Yeah, agreed. Do you find it as interesting as I do that even though DSP 'came of age' in audio over 20 years ago (with the DSP56k) that Meridian is the only high-end company that manufactures digitally XO'd speakers?
I've worked as designer on such a system - I kept all the high dynamic range analog aspects outside the box
Sure do, but I don't think it's because of any problem with making digital filters sound good (if that's what you were implying?). Besides, there have been others.
NHT made a superb system (can't remember model) consisting of mini-monitors with 5" magnesium-coned bass-mid drivers and two seperate subwoofer enclosures with twin 10" drivers mounted opposed to each other to cancel vibrations. The digital crossover/eq was supplied by the company I worked for and they were using >100dB/8ve linear-phase crossover filters. The designer told me the mid-bass drivers were exceptional as they had no break-up below crossover but would be quite unusable with analogue filters as the out of band resonance due to metal cones was so severe they required the very steep filters to attenuate it below audibility. The measured frequency response, phase and group delay of the system were far superior to any analogue only speaker as these were fully corrected by the filters. It really sounded incredible. I believe NHT folded a while back though. Probably didn't get enough favourable reviews in the press
NHT made a superb system (can't remember model) consisting of mini-monitors with 5" magnesium-coned bass-mid drivers and two seperate subwoofer enclosures with twin 10" drivers mounted opposed to each other to cancel vibrations. The digital crossover/eq was supplied by the company I worked for and they were using >100dB/8ve linear-phase crossover filters. The designer told me the mid-bass drivers were exceptional as they had no break-up below crossover but would be quite unusable with analogue filters as the out of band resonance due to metal cones was so severe they required the very steep filters to attenuate it below audibility. The measured frequency response, phase and group delay of the system were far superior to any analogue only speaker as these were fully corrected by the filters. It really sounded incredible. I believe NHT folded a while back though. Probably didn't get enough favourable reviews in the press
This is going way OT, so perhaps the mods can create a fresh thread on the topic of digitally XO'd speakers with these posts in them?
This is a classic example of how things start to go wrong when digital is introduced into the mix Why the choice of linear phase? Did the designer not realise that the pre-ringing associated with a linear phase impulse response would be quite audible? On a full-range digital system that ringing is ultrasonic, on a digital XO its well within the audio band - like around 3kHz for a typical 2-way setup.
Yes, but the nasty resonances can be set off not just by the amplifier output, acoustic coupling from adjacent drivers can (and does) do it. The designer of the B&W Nautilus made one of that speaker's mid-range drivers a flat surface for precisely this reason - the original prototype had used a dome. Admittedly he was only using a 4th order analog XO but the problem wasn't electrical out of band stimulation of resonances.
Could it not have something to do with sucky digital implementation?
This is a classic example of how things start to go wrong when digital is introduced into the mix
Why the choice of linear phase? Did the designer not realise that the pre-ringing associated with a linear phase impulse response would be quite audible? On a full-range digital system that ringing is ultrasonic, on a digital XO its well within the audio band - like around 3kHz for a typical 2-way setup.Yes, but the nasty resonances can be set off not just by the amplifier output, acoustic coupling from adjacent drivers can (and does) do it. The designer of the B&W Nautilus made one of that speaker's mid-range drivers a flat surface for precisely this reason - the original prototype had used a dome. Admittedly he was only using a 4th order analog XO but the problem wasn't electrical out of band stimulation of resonances.
The measured frequency response, phase and group delay of the system were far superior to any analogue only speaker as these were fully corrected by the filters. It really sounded incredible. I believe NHT folded a while back though. Probably didn't get enough favourable reviews in the press
Could it not have something to do with sucky digital implementation?
To quote myself for reference:
"It also helps to bias the op amp output stages into class A (nice bootstrapping trick from you btw jcx!) since this also results in very clean signals on the supply rails."
Owdeo, when driving heavy loads, opamp output stages will be operating in class AB. The load current is reflected in the opamp supply rails as a halfwave rectified version of the output signal. Clearly this will have a lot of harmonics, and if fed from a 'stiff' supply will result in radiated and common impedance coupled grunge on the supply rails. If you bias the opamp output stage up into class A, you theoretically only get DC on the supply rail. Its also important by the way to make sure you terminate the load return back at the local circuit ground point and then have that ground point have a single path pack to the system signal ground. This way, you avoid common impedance coupling noise on the signal ground - a place where just as many problems await the unwary as in the supply rails.
