The best amp and speaker setup for digital playback

[PeterSt]

Reading this from of Per's post, it all seems so easy. But I am afraid it is not;

Speed, speed, speed, yes, but still the best measuring OpAmp doesn't sound (please remember, - to me).
An electrostat may have great speed, but maybe not enough power.
If a high frequency unit works at the proper level, that doesn't tell the bass driver does as well. Let alone what happens at the xover point(s).
When a DAC measures with good THD+N figures maybe it can't drive a main amp (so there go your figures).
When your speakers measure good they may be of less efficiency, and the good measuring amp to drive them may be unaffordable.

And don't think I know it all. But, I try to listen carefully and I always try to reason out why something sounds as it sounds.


If we look at the 1:1 concept of things - meaning that once all is passed with the least distortion possible, currently there is one element in this 1:1 chain that doesn't fit the bill : the analogue stage in the DAC with the OpAmp. I just can't think of why this so good measuring "thing" (disputed by many) won't sound. So, in order to make you crazy once again, this is my experience with it :

In our systems, we often judge the merits of it by cymbals. This may not be for all, because cymbals must first sound good enough to "do" with them what I am about to tell. Otherwise they are merely distorting, and they had rather not be there.
And in between the lines, when you are working on "good sound" at the level I do (meaning with the possible influence, software, hardware), it is a big big challenge to get cymbals right. They can sound too plastic, too light, all the same (no size audible), to rough (all sounding like Chine cymbals) and a few more things. One of these things is quite unknown I think, but it really exsists to my own experience :

The length in time a cymbal sounds.

And this is where my stupid OpAmp solution excelled.
Now try to imagine ... on a random track there's a random hit on a random cymbal, and it takes 5 seconds to die out. Your system sounds wonderful to begin with.
Now I use that good measuring OpAmp in the analogue stage of the DAC, and suddenly the cymbal is audible for 20 seconds. Yes, a completely crazy difference.

In the mean time, this OpAmp doesn't "sound". But why ? this goes against my own theories just following (my) practice; if it measures good it must sound good.

Now, the remainder for now is FYI only, but indicates a bit how I think :


As I said in the other topic the other day : at feeding the DAC with 352.8KHz things suddenly didn't work anymore (while 176.4 did). I seemed to overdrive something. I thought I could hear it was caused by capacitors not performing anymore (mind you, in the digital section of the DAC, and nowhere near of being in the signal path). And I turned out to be right. Fine.

Now, one year after my attempts with the OpAmp, this weekend I will try that again. Hopefully by now I will be able to hear what is wrong with it, or maybe it isn't wrong anymore in the first place. I don't know yet. One thing I do know : they *should* sound right because they measure right. So here's the 1:1 theory and its impact :

If I stuff in the chain something which measures right (think of 0.00001% THD+N) and *that* causes things to sound wrongly, it will have emphasized something which wasn't emphasized otherwise (notice the OpAmp in this case is about gain). This can - IMO - be fed back to something being in lack of drive otherwise (without the OpA). The latter sounds logic to me, because of the crazy long sounding cymbals, which don't sound crazy long without this OpA solution. And the worst of all is : cymbals DO sound long. Try it !

If this isn't about something that gets emphasized, it should be about a feedback of some kind of the device itself. Back into the transformer, while that transformer has other SQ related stuff connected.

But there is a way more creapy possible reason :
Back at that time (a year ago) I was using or no filter at all (this smears to name one nasty habbit of that), or I was using an SRC (just plain wrong), or I was using the common filtering (if AA (now called AI) filtering was introduced around a year ago, it will probably have been that). In either case, all of these distortion types smear and also react to eachother (so the smear smears). And, besides the long sounding cymbals, if I recall one property of that OpA experiment (against all odds as I thought at the time) it was sibilance in woman voices.
And did I mention the higher jitter clock I was using at the time ? another source for smear.

I can add to this, that all what can be wrong with the above mentioned subjects, will be about high frequency products ...

