That's a horizontal spread.. what would you expect of any transducer 3-4 inches wide at freq.s below 2 kHz?
As I see it it, for the Neo10, it will compare to vertical distribution as you have mounted the unit. The window is probably a bit more than + - 15 degrees looking from measurements. But you yourself has to estimate your horizontal spread as you have it. Looking at my thread: The BlindStone OB I really would try it with a small baffle to compare results.
/Erling
/Erling
See the data in the first post on this thread for the Neo10 in the vertical orientation (three inches width). Zaph, however, has published infinite baffle directvity data for both the horizontal and vertical orientations. As skorpion's hinted, the correct comparison to Greg's Z-65 data is Zaph's horizontal data---the relevant dimension there is eight inches.That's a horizontal spread.. what would you expect of any transducer 3-4 inches wide at freq.s below 2 kHz?
Any chance you could send me some of your HolmImpulse Zip files of the Neo 10 & 3 so I can compare them to mine when I mount them in a baffle.
I would also love to see What settings you use so I can learn some.
I seem to learn a lot faster from looking & playing rather than reading, just ask my primary school teacher
David
If I'd saved any of the measurements as .zips, sure, but that's not something I normally do so the screen grabs in the first post of this thread are what I have. I'll be remeasuring eventually to see if the EQ needs to be updated driver breakin; will try to remember to put up .zips. There's nothing special about the process, just log sweeps with an EMM6 from Herb.
I wish I new about that site before I bought my ECM8000, but I'm sure it will do just fine.
Calibrated
Herb will still calibrate any mic you have if you want to send it to him. The ECM8000 usually have some non flat response above 8KHz.
Hi folks,
I'm new to DIY, have been doing a lot of reading recently, and think I might want to try building a 3-way suspended naked dipole system. Here's a list of things I want to achieve:
1) Runs off my Macbook Pro (transport and crossover/DSP)
2) As little cost as possible for drivers and sound card (I have some amps laying around that should do for now)
3) As compact as possible (apartment use)
4) Suitable for nearfield listening (say, 3-6 feet)
5) Down no more than 3db or so at 40hz at moderate listening levels (again, nearfield)
6) Good to excellent off-axis response, low distortion, good FR
Here's what I'm thinking of using:
-Neo3 PDR (have some laying around unused)
-Zaph ZA14W08
-Dayton RSS315HF-4
-Focusrite Saffire Pro 24
-Reaper or some other DAW + misc. plugins
I'm thinking about running a cable high across the top of the room and suspending the drivers from a tube that slides along the cable (so they can slide out of the way when not in use).
I would welcome any helpful suggestions or criticisms, with the understanding that I am a DIY novice and acknowledge that I have a lot to learn with regard to the physics of audio reproduction. In particular, I'm having trouble wrapping my head around the concepts and nomenclature of phase relationships and how to treat phase problems — not that I expect anyone to go out of their way to tutor me, but if anyone can recommend reference materials that aren't forbiddingly technical I would certainly be grateful.
I'm also wondering about driver choices, of course, and about the Dayton 12 in particular. Given that my requirements for SPL and bass extension are quite a bit more modest than that of many others here, I'm hoping it will be sufficient. However, I might still be asking a lot from those 12's —*a lot of excursion, that is — and I don't know how much that would compromise the midbass/lower midrange if I ran them into the low hundreds. Also, I've seen a lot of glowing reports on these drivers, plus a few mentions of notable distortion, so I'm not quite sure what to expect.
For the record, I'm a musician (upright bassist/singer/composer) and have been listening at home through EPI 100's and Stax SR-202.
I'm new to DIY, have been doing a lot of reading recently, and think I might want to try building a 3-way suspended naked dipole system. Here's a list of things I want to achieve:
1) Runs off my Macbook Pro (transport and crossover/DSP)
2) As little cost as possible for drivers and sound card (I have some amps laying around that should do for now)
3) As compact as possible (apartment use)
4) Suitable for nearfield listening (say, 3-6 feet)
5) Down no more than 3db or so at 40hz at moderate listening levels (again, nearfield)
6) Good to excellent off-axis response, low distortion, good FR
Here's what I'm thinking of using:
-Neo3 PDR (have some laying around unused)
-Zaph ZA14W08
-Dayton RSS315HF-4
-Focusrite Saffire Pro 24
-Reaper or some other DAW + misc. plugins
I'm thinking about running a cable high across the top of the room and suspending the drivers from a tube that slides along the cable (so they can slide out of the way when not in use).
I would welcome any helpful suggestions or criticisms, with the understanding that I am a DIY novice and acknowledge that I have a lot to learn with regard to the physics of audio reproduction. In particular, I'm having trouble wrapping my head around the concepts and nomenclature of phase relationships and how to treat phase problems — not that I expect anyone to go out of their way to tutor me, but if anyone can recommend reference materials that aren't forbiddingly technical I would certainly be grateful.
I'm also wondering about driver choices, of course, and about the Dayton 12 in particular. Given that my requirements for SPL and bass extension are quite a bit more modest than that of many others here, I'm hoping it will be sufficient. However, I might still be asking a lot from those 12's —*a lot of excursion, that is — and I don't know how much that would compromise the midbass/lower midrange if I ran them into the low hundreds. Also, I've seen a lot of glowing reports on these drivers, plus a few mentions of notable distortion, so I'm not quite sure what to expect.
For the record, I'm a musician (upright bassist/singer/composer) and have been listening at home through EPI 100's and Stax SR-202.
Hey, missed your post. That's a fine set of drivers for a start and you can always move to larger mids and subs if you find you need more SPL. What I'd do is set up two master rods which slide on the cable---one for the left set of drivers, one for the right---and then hang each driver from its master with a pair of rods. That'll make it pretty easy to slide the speakers back and forth without too much swinging around.
I'm not aware of a turnkey linear phase solution for the Mac so most likely you're looking at warped phase. Maybe Waves LinEQ? The passband on the ZA14W08's wide enough it'll cross from the RSS315HF to Neo3s without great need for correction though. LR4 at 500Hz and 2kHz should be a reasonable starting point. The per channel delay equalization I'm doing here is an optimization to operate the drivers as low as possible---overkill if you just want to try things out.
I'm not aware of a turnkey linear phase solution for the Mac so most likely you're looking at warped phase. Maybe Waves LinEQ? The passband on the ZA14W08's wide enough it'll cross from the RSS315HF to Neo3s without great need for correction though. LR4 at 500Hz and 2kHz should be a reasonable starting point. The per channel delay equalization I'm doing here is an optimization to operate the drivers as low as possible---overkill if you just want to try things out.
Bruno and I discussed the options a while back and you can find my measurements and crossover and EQ configuration at the beginning of this thread. My experience is in room data correlates well with outdoor data. Which is good as around here it's quite difficult to get a day with sufficiently little wind and background noise to get good phase results outdoors. As you can see from the measurements I'm not trying to EQ everything the mic picks up---the reflections aren't early so they seem to be discarded well enough by perceptual processing.
I think I suggested offline processing rather than real time in a VST.
It's been a year and some since I did a survey but the primary options in VSTs seem to be Refined Audiometric's PLParEQ3 and Thuneau Frequency Arbitraror (also included in the full version of Allocator). You can find the Allocator configuration I use at the beginning of this thread; it's a reasonable compromise between CPU load and minimizing step discontinuities between processing blocks (Jan at Thuneau's never indicated one way or the other, but best I can tell his code does no windowing). PLParEQ3 does window but the limited number of filters per VST instance makes it computationally infeasible for my puposes. The full version (PLParEQ, which Stig Erik uses) addresses that but requires deeper pockets than I've got.
Last edited:
It's been a little over 11 months since I started this thread. I was planning on typing up an update on the build right around the one year anniversary. But I find myself with a bit of extra time today. So y'all are getting the update a bit early.
The majority of my DIY time over the past year has been devoted to exploring different options for moderate power amplifiers of a few hundred milliwatts per channel. That's off topic for this forum so I shan't go into it here. Suffice to say the system's currently operating off a precursor to this board and no one's complained about its ability to play comfortably loud yet despite dipole inefficiencies and the LME49600s clipping around 300mA per driver.
The other major item I've been working on has been figuring out how to do linear phase without requiring a PC crossover. I've had this working in various forms for the past several months but just recently got things tidied up to the point where they're worth discussing. I've changed over from playing CDs as-is on my increasingly unreliable Azur 640C v1 CD player to ripping the tracks on my laptop, running them through a time reversed IIR tool I wrote called Cross Time DSP, and burning them onto a USB flash drive for playback from a Squeezebox Touch. The Cross Time DSP home page (previous link) and pre-alpha documentation discuss the tradeoffs of this approach, the theory behind it, and a method for linearizing dipole phase at some length. So you can read about it over there if you wish and download today's pre-alpha bits to fool around with if you want.
This, of course, calls for pictures. The first two figures are output discontinuities caused by Arbitrator's lack of windowing not creating smooth transitions between blocks. As the 18SWS1100s output significant SPL up to 13kHz or so these discontinuities effectively reduce the speakers' SnR pretty much everywhere. The same issue occurs with the Neo10s, though the discontinuities are smaller. The whole point of writing Cross Time DSP was to get rid of these glitches (see discussion of stream based versus block based DSP in the links above) and, while I'm a little too close to this to qualify as an impartial observer, I'm scoring 100% discrimination in favor of stream based time reverse IIR in something that's arguably a reasonable approximation of single blind A/B testing. Changing to over to the Squeezebox and uprezzing to 24 bit seems to have wrung a bit more performance out of the system as well. A lot of that's probably just from getting away from the CD player's failing transport but blind A/B between 16 and 24 bit source files on the Squeezbox is strongly in favor of 24 bit. I'm not sure where the difference comes from, but I'll take it. When A/Bing between the two 16 bit sounds somewhat grainy.
The last two figures are the Allocator Lite and Bidule configurations I'm currently using. The Allocator Lite settings are just a copy of what I was using previously in the forward time portion of full Allocator. Since Thuneau VSTs don't support all pass filters the ReaEQ VSTs are added to Bidule to provide the dipole cancellation allpass discussed in the Cross Time DSP docs. Only the ReaEQ_0 instance actually has a cancelling allpass---the other instances are just there to match delays between channels.
So what's coming next? Finishing some power amp builds. Stabilizing the Cross Time DSP codebase and releasing a 1.0 (the DSP core is stable, verified to be accurate within numerical precision, and has processed hundreds of GB of data---the issue is all the glue code around it to make the engine go in a reasonably user-friendly fashion). Getting the code together for the Atmel AVR on the SigmaDSP + ESS DAC board so I can finalize the AVR pin assignments, post the schematic over on this thread, and start layout work. Sure, it'd be quicker and easier (and probably cheaper) to get rid of the PC crossover by switching over to a MiniDSP or DCX. But where's the challenge in that? I'm also fooling around with some less than half baked mechanical ideas involving counterweighted pendulums and thrust bearings.
The majority of my DIY time over the past year has been devoted to exploring different options for moderate power amplifiers of a few hundred milliwatts per channel. That's off topic for this forum so I shan't go into it here. Suffice to say the system's currently operating off a precursor to this board and no one's complained about its ability to play comfortably loud yet despite dipole inefficiencies and the LME49600s clipping around 300mA per driver.
The other major item I've been working on has been figuring out how to do linear phase without requiring a PC crossover. I've had this working in various forms for the past several months but just recently got things tidied up to the point where they're worth discussing. I've changed over from playing CDs as-is on my increasingly unreliable Azur 640C v1 CD player to ripping the tracks on my laptop, running them through a time reversed IIR tool I wrote called Cross Time DSP, and burning them onto a USB flash drive for playback from a Squeezebox Touch. The Cross Time DSP home page (previous link) and pre-alpha documentation discuss the tradeoffs of this approach, the theory behind it, and a method for linearizing dipole phase at some length. So you can read about it over there if you wish and download today's pre-alpha bits to fool around with if you want.
This, of course, calls for pictures. The first two figures are output discontinuities caused by Arbitrator's lack of windowing not creating smooth transitions between blocks. As the 18SWS1100s output significant SPL up to 13kHz or so these discontinuities effectively reduce the speakers' SnR pretty much everywhere. The same issue occurs with the Neo10s, though the discontinuities are smaller. The whole point of writing Cross Time DSP was to get rid of these glitches (see discussion of stream based versus block based DSP in the links above) and, while I'm a little too close to this to qualify as an impartial observer, I'm scoring 100% discrimination in favor of stream based time reverse IIR in something that's arguably a reasonable approximation of single blind A/B testing. Changing to over to the Squeezebox and uprezzing to 24 bit seems to have wrung a bit more performance out of the system as well. A lot of that's probably just from getting away from the CD player's failing transport but blind A/B between 16 and 24 bit source files on the Squeezbox is strongly in favor of 24 bit. I'm not sure where the difference comes from, but I'll take it. When A/Bing between the two 16 bit sounds somewhat grainy.
The last two figures are the Allocator Lite and Bidule configurations I'm currently using. The Allocator Lite settings are just a copy of what I was using previously in the forward time portion of full Allocator. Since Thuneau VSTs don't support all pass filters the ReaEQ VSTs are added to Bidule to provide the dipole cancellation allpass discussed in the Cross Time DSP docs. Only the ReaEQ_0 instance actually has a cancelling allpass---the other instances are just there to match delays between channels.
So what's coming next? Finishing some power amp builds. Stabilizing the Cross Time DSP codebase and releasing a 1.0 (the DSP core is stable, verified to be accurate within numerical precision, and has processed hundreds of GB of data---the issue is all the glue code around it to make the engine go in a reasonably user-friendly fashion). Getting the code together for the Atmel AVR on the SigmaDSP + ESS DAC board so I can finalize the AVR pin assignments, post the schematic over on this thread, and start layout work. Sure, it'd be quicker and easier (and probably cheaper) to get rid of the PC crossover by switching over to a MiniDSP or DCX. But where's the challenge in that?
I'm also fooling around with some less than half baked mechanical ideas involving counterweighted pendulums and thrust bearings.Attachments
Thanks a lot for sharing. You provided a lot of information that takes time to digest. (and, you didn't mention how it sounds after those you've done)
DSP stuff is beyond me, but this caught my eyes. Are you sure the HF response of your 18SW1100s is intrinsic, or something related to the source, especially those digital things?
Something related. Recently I tried upside down pendulum (metronome) on one of my sub. It's a very small OB, with 2 little rigid feet under its base. These 2 feet are alingned in lateral, placed more or less at the center (or under the center of gravity). So the assembly is easily swinging back and forth within a small angle - basically it can't stand upright by that 2 suppoting points.
So I added very soft foam pads at the front and rear to act as spring and damper to maintain the upright position.
The combo works pretty good. Very calm under high excursion operation of the woofer. It's not as perfect as a pure swing, which is deadly steady, but very close. With such a simple implementation, I'm happy with the result.
DSP stuff is beyond me, but this caught my eyes. Are you sure the HF response of your 18SW1100s is intrinsic, or something related to the source, especially those digital things?
Something related. Recently I tried upside down pendulum (metronome) on one of my sub. It's a very small OB, with 2 little rigid feet under its base. These 2 feet are alingned in lateral, placed more or less at the center (or under the center of gravity). So the assembly is easily swinging back and forth within a small angle - basically it can't stand upright by that 2 suppoting points.
So I added very soft foam pads at the front and rear to act as spring and damper to maintain the upright position.
The combo works pretty good. Very calm under high excursion operation of the woofer. It's not as perfect as a pure swing, which is deadly steady, but very close. With such a simple implementation, I'm happy with the result.
So what's coming next? Finishing some power amp builds. Stabilizing the Cross Time DSP codebase and releasing a 1.0 (the DSP core is stable, verified to be accurate within numerical precision, and has processed hundreds of GB of data---the issue is all the glue code around it to make the engine go in a reasonably user-friendly fashion). Getting the code together for the Atmel AVR on the SigmaDSP + ESS DAC board so I can finalize the AVR pin assignments, post the schematic over on this thread, and start layout work. Sure, it'd be quicker and easier (and probably cheaper) to get rid of the PC crossover by switching over to a MiniDSP or DCX. But where's the challenge in that?
Woah !
I will follow this thread closer than ever now....
Best from France
Jean Claude
Well, CD players with tracking troubles are nothing new. Neither are 24 bit samples. Removing the discontinuities from unwindowed, block based time reversed IIR is what you'd expect from a global SnR improvement; sound's a bit smoother overall and you notice more details if you're listening critically.