It goes without saying that the decoupling ground and the signal ground are not one and the same - they need to be separate.
WRT the size of the R's in the RC supply filters. This will in many cases be a matter of preference and consideration of the probable loads that will be driven by the opamps. I think 47 Ohms is high if you are driving a 600 Ohm load for example, but if you were driving 2 or 3K, it would probably be acceptable, provided the C is large enough. As mentioned earlier, I have settled on 22 Ohms and 100uF which seems to serve me well on LM4562's and NE5534/32 devices. The short correct answer: experiment!
I can feel another 'article' coming up on my website about noise and decoupling!
"It also helps to bias the op amp output stages into class A (nice bootstrapping trick from you btw jcx!) since this also results in very clean signals on the supply rails."
Owdeo, when driving heavy loads, opamp output stages will be operating in class AB. The load current is reflected in the opamp supply rails as a halfwave rectified version of the output signal. Clearly this will have a lot of harmonics, and if fed from a 'stiff' supply will result in radiated and common impedance coupled grunge on the supply rails. If you bias the opamp output stage up into class A, you theoretically only get DC on the supply rail. Its also important by the way to make sure you terminate the load return back at the local circuit ground point and then have that ground point have a single path pack to the system signal ground. This way, you avoid common impedance coupling noise on the signal ground - a place where just as many problems await the unwary as in the supply rails.
It goes without saying that the decoupling ground and the signal ground are not one and the same - they need to be separate.
WRT the size of the R's in the RC supply filters. This will in many cases be a matter of preference and consideration of the probable loads that will be driven by the opamps. I think 47 Ohms is high if you are driving a 600 Ohm load for example, but if you were driving 2 or 3K, it would probably be acceptable, provided the C is large enough. As mentioned earlier, I have settled on 22 Ohms and 100uF which seems to serve me well on LM4562's and NE5534/32 devices. The short correct answer: experiment!
I can feel another 'article' coming up on my website about noise and decoupling!
Last edited:
Abraxalito, you have just written off a product that you presumably haven't seen, heard nor know anything about other than what I've written and yet have already offered various arguments for why it doesn't sound any good and therefore did not sell well and also implied that you know better than the designers did. By all means start a new thread but I'll pass on reading it thanks.
Perhaps there were things wrong with it, but it measured superbly and sounded fantastic to all that heard it that I know of. I also can't agree re pre-ringing being audible. I experimented alongs those lines and concluded that it isn't, but let's leave it at that.
I wonder how much that same kind of thinking influences the "gods of hifi review" when they pass their subjective judgement on a product full of "nasty digital" tech... unless it comes from one of the established big names of course in which case it will be suberb, once it has "burned in"...
Perhaps there were things wrong with it, but it measured superbly and sounded fantastic to all that heard it that I know of. I also can't agree re pre-ringing being audible. I experimented alongs those lines and concluded that it isn't, but let's leave it at that.
I wonder how much that same kind of thinking influences the "gods of hifi review" when they pass their subjective judgement on a product full of "nasty digital" tech... unless it comes from one of the established big names of course in which case it will be suberb, once it has "burned in"...
Gotcha, sorry I misread your post and thought you were saying that RC filtering helps to bias the opamp into class A
.That makes perfect sense now.WRT the size of the R's in the RC supply filters. This will in many cases be a matter of preference and consideration of the probable loads that will be driven by the opamps. I think 47 Ohms is high if you are driving a 600 Ohm load for example, but if you were driving 2 or 3K, it would probably be acceptable, provided the C is large enough. As mentioned earlier, I have settled on 22 Ohms and 100uF which seems to serve me well on LM4562's and NE5534/32 devices. The short correct answer: experiment!
And now for the million dollar question: experiment how? What measurements did you use to determine that those values worked well? How were they different to if you hadn't used any RC filtering? Unless someone can answer this credibly the whole thing seems to be another article of faith to me...
Nowhere have I 'written it off' - but from what you have told us right here the designer made some highly significant errors in its design. Those errors were entirely avoidable ones. Bruno Putzeys has written about the mistake of using linear phase filtering and explains how to get around that as well as some more detailed potential problems with equiripple filters.
I'm not contradicting any of that. Its an example that we still don't know well enough what to measure.
Suggest you read Bruno's writings.
I wonder how much that same kind of thinking influences the "gods of hifi review" when they pass their subjective judgement on a product full of "nasty digital" tech... unless it comes from one of the established big names of course in which case it will be suberb, once it has "burned in"...