And NOW I am going to add that device which can drive and has sufficient power and SPEED to ... emphasize just that.

By now I seem to have already convinced myself that the OpAmp today will be working. But many things have changed, with as the most important (for this subject) the Arc Prediction Upsampling and probably a low jitter clock.


Well, you see ? this is how to build "a chain". This is quite contrary to removing a device which doesn't sound good (and mind you, devices go up to resistors !). You'll have to *understand* why it doesn't sound good, or otherwise you may be left behind with a not optimal system.

It all takes quite some learning and experience, plus the not-wrong attitude. I mean, it is maybe two months back only that I said that OpAmps unconditionally don't sound good, no matter how good they measure. But it is completely against my own thinking and experience (except for this one then), and it is difficult to sleep on it.
I am not sure how this one will turn out, but supposed I get it to work now (good sounding), there is a definite rule to remember for the future for everybody :

When an element is added to the chain, and which element measures at least better than the remainder of the chain while the sound gets worse from it, don't remove that element, but keep on looking for something else which cannot cope.
"Cope" is in the line of speed, but with digital all just *is* about speed (think possible transients, which no vinyl can follow).

This should be sufficient for the weekend !
Peter

[Telstar]

And now to my question to Peter and you guys / girls (?)

Would very fast amplifiers that can swing great voltage with ease (like huge Class A power amps) and planar speakers with super light mylar speaker membrane be the ideal setup?

This is my recipe. Horns and chipamps can be another one. Etc.

[manisandher]

I'm currently driving a pair of Quad 2805s (planars) with a Pass Labs Aleph 4 (single-ended, pure class A, 100wpc... and lots of heat!) and getting a really nice sound. In fact, the most 'coherent' sound I've ever achieved from speakers.

I think this is one (perhaps of many) recipes that actually works... thought not if you're into raves.

If only the Quads were more efficient, I'd be driving them with my Berning Siegfried.

Mani.

[manisandher]

Back at that time (a year ago)...

Wow, it's amazing how much your thinking has evolved in a year!

When an element is added to the chain, and which element measures at least better than the remainder of the chain while the sound gets worse from it, don't remove that element, but keep on looking for something else which cannot cope.

When you refer to 'measurement', what else are you measuring beyond just THD?

E.g., how important is slew rate in amps? Here's what Nelson Pass thinks (taken from the Aleph 4 manual):

"The slew rate of the amplifier is about 30 Volts/uS under load , which is about 30 times faster than the fastest signal you will ever see, and about 100 times faster than what you will be listening to."

And yet, if you look at the specs for something like the Spectral DMA-360, the slew rate is quoted as "600 volts/microsecond"!

Thoughts?

Mani.

[Telstar]

E.g., how important is slew rate in amps? Here's what Nelson Pass thinks (taken from the Aleph 4 manual):

"The slew rate of the amplifier is about 30 Volts/uS under load , which is about 30 times faster than the fastest signal you will ever see, and about 100 times faster than what you will be listening to."
And yet, if you look at the specs for something like the Spectral DMA-360, the slew rate is quoted as "600 volts/microsecond"!

Thoughts?

It depends first of all on the fastest signal that will pass through the amplifier. with the NOS1 that's 352khz.
Then on the maximum power at which that signal will be amplified.

There are some simple formulas. in short the 30V/us are good only for low powered amps (<40W).
The classic formula considers 20khz as the maximum frequency containing music signal, which is not the case with upsampled and unfiltered source.

Unfortunately the slew rater measure only vertical speed. I haven't seen any (way to) measure the horizontal speed. But it's the only bit of information that we have that gets close to the definition of speed.
The open loop bandwith is another one, but almost impossible to know unless you can simulate or measure the circuit.

PS: high-efficiency planar(or ribbon) extended midrange is the goal for my next speakers.

[PeterSt]

I haven't seen any (way to) measure the horizontal speed.