One possible explanation for 24 bit sounding better is it's a DAC thing. The DACs on the CS4272 codecs in the Saffire 40 I'm using for playback are 19 bit. The NJM4565 op amps Focusrite uses as output buffers are quiet enough to hold that, as are the LME49710/49990 and LME49600s on the power amp. The LME49713 channel loses a bit or so to current noise but, other than the noise floor, seems objectively and subjectively identical to the 49710 and 49990 channels. In other words, whatever's going on shouldn't be the result of hearing differences on the 17th or 18th bit. The power amps have unity gain and there's about 6dB of gain built into the Saffire's outputs, so with this system's efficiency and my typical playback levels the CS4272s are usually operating between -6 and -30dBFS, depending on which driver you're looking at and the music's PSD. So the music is playing back with around 12 bits RMS of dynamic range with 16 bit samples and 13 or 14 bits with 24 bit samples. That 16 bit sounds grainy and 24 bit sounds smooth when A/Bing between them therefore seems to suggest the difference between 72ish dB SnR and 85dBish SnR is audible. My place is pretty quiet but the ambient noise floor still is maybe 55dB SnR in the midrange and 70dB on the highs. The dynamic range of any given type of hair cells is about 45dB with the cochlear amplifier acting like a class H preamp to get the ear up to 130dB. While the brain's spatial processing allows you to hear things below the ambient noise floor of an omni measurement mike, directly hearing the SnR difference between 16 and 24 bit would seem to require ~25dB of noise rejection and cochlear amplifier slew rates an order of magnitude higher than published psychoacoustic data. Both of these seem improbable, so I'm inclined to reject this line of reasoning.
An alternate explanation for 24 bit sounding better is reduced truncation error when converting in and out of the double processing performed within Cross Time DSP and Allocator Lite or full Allocator. Since the source data from the CD is 16 bit it would initially seem improbable this matters. However, since inverse allpass processing moves spectral content between samples there's quantization in phase as well as in magnitude. With a typical music PSD and a typical recording a 1kHz sine would usually be around -40dbFS. At 44.1kHz positioning such a sine with 1 degree accuracy requires resolution to -75dBFS at its zero crossings, rising to the 24 bit precision limit of -144dBFS within a sample or two of its crest. The audibility threshold for phase errors depends somewhat on what psychoacoustic data one's looking at but the rule of thumb I've arrived at is you want it to be flat to better than 15 degrees between a few hundred Hz and a few kHz, ideally within 5 degrees if you can do it. This would imply 24 bit sounds better because the extra two or three bits it makes available allow more accurate time domain placement of waveform crests.
Yes; I have a variety of characterization data with various amounts of forward and reverse time processing. The only thing which moves the SPL is deliberate EQ. Besides, the primary value proposition for swept sine or chirp measurements over of MLS is their significantly better noise rejection---tossing delta functions into a properly implemented swept sine raises the reported noise level but hardly moves THD.
I've been fooling with similar ideas for providing restoring force on an invered but haven't actually built anything. Good to hear you're getting good results---something I'm realizing is I'm used to working with much longer travel systems than what's required here and don't really know much about the materials choices for short throw.
Something I've been thinking about for regular magnet and cone drivers is a one sided version of what you did, albeit with with a longer arm on the inverted pendulum. Set up right the arm would take just about all the weight of the driver, which would then rest gently on some sort of foam/sorbothane/low durometer whatever pad. One would probably want to arrange this so the driver was allowed to move by flex in the arm; I'm not sure how good bushings are for handling very small, very repeated angular displacements but I know bearings don't like that.
Another option is similar to the pantographs I'm already using. If you look back at the pictures in this thread you'll see my 18s are held up at the magnets and pulled down at the baskets since they're so magnet heavy I'd have to install fittings on the magnets to support them at or behind their center of gravity. The line which goes around the magnet and carries most of the weight of the driver could be replaced by a clevis sitting on a flexible post and both the post and basket tie down anchored to a plinth. That'd create this sort of elegant inverted pantograph kind of thing of about 45cm height which would be fairly difficult to tip over accidentally. Just have to choose the post material and size it to get a good spring constant.
In particular, the idea of two posts with a sling between them to hold the driver magnet seems worth trying. That's about halfway between regular post and lintel construction and some of the suspensions Stig Erik used for his subs before he switched over to bicycle tubes. It's possible a rope or webbing sling might provide all the flex needed just by itself.
Thanks. I've been kind of laid up the past year with a nasty deviated septum situation so have had more DIY time than usual. Spending more time outdoors and less time inside with the laptop now, so if I get Cross Time DSP finished up and the amps built over the winter that'll be pretty good progress. Figure a couple years past that to get the SigmaDSP running and improve the mechanicals.I will follow this thread closer than ever now....
I haven't decided where to post the amps yet. Maybe this thread for most of them, though the LM4675 will need to go in the class D forum. Assuming I manage to solder the 10-LLP, anyway.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.