L-R24 (only tweeter moved), at 3kHz?
With a DSP, takes about 20 seconds to switch from one slope to the other. Or to change the XO frequency a bit. Or to..... And it takes just 3 seconds to replace the old setting memory with the new settings (Behringer DCX2496).
Attached are L and R systems with the flatter curve in each being the .... you can guess for yourself (or read the labels). Distortion a bit lower with 48 but impulse - at least in first milli-second something of a toss-up.
I just can't understand how a DIY person can be without DSP, even if just for setting up their HiFi.
B.
With a DSP, takes about 20 seconds to switch from one slope to the other. Or to change the XO frequency a bit. Or to..... And it takes just 3 seconds to replace the old setting memory with the new settings (Behringer DCX2496).
Attached are L and R systems with the flatter curve in each being the .... you can guess for yourself (or read the labels). Distortion a bit lower with 48 but impulse - at least in first milli-second something of a toss-up.
I just can't understand how a DIY person can be without DSP, even if just for setting up their HiFi.
B.
Attachments
Last edited:
The time domain response of a 48dB/oct filter (or higher) "traditional/analog like" (e.g. IIR digital) filter will have very large group delay peaking at the crossover point because there are one or more higher Q poles. The passband group delay will also be different near the crossover point compared to a filter with a different attenuation rate.
GD peaking will create some ringing and waveform distortion in the time domain. 4th order e.g. LR4 is already has some GD peaking but it's relatively benign in the time domain. SL wrote some info on this, which can be found at the bottom of this section:
Frontiers - Group delay and transient response
I'm not sure how the Behringer unit implements this filter, but my guess is that it is not the linear phase FIR version. Using linear phase FIR filters to compare attenuation rates would be a better way to isolate the influence of the rate without the additional confounding influence of a difference phase response and a changed passband group delay.
GD peaking will create some ringing and waveform distortion in the time domain. 4th order e.g. LR4 is already has some GD peaking but it's relatively benign in the time domain. SL wrote some info on this, which can be found at the bottom of this section:
Frontiers - Group delay and transient response
I'm not sure how the Behringer unit implements this filter, but my guess is that it is not the linear phase FIR version. Using linear phase FIR filters to compare attenuation rates would be a better way to isolate the influence of the rate without the additional confounding influence of a difference phase response and a changed passband group delay.
You say, "...to compare attenuation rates would be a better way to isolate the influence of the rate without the additional confounding influence of a difference phase response and a changed passband group delay..."
I do not actually share your view that the without-end arm-chair contemplations you endorse is better than quick tests with a DSP.
I also think the basic criterion - frequency response - defines sound quality a lot more than the generally dismissed notion of hearing phase in music reproduction.
B.
I do not actually share your view that the without-end arm-chair contemplations you endorse is better than quick tests with a DSP.
I also think the basic criterion - frequency response - defines sound quality a lot more than the generally dismissed notion of hearing phase in music reproduction.
B.
....individually equalize the driver's response, apply time alignment to achieve a smooth amplitude (and phase) response, which each of the four examples in the OP are grossly lacking.
I just can't understand how a DIY person extolling the virtues of DSP would not use the functions it has to achieve a reasonably flat response 😕
The Behringer DCX2496 uses Infinite Impulse Response (IIR) filters which behave exactly like their analog counterparts.
FIR filters do not, they can be arbitrarily set to any combination of amplitude and phase response but they need a much more powerful DSP engine than the old DCX2496 has.
Last edited:
When you select misbehaving drivers, no amount of slope will rescue the outcome.
What about phase? Have you ever consider that?
What about phase? Have you ever consider that?
Last edited by a moderator:
I can see you have been looking at one-meter FRs which can be quite well behaved. Or perhaps those swell sim plots.
I measure at my seat which FRs are rarely well-behaved - and always messy with my dipoles. Therefore, I will not take offence at your implied criticism.
By the way, I use electrostatic panels and Foster E120 ribbon tweeters (may have been used in the wonderful Infinity top speakers). If you have better behaved drivers, you should share your secrets with the rest of us.
Now and then REW puts phase information on my screen but I remove this rarely useful information as quickly as I can. Anyway, you can't be serious about wanting to see what phase information my mic is falsely picking up at my chair.
I have no personal "agenda" for starting this thread except to remind people how convenient it is to empirically test their systems and iterate improvements using dial-in DSPs.
B.
Ben,
In terms of polar response, electrostatic panel behavior is quite erratic compared to any well designed standard cone or horn speaker using decent components.
Ribbon tweeters are generally thin enough in width that their horizontal dispersion is smooth, though vertical dispersion is quite narrow, and also erratic outside a narrow "window".
Your choice of speaker design does make achieving flat (or smooth) frequency response impossible other than in a single location, regardless of room response.
Are you also low passing the panels at 3kHz, or just adding the tweeters with the LR24 or LR48 filters ?
Art
Last edited:
It is a baffling one, isn't it? 🙂
The fact is, some folks like to understand the underlying concepts and implementations of analog solutions, and enjoy that process. You learn much more doing it that way.
DSP is good for development purposes and for high-slope crossovers. But pretty much any workable solution can also be achieved via analog means.
Dave.
Yes, workable solutions can be achieved by analog means (esp active analog). Very good to understand the principles. Agreed.