This isn't necessary either, because it is a state (like continues 6V). BUT :

There's always ringing, and here is a key which influences the projected vertical speed again, *as* it does with that "6V state". Thus :

Before any transient can start (but think of a real steep one), preringing is needed. This is physics. But, it depends on the power how long that will be (the more power the less). The same happens when the 6V is reached. Now we have post ringing;
What I have seen from my measurement post ringing can be virtually endless. Maybe in electronics some kind of feedback exists to get the state as quickly as possible stable, but the feedback by itself will have to be very smart (and super fast) or otherwise it creates the "ringing" by itself.

We see the above when looking at squares, but of course these kind of things happen in real life music all the time; You may recall my measurents actually showing these things. This one is always the best example : Re: Measuring XXHighEnd ... which as far as I am concerned is one pulse with an after swing. Notice this is just music data and no test signal.
I happens all the time, and is not good for the best sound.

Of course, when we'd have these things under perfect control, sound may start to sound very analytical. But on the other hand it should sound as intended.

The foremost thiong for me remains : when this would be output from an OS DAC (or with normal filter anyway) such pictures would be a complete mess with no reference at all (I tried it). This doesn't imply you won't here the gueste of what the artist intended. But it can't be more than that. You can just see it ...

Peter

[bhobba]

Well I don't know if its the best but after a lot of careful consideration I have chosen my amp, louspeakers, and DAC

Loudspeakers - Aether Audio Timepiece Mini:
http://www.aetheraudio.com/
http://aetheraudio.com/timepiece_mini.html

And Bob (that the hippy looking guy in the picture - great bloke IMHO) has a new development that allows a crossover at 527hz.  Both the frequency and phase response are virtually ruler flat and talk about power handling and speed - well their speed rivals electrostatics while their power handling is simply unbeatable - they handle 1K monster amplifiers with ease - no distortion whatsoever.  Now I won't be using them with anything like that but it is good to know they can handle it.  Don't take my word for how good they are.  See what the great audio engineer Jim Merod, who has heard everything and anything in his time has to say:
http://www.positive-feedback.com/Issue30/sptech_2.htm


Basically as one person who heard them said - they are shockingly naked.

For amp duties I have chosen the latest design by Hugh Dean the Maya.  Its very new and he hasn't got it up on his site yet but here is what he does have up:
http://www.aksaonline.com/products/products.html

The Maya is supposed to be a touch better than the Soraya and that is some feat IMHO.  Check out: 
http://www.diyaudio.com/forums/solid-state/108756-soraya-cb105.html

As destroyerx wote (what a name): This Soraya is not more, not less, than the best.  That must be some statement because he has had '49 years buiding audio amplifiers, more than 4 thousand units made and tested'.

For the DAC I have chosen the new DAC2 from Wyred For Sound:
http://www.wyred4sound.com/webapps/site/74030/117839/shopping/shopping-view.html?pid=457975

I suspect some dacs like the overdrive and new tranquility dac may have a slight edge on it in sonic terms (as judged by some) however I chose it because it uses the Saber 32 bit dac which headphone guys are really impressed with.  They use words like - removes a layer of grit from CD's.  They have heard nothing to beat it and that is good enough for me.  Both the overdrive and tranquility are nos dacs.  I have not had good experiences with nos dacs.  All the dac's I have heard that I like were oversampling.  Others however have commented they were a bit too clinical and detailed - they liked the more laid back sound of nos DAC's.  I am the exact opposite - to me clinical and detailed just sounds REAL.  I may however upgrade later.  Oh and the tranquility DAC is only 16 bits so a digital volume control which I also like is out.

Thanks
Bill

[PeterSt]

Hi Bill, nice story and thanks for sharing.

Maybe it's allowed to tell that this topic is supposed to be about theoretical best stuff and less about our subjectiveness of hearing ? Thus, this isn't about a bunch of head-fi people noticing layers disappear after using OS DACs all their life with layers to begin with (in the middle of an environment where you're shot when you dare to talk "NOS" ). It also isn't related to what I find sounding best. Nor anywone else.

It is about what *IS* the best. This is an absolute thing and not subject to what we (want to) perceive from it. But :

Even when listening would be involved, it is my firm opinion that this is not subjective IF you have the reference first. And the fun is, today's reference allows us easily to listen to the instruments how they are, or at least I can do that here (with my Phasure NOS1). The world of audio really changes from that. What I mean is this (and it has been told before) :
Without the reference, we will be used to let our brains speak for whether we find something good or not. "Good" means acceptable. Telstar might jump in to confirm this, at a comparison with an ESS implementation (24 bit) we did a year back or so. Thus, without the reference the owner of the ESS (which was not Telstar) was completely satisfied. He had no other reference. Telstar was not satisfied because one "hears" better than the other. Also, at that time Telstar didn't care much about OS vs NOS. I could hear it was wrong all over, because I did have a reference (for the better, which was my own DAC). Now :
As soon as we hopped over to my DAC, the ESS seem to fail all over. It really needed 2 seconds of listening (and I mean this) to judge the ESS as broken.

I can tell you, at that time my DAC was  just "NOS", although technically very good. But nothing about the needed filtering which is in there today (and which in the end is Arc Prediction from XXHighEnd). So, my reference already changed again, and something like the ESS would be the worst (I am just using superlatives because it is so).

Keep in mind, the last story was about listening not being subjective at all, because an instrument is not subjective (if you only heard it live once) while todays capabilities of audio allow us to listen to that instrument. So, it is the instrument (good playback) or it isn't the instrument at all (bad playback).

Once we are in the stage of listening to real instruments, we are still far from being there, because there's also the foot-tapping or tear drawing mechanisms (analytical vs emotion). They are hard to control, but *are* a phenomenon which can be explicitly tested. And the worse is, emotion can work just the same with a -30dB noise audio cassette tape. Things like noise even seems to be a prerequisite sometimes (vinyl). So, it is about bringing this all together, and luckily the Engine#4 version of XXHighEnd contributed a lot to the emotion part.

Back to the real thing : To my findings - and I am really explicitly working on that for quite some years now - all works out the best if digital is treated how digital should be treated. This is not massive oversampling and (today) it is very easy for me to proove where it destroys. Mind you, I could always do that, but before this was without alternative. So, before this was about NOS being wrong as well, but in such a completely different way that it came down to a subjective choice what was better. NOS with heavy harmonic distortion (and I mean HEAVY) or OS with EVERYTHING wrong, but unmeasureable (because the used filtering is theory only and doesn't work in practice). But this changed since decent filtering can be applied to NOS, and now it only has benefits.

All 'n all I am trying to pass through as 1:1 as possible that stupid digital redbook data, which can't be 1:1 to begin with because of a too low sample rate in redbook. Still the 1:1 principle counts, because in every aspect where I achieve an improvement regarding this, it sounds better (not subjective listening !). This is important to know, because we'd expect it the other way around. If something sounds "digital" it must be rounded to more analogue. WRONG. If something sound digital, something can't cope and you won't be hearing the bits or anything, but just wild distortion.

Two or three days ago I talked about the OpAmp thing. I guess that was a less blabbering story than this one. But I can tell you now : I was right, and it now works. Remember, the OpAmp "gain" wasn't change a bit. The rest did though, and is now ready to receive the high frequency data which it couldn't cope with before. It's that easy, but you have to know where to be to get there.
I think it is very important to realize that when e.g. your speakers can't cope with high transient data (because that is what we're striving for) it is not that you won't hear the transients as should, but you'd merely have distortion;
If you could look at a wave which can't follow what's intended (which is what I could measure as well) it is easy to see what happens with it. It is not only the first cycle which can't reach its peaks, it is merely the second and further which become a great mess. Thus, before the first cycle is at its peak, the now negative voltage draw will try to pull the wave down but that works out too late and in the end a sine has become, well, a nothing. Noise. I am just talking electrical here, let alone mechanical loudspeaker diaphragms.

