Magnetar said:
Not a problem on the encode (A -> D) side of things. ADC's require a lot of signal conditioning, which is why transfers from magnetic-tape masters sound so good. The mag-tape process is a near-ideal signal conditioner, supplying limiting, plenty of dithering, and careful tailoring of dynamic range all in one go.
It is true the first-generation Sony 1630 ADC/DAC had the worst opamps of the day - the notorious 741, with its slew rate of 3V/uSec, a full-power bandwidth less than the audio spectrum, and lots and lots of HF distortion thanks to not-much feedback at HF and the quasi-complementary output stage. In other words, a really bad power amp in a tiny package. And the 1630 had plenty of these things, along with a generous handful of electrolytic coupling caps at each input and output of the 741.
Another gotcha of digital conversion is non-integer sample-rate conversion. It took almost a decade for the industry to realize that this klnd of conversion, even in the digital domain, requires dither in order to keep distortion to a minimum - and preferably dither with a specially-shaped spectrum, not just white noise. The Apogee UV22 algorithm, dating from the early Nineties, is probably one of the earliest to use this approach.
As for the tests Matt and I made about ten years ago, the transition speed coming out of the DAC is basically a function of the internal architecture of the chip. The old Philips 16-bit chip is internally slower than the very fast Burr-Brown 20-bit chip, so the transitions coming out of the current-mode output have a different dV/dT as well. It takes a very fast scope to actually see the dV/dT of the transitions, and verify the magnitude (if any) of glitching that might be present. This is where RF-style board layout and power-supply decoupling make a big difference to the overall outcome. RFI-emitting or RFI-susceptible loop areas have to be kept as small as possible.
I'm still in the camp that favors current-mode output that is coming directly off the ladder array, passive I/V conversion, and passive lowpass filtering using RF techniques mounted very close to the chip. Once all the RF stuff has been reduced, then the analog design becomes a more relaxed proposition, basically a linestage with 20~30 dB of gain, depending on the choice of I/V resistor. Tube, transistor, or opamps, whatever you like. One of the charms of transformer coupling is balanced-to-single-ended conversion (if you want that), studio-grade RFI filtering, and most importantly, isolation of the grounds of the digital and analog sections (no actual physical connection between D and A sides of the circuit, in other words).
Fortunately, measurements with the PCM-63 showed that once the I/V resistor fell below a critical value (100 ohms), then the distortion fell to the residue of the DAC itself (using test gear with a residue of 0.0008% at 20 kHz). Current-mode DACs give a certain amount of "voltage compliance" but that's all you get.
The power supply design for the analog section is not trivial - it requires techniques similar to what you'd do if you wanted to build a quiet phono preamp directly adjacent to a flourescent light. The digital lowpass filter is the main RFI emitter, since it is essentially a dedicated single-function computer that sits close to the DAC (or inside it), and runs a lot faster than the input receiver or the DAC conversion process.
Lynn Olson said:
I find it almost as hard as to select a proper video- opamp to find suitable passive parts in the RF department - like chokes - with sufficient audio qualities.
Also caps are not really easy to select as they can have their resonance and following impedance rise around the frequencies of interest and at the values needed for a purely passive first filter stage.
Some caps with very good RF performance – consulting data sheets – didn't sound best to me like polycarbonate types, providing an unnatural emphasised sound at the top end for example.
Greetings
Michael
pdan said:
Nice link, much appreciated. As someone that has always considered the Quad ESL57 and BBC monitors to be the touchstone of loudspeaker design, I am in 100% agreement with the interview. If a loudspeaker cannot reproduce natural, lifelike speech, I don't really see how it can be called "high fidelity" in any meaningful sense of the word.
And I am compelled to agree with Alan and Derek that fewer than 10% of the "high-end" speakers on the market can reproduce speech in a lifelike way - sad but true. I've heard speakers with very famous names at the $60,000 and $100,000 price points that cannot pass this elementary test. Most of the speakers at the CES and RMAF would fail, for example, since it was obvious just by walking into the room that the balance was grossly unnatural (usually very tipped-up with presence-peaking in the 3 to 5 kHz range).
I was talking to Bud Purvine about the currently fashionable sound in the high-end business - and he was pretty direct: he felt the market for this kind of sound are old gents who are losing their hearing, and need a lot of presence-peaking just to enjoy the music. He had just come back from the Portland VSAC show, and noticed the same thing as I did at the last RMAF show - the majority, maybe as much as 80%, of mainstream high-end exhibitors had aggressively presence-peaked loudspeakers that just sounded very unmusical. I am at a loss why anyone would want a harsh, metallic-sounding speaker, particularly if it cost many thousands of dollars.
I was honestly wondering if this was just me, but Bud confirmed what I noticed, and now the Harbeth team say the same thing. It's just the current fashion. Has the average high-end customer forgotten what reality sounds like?
Quote by Lynn Olson:
"noticed the same thing as I did at the last RMAF show - the majority, maybe as much as 80%, of mainstream high-end exhibitors had aggressively presence-peaked loudspeakers that just sounded very unmusical. I am at a loss why anyone would want a harsh, metallic-sounding speaker, particularly if it cost many thousands of dollars."
I agree with your observation about the aggressively presense-peaked loudspeakers.
RMAF 2007 was my first attendance to a large scale audio show where I was exposed to so many different systems.
I was actually dumbfounded by this common peaked sound characteristic of the majority of the systems exhibited, jaded in fact.
I also noticed a huge disconnect from speakers that used large diameter woofers with horns and small format compression drivers that were crossed over above 1kHz, especially in the upper female vocal region. Some of these speakers sounded smeared and shrill in that area, with a few beaming like a lazer.
I hope you will be able to get past these problems with your proposed lower crossover point and limited frequency response using the larger 1.4" compression driver, LeCleach horns, above large diameter woofers for the midrange in your final speaker design.
Your comment surprised me earlier in this thread where you felt that the Audio Kinesis and Feastrex systems were some of the best sounds of the show, as those were my choices as well.
As they were at LSAF in Dallas this past May.
Norris Wilson
"noticed the same thing as I did at the last RMAF show - the majority, maybe as much as 80%, of mainstream high-end exhibitors had aggressively presence-peaked loudspeakers that just sounded very unmusical. I am at a loss why anyone would want a harsh, metallic-sounding speaker, particularly if it cost many thousands of dollars."
I agree with your observation about the aggressively presense-peaked loudspeakers.
RMAF 2007 was my first attendance to a large scale audio show where I was exposed to so many different systems.
I was actually dumbfounded by this common peaked sound characteristic of the majority of the systems exhibited, jaded in fact.
I also noticed a huge disconnect from speakers that used large diameter woofers with horns and small format compression drivers that were crossed over above 1kHz, especially in the upper female vocal region. Some of these speakers sounded smeared and shrill in that area, with a few beaming like a lazer.
I hope you will be able to get past these problems with your proposed lower crossover point and limited frequency response using the larger 1.4" compression driver, LeCleach horns, above large diameter woofers for the midrange in your final speaker design.
Your comment surprised me earlier in this thread where you felt that the Audio Kinesis and Feastrex systems were some of the best sounds of the show, as those were my choices as well.
As they were at LSAF in Dallas this past May.
Norris Wilson
Thanks Lynn, I'm not too much into the digital thang but over the years have found a player that sounds good to me. I have one of those Heart Audio Marantz 6000 and it is pretty good at not sounding artificial. I believe it shares a lot of what you have explained below.
To me it still has faults with many discs, I think it's the process used by the producers of the discs more than any thing else because I have many discs that sound just lovely.
Heart Audio digital
To me it still has faults with many discs, I think it's the process used by the producers of the discs more than any thing else because I have many discs that sound just lovely.
Heart Audio digital
Lynn Olson said:
Norris Wilson said:
I am well aware of the problems of a 15" woofer and a small-format compression driver - that's an Altec A7, one of my least-favorite loudspeakers. Your description of what you heard is exactly what an A7, or its modern derivatives, sound like to me. Shrill, harsh, completely unacceptable female vocals, and thanks to the harsh midrange, no rendition of spatial qualities whatsoever.
I had written off all horns and compression drivers until I heard Altec A5's at a meeting of the Bay Area Tube Enthusiasts. In fact, I had to repair the crossovers, since Gary Pimm and I refused to demo our amplifiers until the speakers were repaired - and guess what, when you want to do a job right, you do it yourself.
The crossover housings were beautiful cast aluminum with that entrancing "Altec" logo on the outside, but what was inside was sheer garbage. Badly-soldered Mylar caps, extremely corroded L-pads, and cheapo inductors, too. I can see why Altec wanted to hide what they were selling. We got rid of the corroded and unusable parts, replaced them with at least half-decent polypropylenes, and lit up the speakers with minutes to spare before the conference got started. At least the A5's were halfway listenable after that, although nowhere close to flat - the stock crossover topology was the typical basic mid-Fifties textbook 12 dB/octave network, no compensation or EQ at all. One of the "mods" I made, if you could call it that, was a 20-ohm resistor in parallel with the HF driver, to damp the extreme impedance excursions of horn and compression driver.
The HF was a lot better than I expected, and nothing at all like the crudity and harshness of the A7. Sweet, not very extended, but quite effortless in filling a 200-seat auditorium. Surprisingly, I didn't hear any metallic "ting-ting" at all, and I was listening for it. And the big multicells (probably 1505's) sounded really very good - the best horns I'd heard to date. The upper midrange (800 Hz to 3 kHz) was a lot better than I expected.
So I had to take the Altec 288 and 1505 off my "bad-bad" list. A little bit of EQ - not really that much - and they could keep up with a lot of midranges and tweeters. I later found out there were very serious problems with the small-format Altec 511 and 811 sectoral horns - the "pinch" in the throat region is a massive diffraction generator, and even back in the Fifties, those in the know were aware that multicells were superior to sectorals, which were essentially an exercise in cost-cutting.
So that led me to re-think horns and compression drivers. What I didn't like was the sound of overstressed small-format drivers and the diffraction and internal reflections of poorly-designed mid-Fifties horns. Hearing the Dr. Geddes Summa and the local Denver-area Le Cleac'h horns confirmed that impression - if the driver had good (low-distortion) midrange reproduction and the horn profile was designed to minimize diffraction, then the bad old Altec (A7), JBL (L200, L300) and Klipschorn harshness was no longer there.
Which is why you see me leaning towards the Altec 288 and the modern version, the Great Plains Audio 399 with 16-ohm aluminum diaphragm, along with a LeCleac'h horn. Will it sound like a beryllium TAD or JBL? Probably not. But then again, I'm not sure I want that sound. I've heard these drivers in Japanese-market high-efficiency speakers, and I am not convinced that beryllium is all that special - measures good, yes, but more natural, more lifelike than aluminum or phenolic? Not sure about that.
I do trust my senses, and people who know me know that I assess a loudspeaker very quickly, in a matter of minutes, and my impressions don't really change that much after several hours or days of listening. I like a speaker or I don't, or more commonly, I may hear some things that are interesting in an otherwise flawed loudspeaker. The Hill Plasmatronic was the most dramatic example - the best treble I've ever heard, but the rest of the speaker was very mediocre, and did not harmonize with ionic tweeter at all.
Gary Pimm and I would joke about how fast we'd hear things - we'd make a circuit change, or twiddle with the loudspeaker, and most of the time, if the change was anything significant, we'd hear it before we sat down in the listening seat. In fact, that became one of our subjective yardsticks - a change that was really worthwhile would be audible while you were in transit, while the small-change stuff took longer, maybe listening to a whole side of the record, or playing through the whole CD. But the stuff that was important was something you'd hear right away - even if you were in another room of the house.
If you hear something unexpected (which happens fairly often), it's always interesting to find out why it sounds the way it does, especially if it violates existing preconceptions. In amplifiers, for example, surprisingly small circuit changes can be immediately audible, and just to double-check, immediately audible to non-audiophile listeners. In a way, my non-audiophile friends are my reality check - they hear more things than I do, and have trained me to listen for qualities like excitement, a sensation of you-are-thereness, and just the simple reality of "does it sound like music?"
Sometimes I wonder how such people as the Mercury team could produce recordings of enduring quality on monitors of their era.
I guess that they knew so well how it would go to the disks because they could replay it in their heads and subtract the equipment colorations mentally. The huge involvement with live music in halls must have been their compass.
I guess that they knew so well how it would go to the disks because they could replay it in their heads and subtract the equipment colorations mentally. The huge involvement with live music in halls must have been their compass.
Attachments
In case the LeCleach horn becomes the choice, here are various places to learn how to make one yourself- including how to make one out of petal segments
http://www.diyaudio.com/forums/showthread.php?postid=1528948#post1528948
Posted by the designer himself, in the Horns vs Waveguides thread..
http://www.diyaudio.com/forums/showthread.php?postid=1528948#post1528948
Posted by the designer himself, in the Horns vs Waveguides thread..
four Pi loudspeaker response curve
Regarding the "problems of a 15" woofer and small format compression driver", certainly both are being required to cover a fairly wide bandwidth. The midwoofer is past its pistonic range, much like a single driver speaker. So cone flex resonants must be well damped. Same is true for the compression driver. Intermodulation is higher when bandwidth is higher, again, like a single driver speaker. However, it is a simple speaker with only one crossover, and it can be placed in a range where crossover is least objectionable, where summing is coherent and where directivity matched. In my opinion, this is a very attractive philosophy. Not the only good way to make a speaker, but certainly one of the better ones, in my opinion.
This the response curve of a four Pi loudspeaker, which is a DI-matched speaker using a 15" woofer and 1" exit compression horn tweeter. Crossover is asymmetrical and is above 1kHz. Measurement was made outdoors, ground plane, using a swept sine. No smoothing or post processing was applied.
Regarding the "problems of a 15" woofer and small format compression driver", certainly both are being required to cover a fairly wide bandwidth. The midwoofer is past its pistonic range, much like a single driver speaker. So cone flex resonants must be well damped. Same is true for the compression driver. Intermodulation is higher when bandwidth is higher, again, like a single driver speaker. However, it is a simple speaker with only one crossover, and it can be placed in a range where crossover is least objectionable, where summing is coherent and where directivity matched. In my opinion, this is a very attractive philosophy. Not the only good way to make a speaker, but certainly one of the better ones, in my opinion.
This the response curve of a four Pi loudspeaker, which is a DI-matched speaker using a 15" woofer and 1" exit compression horn tweeter. Crossover is asymmetrical and is above 1kHz. Measurement was made outdoors, ground plane, using a swept sine. No smoothing or post processing was applied.
agent.5 said:
Very poor. My feelings about the big Onken are the same - no thanks. Big, resonant, and floppy is not my thing. I don't care for "vintage" bass at all.
The Altec Alnico drivers, though, are another story. I've played around enough with test baffles to know that wonderful drivers can lurk in the worst cabinets, and that some pretty funky drivers can also have their deficits concealed by cabinets. In my experience, which I know is not the same as others, direct-radiator drivers are rather well-behaved in the 300 to 1 kHz region (aside from IM distortion), while cabinets are at their worst in the same region.
And unfortunately just raising the mass of the bass cabinets to ridiculous levels (like the 200~300 lbs we now typically see in audiophile speakers) just results in slow, soggy sound thanks to the enormous amounts of energy storage and very long discharge times. Although the wall vibrations may be small, what offsets that are the radiating areas, which are many times larger than the cone area. When a cabinet wall goes into resonance, it becomes acoustically transparent at that narrow frequency, and energy flows from the inside of the cabinet to the outside - and the high Q of the resonance stores the energy for a long time after the initial excitation has ended.
I am more and more convinced that well-designed bass-horns and transmission lines share a common feature - greater rigidity than the typical big bass-reflex box, and this is immediately audible as greater clarity and less "loudspeaker" coloration for material like piano recordings, which tend to excite cabinet resonances (a piano has a big resonant wood box too, but it's one we want to hear). The greatest merit of open baffles may be nothing more than small emissive areas compared to conventional boxes - and that all this talk of radiation patterns has been distracting us from the main attraction, which is reduction of energy storage.
Just got a very nice e-mail from my friend Allen Wright, who describes David Haigner's speakers as the least "hornlike" sounding system he's heard to date, but with wonderful dynamics. The Alphahorn, Betahorn, and Gammahorn all look interesting. There are PDF measurements of a Titanium driver and a Phenolic driver. The elliptical horn shapes are intriguing to say the least.
Allen mentioned that Haigner designs opera houses (!) - not something you see every day here in the USA, where - how shall I say this - other types of music dominate the cultural scene. One of the things I liked best about the European musical scene was the freaky combination of classical, opera, and techno. Works for me.
Well, in all truth, I think all compression drivers are getting a little shaky above 12 ~ 15 kHz. When you remove the smoothing, there are a lot of ups-n-downs, and the impulse response is not a thing of beauty. Entirely aside from large-diaphragm breakup, the wavefronts coming out of the phase plug are most certainly not well-behaved at those frequencies, which then violates all of the design assumptions of the horn or waveguide.
Horns do not have a magical ability to "straighten-out" a ragged wavefront. No device can, although I guess you could scramble it even more and hope for the best.
Horns do not have a magical ability to "straighten-out" a ragged wavefront. No device can, although I guess you could scramble it even more and hope for the best.
Lynn Olson said:
Thats about the 1.4 inch compression drivers quality. Very acceptable. But I just think that those horns in the links cant be constant directivity and having FR up to 20k or beyond at 103dB sensitivity as quoted by their manufacturer. They must be narrowing up high so to keep SPL up.