But there are real limits to what arm-chair analysis can anticipate for a given room dimensions and furnishings, for drivers of various polar responses, how the speakers are placed relative to floor, ceiling and walls, and so on... and the material used in your carpet.*
Analysis and simulation - even to three decimal places - can be great fun but it is still a very loose guess at how your mic will respond at your chair and even far looser for your ears.
Anybody who designs a XO without an L-pad or other empirically-derived driver gain management, has too naive a grasp of room reality and human hearing. Not to mention adjustment of lots of other parameters.
B.
* Would an audiophile with good hearing ever play music a room without a large wool carpet?
But there are real limits to what arm-chair analysis can anticipate for a given room dimensions and furnishings, for drivers of various polar responses, how the speakers are placed relative to floor, ceiling and walls, and so on... and the material used in your carpet.*
Analysis and simulation - even to three decimal places - can be great fun but it is still a very loose guess at how your mic will respond at your chair and even far looser for your ears.
Anybody who designs a XO without an L-pad or other empirically-derived driver gain management, has too naive a grasp of room reality and human hearing. Not to mention adjustment of lots of other parameters.
B.
* Would an audiophile with good hearing ever play music a room without a large wool carpet?
Last edited:
Such a newcomer may be unfamiliar with the effects of a room. What are your thoughts on creating a useful speaker/room interface?
You're veering way off your original premise now. 🙂
Altering speaker response to alleviate room interactions or measurement anomalies is quite another thing.
Also, I don't how the topic of tweeter attenuation and L-pads came up.
You were originally talking about steep-slope crossovers. Any DIY'er knows driver relative sensitivities might not match and you need adjustment. That's not really related to whether the solution is done in the analog domain or DSP. It can be accomplished in either way.
Dave.
Having originally made reference to the manger msw, FIR filters can work wonders making this driver sing. Problems with cheaper DSP however are noisey input/output buffer stages that degrade macro-dynamics and add high order harmonics, spoiling the benefits and making the DSP implementation unsuitable for a serious high end audio experience. I've modified my fair share of "cheap" processors and they all end up costing a fortune to properly mod into a high end performing device. That being said, I use an RME ADI Pro2 for playing digital sources, running them through DSP via PC and then converting to analog.
I have a set of Manger 109s that are passively corrected, paying careful attention to the dreaded 1.5k THD peak with a series resonance notch. They sound magical in a pure analog signal path. I tried all sorts of DSP with these (a la Overkill Audio) and never liked the results. The more beneficial, steep DSP filters would shift the imaging in a weird way vertically and you would notice the transition point from cone driver to Manger. Shallower slopes help, but that defeats the benefit which is also negated by the added AD/DA stages when playing pure analog sources. Carefully implemented passive LR2 filters do the trick for me without the extra active audio stages adding their own sound signature.
I have a set of Manger 109s that are passively corrected, paying careful attention to the dreaded 1.5k THD peak with a series resonance notch. They sound magical in a pure analog signal path. I tried all sorts of DSP with these (a la Overkill Audio) and never liked the results. The more beneficial, steep DSP filters would shift the imaging in a weird way vertically and you would notice the transition point from cone driver to Manger. Shallower slopes help, but that defeats the benefit which is also negated by the added AD/DA stages when playing pure analog sources. Carefully implemented passive LR2 filters do the trick for me without the extra active audio stages adding their own sound signature.
I agree with Ben. DSP should be ahead of DAC for best results for sure.
I currently use LR48db all around; it seemed hopeless in the beginning though.
I currently use LR48db all around; it seemed hopeless in the beginning though.
"Any DIY'er..." - funny, I rarely see folks talking about L-pads in this forum. Seems more talk about buying drivers and blindly trusting the manufacturer's specs and your room's absorption.
If you need to reduce your woofer, you can't do it with the usual passive analog XO because you can't stick resistance in series with the driver. Any DIY'er knows that.
But with DSP or other bi-amping, simple to test and adjust.
I used to have a math teacher who said you can create a formula to graph an image of an elephant - with enough variables. Likewise for analog XO. The famous BBC LS3/5a two-driver speaker had (that is, needed) 18 elements in the XO.
B.
Last edited:
Agreed, but can you give me one reason why I should use an L-PAD in an active system with a DSP crossover ?
Regards
Charles
I meant no offense. I did not know you measured that far away. If you do, you are essentially measuring the room.
When I work on the crossover, I work in close proximity, because I want to design crossover without the influence of the room. I want to design speaker to be used in any room.
Once the speakers are designated to be in particular room, then I use behringer ultracurve to dial in flat fr response for that particular room.
So I do not need to mod crossover to do so.
That's just my rational.
To reduce the noise output from a very high sensitivity tweeter, where some attenuation is necessary anyway, like with the JBL M2 for example.
Any DIY'er knows that 99% of the time the woofer is more sensitive than the companion tweeter and doesn't need attenuation.
But, as a matter of fact, there are some designs that have a series resistor on the woofer to alter the Q behavior and/or reduce sensitivity. (It used to be more popular years ago, but I don't see it done much lately.)
Anyways, your premise seems to be DSP can solve all problems so why use anything else? I think that question has been answered. You just don't seem to understand the answer.
Dave.
Last edited: