Edgar

Oh, good heavens. I'm afraid that you have been misinformed. First, in a 16 bit CD, all 16 bits are available for dynamic range. Error checking is a completely separate issue, and there are many other bits devoted to that. Second, with proper dithering, if the recorded sound "falls below the 1-bit threshold", it does not disappear. Modern dithering and noise-shaping schemes can yield 16-bit dynamic range approaching 120 dB, at frequencies below ~10 kHz or so, at the expense of less dynamic range at higher frequencies, where it doesn't matter anyway because we don't hear so well up there. See, for example, here. Greg

Absolutely. It can be a problem when using pro equipment in a consumer situation. The expected input and output levels are different, so the output amplification on pro equipment can be so high that what would be a perfectly tolerable noise floor to other pro equipment is intolerably high to consumer equipment, even though there is absolutely nothing wrong with either one. The right way to deal with it is to place some attenuation between the outputs of the pro equipment and the inputs of the consumer equipment. Greg

Not set up at the moment, but back when I had my full system up and running, I ran directly from the SPDIF output of my CD player to the AES/EBU input (yes, they are compatible) of my digital crossover. Had a six-channel passive volume control (two speakers, each triamplified) between the outputs of the DACs and the inputs of the amplifiers. Wonderful. Greg

Anything that does not use a dual-range ADC. Modern monolithic converters are achieving very high dynamic range and SNR, and good sound, without the need for range switching.

Holy cow, has it actually been that long? The older I get the more quickly time passes. Nope; sorry, I've been out of that end of the biz for quite a long time. Greg

Sheesh, I'd have hoped that they would have gotten past that tired old design by now. Way back in 1995 or thereabouts, when I worked in the Merlin group (now long gone) at EV, we did comparative listening tests between our single-range ADC and the then flagship Klark-Teknik (also part of EV) dual-range ADC that they were very proud of. The range-switching in the KT converter was brutally obvious to anyone who knew what to listen for (DrWho, you describe it perfectly), and we told them so. But I guess you can't teach an old dog new tricks. Here it is, fifteen years later, and evidently their dual-range ADC still has the same problems. (Don't get me wrong -- it is still a very, very good sounding converter. It's just that more modern designs sound even better.) Greg

Amen. It amazes me that people will think nothing of spending thousands on a preamp, and then cheap-out on a $200 crossover. If you think about it, the crossover has a more difficult job than the preamp -- the preamp has to act as much like a "wire" as possible, and still sound good, while the crossover has to do some serious modification to the signal, and still sound good. Greg

I've never seen nearfield specs for any acoustical horn. Nearfield is usually not important for sound reinforcement. Greg

Be aware that a horn's polar response is only applicable to the "far field", while in a living room you are generally listening in the "near field". The coverage in the near field can be significantly different than in the far field. Where the near field ends and the far field begins is open to debate, but just understand that a horn is not like a LASER, where if you're not in the beam you don't see anything. The horn coverage varies gradually. The best way to see if what you propose works is to try it, if you possibly can. Greg

The beamwidth is defined by the angle at which the response is down by 6 dB. Greg

No, a 100° beamwidth means ±50°. Ninety degrees vertical would be 45° up to 45° down. Greg

Coincidentally, I was just discussing exactly this idea with a friend. My thought is that, at least in a living room situation, the horizontal beamwidth is limited by the room walls anyway, so it is more important to have tight beamwidth control in the vertical direction. Furthermore, a CD horn in portrait mode, in the corner of a room, might be a pretty good approximation to a line source. There is precedent for this here. Greg

USPS averaged 0.5 drops over 6 g's per trip. FedEx averaged 3 and UPS averaged 2. USPS averaged 12.5 position changes per trip. FedEx averaged 7 and UPS averaged 4. They also noted that packages marked "Fragile" received more abuse, and packages marked "This Side Up" were flipped more often.

There's an article about shippers in the Dec 2010 issue of Popular Mechanics. FedEx did not fare well in the tests.

In the paper, L&V provide examples of linear PCM configurations in which the data rate is lower than DSD, the noise-shaper design is simpler and the noise-shaping less severe, and the signal can be properly dithered (a 1-bit DSD signal can not). That is all that I meant by "outperform DSD". That is a very strong argument, indeed. From that standpoint, DSD is very attractive. I can do it. I hate to be mercenary, but is there any money to be made from it? Greg

People have definitely been working on it. I know of at least one patent for processing 1-bit signals directly (5,990,818; "Method and apparatus for processing sigma-delta modulated signals"), but it is not nearly as simple as processing linear PCM. Also, the infamous Lipshitz and Vanderkooy paper is worth reading, not because it points out the flaws in the format, but because in section 3 it shows how linear PCM can easily be made to outperform DSD, while retaining the inherent ability to be filtered. Greg

A little bit of both. DSD is a 1-bit format (the commercial version; the "Pro" version is multibit). Filtering involves scaling and adding multiple samples together. Do that to 1-bit samples and you get multibit samples -- there's no way around it (well, no simple way ... afterward you can re-encode to 1-bit, but that's not quite the same as staying in the 1-bit domain). In addition, trying to design filters that operate below 100 kHz, when the sampling rate is 2.8224 MHz, inherently causes numerical precision problems. Not insurmountable, but much easier to deal with when the ratio of the sampling rate to the frequency of interest is lower, within Nyquist constraints. It's not a coding problem. It's a mathematical problem. Greg

It's a fine idea, except that it's extremely difficult to apply EQ directly to the DSD bitstream -- it has to be converted to PCM first. Once in the PCM domain, techniques have already been implemented to match analog filters extremely well, either as magnitude-only (See posts by Robert Orban -- yes, that Robert Orban -- at http://groups.google.com/group/comp.dsp/browse_thread/thread/59232b93211c041a/1786c9d17474f221. Even though he discusses de-emphasis filters there, the same technique can be applied to EQ.), or as magnitude and phase (See posts by Greg Berchin -- that's me -- at http://groups.google.com/group/comp.dsp/browse_thread/thread/ce5cc4bfe193070/b05e9312fa36a7be.). So as long as you aren't adamant about keeping the data in the DSD format, yes, we have the technology. Greg

The graphic format is intentionally similar to that shown here. To me it is more intuitive than a polar plot.

I'm a little reluctant to post this, for fear that it will be misinterpreted. The attached graphic is the result of simulation, not measurement. It is not a polar plot, but instead it shows what the actual waveform might look like at an instant in time if the sounds were waves in a water tank, and you were looking down at the tank from directly above. The speaker is located in the upper right corner. The red areas are peaks and the blue areas are valleys. The simulation takes into account side wall, front wall, floor, and ceiling reflections, but it does not assume that those reflections are perfect (reflection coefficients of 0.7, 0.7, 0.7, and 0.5, respectively). This particular simulation shows 300 Hz, but I can simulate any frequency. Greg

Interesting. Danley shows the SPUD to be a lot flatter than that below 60 Hz. Where is your SPUD located in your room? It looks like there may be some significant room interactions in your frequency response graph. EDIT: Just read your other post regarding placement of the SPUDs behind the Jubilees. Perhaps orienting the SPUD mouth differently would change some of those peaks and valleys. Might be worth experimenting.

The H11 only goes down to about 70-80 Hz, however, compared with around 40 Hz for the Jubilee. THe H11 would make an interesting DIY project. Each contains a single Electro-Voice DL10x driver, though an EVM10m or Klipsch K41m would be essentially equivalent. It's a basic "W" fold design, and appears to be fairly simple to construct. It would be a good match for many curved-front radial HF horns like the HR9040a shown in the photo.

That's a Ray Newman design known as "H11". Jim Long is the one to ask about them as I believe he has the only pair in existence. Mr. Newman perished several years ago, and left only incomplete notes and sketches in addition to the two cabinets, copies of which Mr. Long has been kind enough to provide to me. I have a Hornresp model based upon assumptions that I had to make about many of the measurements -- as I said, the notes that Mr. Newman left are incomplete. I would be willing to make the information that I have available, but only if Mr. Long approves. Greg

You and I are in almost complete agreement. In fact, in an earlier post on this thread I said, "an argument could be made that curling could be better at really low frequencies". By that I meant that really low frequencies are able to turn corners just fine, whether or not a reflector or a radius is present, so a reflector is seen by the low frequency wavefront as a constriction. A constriction will show up as a discontinuity in the acoustical impedance, which affects the loading and the frequency response. The magnitude of the problem is entirely dependent upon the relative sizes of the constriction and the wavelength. Greg

Entirely true, but there's more to it than that. As can be seen in many of the diagrams shown here, a bend without a reflector can cause standing waves inside the horn that affect the polar response of the horn. What the diagrams don't show is that those standing waves also affect the loading characteristics of the horn, which, in turn, affect the frequency response. Basically, the only time that reflectors aren't necessary is when the wavelength is much longer than the dimensions of the horn (see here). In any other situation, reflectors are the way to go. Bruce Edgar (no relation) writes about exactly this here. Greg