In the very end it is just about measuring, but the trick is that it is about the whole chain. Thus again : when something in the front-end measures better, it may sound worse and we say "see, measuring tells nothing". WRONG. We may try to feed our speakers with transients which weren't there before, and because of the speakers can't cope it sounds worse. And keep in mind, this is just a rough example, because it really happens from the one electronics part to the other as well.

Blahblah Peter.

[Telstar]

Just two quick comments, Bill.
1) Try to listen to something before you buy it, unless you are totally sure that your tastes and judgement closely matches the one of the reviewer.
2) The speakers are rather unefficient, that requires lots of power to sonorize a medium room (the Soraya is only 105W).

I dont think you are up to a big disappointment, but a not ideal match is very likely.

[bgjohan]

Peter

You have stated that sufficient speed in the amplifier is essential for excellent sound.

I am just wondering if running a stero amp in balanced (mono) mode is a way to increase speed, or this just adds power to the amplifier whithout any impact on the speed.

Bjorn

[Telstar]

Peter

You have stated that sufficient speed in the amplifier is essential for excellent sound.

I am just wondering if running a stero amp in balanced (mono) mode is a way to increase speed, or this just adds power to the amplifier whithout any impact on the speed.

Bjorn

Actually is none of this, although using a mono (or dual mono) amp has several advantages.
It's how the circuit has been designed and build.

[PeterSt]

What Telstar says. I GainClone - and how it should be built - is a good example of it. All the signal wires involved can take a few mm only really. And even short wires me contribute to speed (think of "elasticity" at a longer wire, or better : reaction (stiffness) at the other side). Of course, the longer wires (like for interlinks) can't be avoided, but it is here where "drive" (power, or actually current) comes into play.

Don't misunderstand me with the example of the GainClone (or the real Gaincard for that matter). You may gain on the short wires, but you may (or will) loose it on the high distortion such a chipamp produces in the first place (think near 1% at full gain).

Bringing all the virtues together is the big trick, and this starts with the "design" and (knowledge of !) available parts.

Peter

[bhobba]

Hi Peter and All

Back to the real thing : To my findings - and I am really explicitly working on that for quite some years now - all works out the best if digital is treated how digital should be treated. This is not massive oversampling and (today) it is very easy for me to proove where it destroys. Mind you, I could always do that, but before this was without alternative. So, before this was about NOS being wrong as well, but in such a completely different way that it came down to a subjective choice what was better. NOS with heavy harmonic distortion (and I mean HEAVY) or OS with EVERYTHING wrong, but unmeasureable (because the used filtering is theory only and doesn't work in practice). But this changed since decent filtering can be applied to NOS, and now it only has benefits.

Thanks very much for posting that.

Although I perceived the sound of oversampling DAC's to be better your point is well taken.  My preference may be for a form of distortion and I have had bad experiences with that in the past.  My first speakers were the legendary Gale 402's which had sealed bass.  When it came time to retire them I bought a pair of LS88's which had ported bass.  Initially I thought the ported bass more full sounding and richer - it had the  more immediate appeal.  But over time I came to realize this was false - it grew tiring.  The sealed bass had a rightness about it which a ported design simply could not match.  The 88's had other qualities I liked a lot that more than compensated, but the bass was definitively a minus.  That's one thing I like about the speaker I am interested in.  It is not ported bass - it is a hybrid transmission line ported  design that its designer claims (and I tend to believe him) gets rid of most of the problems with ported designs.

Anyway another choice of DAC would be Steve Nugents overdrive.  It is more expensive, does not have a remote volume control (which I really like as I have bad arthritis), and does not have a home theater bypass mode.  To me all these are also important.  My gut feeling is I may in the end settle for the DAC2 - at least initially - with the idea of upgrading later.  But I may also just decide on a 'better' dac such as the overdrive or Peters Phasure NOS1 (which right now I don't know too much about but will do a bit more investigation).