By 'that same kind of thinking' you're referring to your own cynical thinking behind the remarks on my words right? Like taking my pointing out of very real technical drawbacks in a design approach and calling them 'writing off' ? Or implying that I say it 'doesn't sound any good' when I merely claimed (much more conservatively) that pre-ringing was audible?
Sounds interesting Alex, and I imagine it could well be. But I fail to see how this relates to pre-ringing being audible per se. Was that the premise of the test and if so on what basis?
Cheers,
owdeo
Alex that sounds to be John Kenny's test over on WBF - which was for pre-echo, not pre-ringing. Different beasties
Alex that sounds to be John Kenny's test over on WBF - which was for pre-echo, not pre-ringing. Different beasties
O.K. I will go quietly , I am out of my depth here.
You 2 can slug this one out.
Anyway, some preferred the modified version. I didn't, and ordered a copy of the CD. If I had only heard the modified version, I wouldn't have.
There were no errors in the design, it was a deliberate choice. Unless there is broad concensus industry wide as a result of extensive scientifically conducted tests that pre-ringing is audible per se in any application (in which case just about all currently available DAC ICs are evil) these are only your (and presuably Bruno Putzey's) opinions. I really think you are just proving my point further - please let's leave it there.
Those guys can create pre-echo for themselves then - just takes a couple of operations with Audacity to add it in
My blog gives details of a similar operation in the article 'Correcting NOS droop with Audacity'.
You misunderstood my point about pre-ringing as you've made it too general. DAC pre-ringing is ultrasonic (>20kHz even with Redbook CD), digital XO pre-ringing (being at the much lower bass/mid cutoff frequency) isn't.
Regarding the op amp's choice by D.Self for his new preamp.I can confirm that my experience shows that the op amp specified based on price alone is justified. I have used both the expensive Burr Brown version on a preamp and the NE 5532 and i can honestly say that whatever sonic difference there was, could not be heard.
While that is certainly true, "iron" also has its share of problems that limit its use to special cases (RF frontends, MC pre). In a preamp's gain stage you're mostly after maximum dynamic range, and transformers doing well in that regard - towards the low end of the audible range in particular - tend to be unreasonably expensive and clunky.
SNR is fundamentally a matter of signal power vs. noise power, so for maximum dynamic range you need to maximise signal power transfer. That means you can go down in impedance until either your active devices or your passives get unhappy (not to mention increased power consumption). The same dynamic range can be achieved using either high voltages and high impedance or lower voltages and low impedance (V ~ |Z|²) - what's most practical depends on the amplifier technology at hand. If the world still were all hollow state, we'd certainly be seeing low-noise preamps with +/-150 V supplies, 100k pots and internals levels about 10 times as high compared to what we're used to.
BTW, has anyone ever gotten a common carbon log pot to measurably "break into sweat" (generate distortion) within its rated power? I would expect a lin pot to heat up fairly uniformly, so any variation would cancel out in a voltage divider, but what about log pots and their variable track width?
Last edited:
"And now for the million dollar question: experiment how? What measurements did you use to determine that those values worked well? How were they different to if you hadn't used any RC filtering? Unless someone can answer this credibly the whole thing seems to be another article of faith to me..."
No articles of faith. The coupling and layout techniques I listed work as advertised. WRT the experiment bit, get a scope, drive a load and look at the supply rails. Then make a selection of whatever value your engineering judgemnt tells you is best. You can take a listen as well BTW
No articles of faith. The coupling and layout techniques I listed work as advertised. WRT the experiment bit, get a scope, drive a load and look at the supply rails. Then make a selection of whatever value your engineering judgemnt tells you is best. You can take a listen as well BTW
Not in audio, it is not a match terminated system. A 10 Ohm MC has .4 nV of noise, a JFET preamp could be concocted with .2nV voltage noise and virtually no current noise. No power is delivered to the JFET since the input impedance is >>10 Ohms.
Which begs another question, the synthesized 47K still looks like 47K to the current noise of the op-amp.
On a sarcastic reply Nico ...Just a sarcastic note.... why hasn't anyone replaced Doug's op amp choice with the highly aclaimed Burson op amp as seen in some threads on the forum. Obviously this would solve all the shortcomings in the design.
Nice web-site Bonsai
At $180USD per pair as shown on the Burson website, that prices out at $1,620USD if you replace all 18 op-amps on Doug's pre-amp PCB! 