Thanks
Bill   

[bhobba]

High Telstar and All

Just two quick comments, Bill.
1) Try to listen to something before you buy it, unless you are totally sure that your tastes and judgement closely matches the one of the reviewer.
2) The speakers are rather unefficient, that requires lots of power to sonorize a medium room (the Soraya is only 105W).

I dont think you are up to a big disappointment, but a not ideal match is very likely.

Thanks for posting that.  This is something that has been subject to a lot of debate over on Aether Audio's forum with a lot of very interesting discussion.  A couple of points first.  The Timepiece Mini I am getting, while audibly virtually identical to the Timepiece (it has 85 db sensitivity which is low) except for a bit less bass extension is actually not too bad in the sensitivity department at 87 db (they actually do one thing better than the Timepieces - since they are smaller there is less refections so they disappear - as their designer says 'thought we were listening to a standard Timepiece that had been tweaked...except that this pair of TPs completely disappeared - poof...gone.   You close your eyes and float away to this enormous soundstage and 3-D image.  Then you open your eyes and the brain just dis-connects.  There is no way on earth such a soundstage and dynamics can come from a speaker that small').  The designer Bob thinks 100W is more than good enough to 'send the willies up you' (his words not mine).  The other thing is the Maya that is Hugh Deans latest amp is 150w - not 100w.  Combine this with the fact I listen at lowish volume levels I think it should be fine.  But to be on the safe side I have corresponded with people who have this speaker and they have checked them out with the Soraya and it is fine with that amp - even at the levels they listen to which is well above what I do.  They like them with 400W Cherry amp monsters.  Thats the thing about Bobs speakers they can take that type of power with zero distortion no problemo - it has to do with the waveguide design he uses (he has to pad down the output of the upper end of the tweeter with the crossover - because of that they can handle god awful amounts of power without blowing them or suffering compression of any kind - I won't use them at that sort of power - but its good to know they can handle it).  Anyway despite the fact the Cherry is a bit cheaper I still prefer Hugh's amps - people whose ears I trust say they work very well together.  Also remember the ear is logarithmic in its response - 100W amp only sounds half as loud as a 1000W monster.

I could write a bit about 'listening' to equipment and I will probably do a much longer post on it one day.  Suffice to say I am not that big a fan of it despite the fact everyone says its what you should do.  One reason is you can be easily fooled - and I have been.  I tend to trust guys who listen to a lot more stuff than I do and who technical skill is beyond reproach.  These are guys like Hugh Dean at Aspen and Bob Smith at Aether Audio.

Thanks
Bill

[xp9433]

Bill

Speaker sensitivity and amplifier power rating are never as simple as people expect. For example a 3-way speaker makes greater demand on an amplifier (both current and voltage) than a single cone of the same sensitivity. It is very easy to overstretch an amplifier and greatly increase distortion/clipping on musical transients. The arithmetic doesn't always work and your ears will tell you easier than the maths will, providing the listener recognises the symptoms.

I can remember at an CES where a new open baffle speaker design (94dB senstivity 8 ohms) was being demonstrated using a respected 35watt tube amp driving the satellites. The music was hardening on peaks, for example when a soprano was singing. I offered the use of a Plinius SA250, which I was taking for review, and the problem disappeared. Linkwitz (one of the speaker designers) measured the output from the Plinius on "soprano peaks" at in excess of 180 watts, and this was in a smallish room. Go figure.

Needless to say I am a big fan of powerful amps rather than the other way around. I believe in amplifier headroom. I have seen too many speaker drivers destroyed with underpowered amps, and many owners suddenly relax with the new "natural unstressed reproduction" on dynamic musical passages when a powerful amp was inserted into their system. Most didn't know they had an underpowered amplifier problem until the stress and distortion problems were taken away.

Frank