Tuesday, September 11, 2018

XSP-1

No doubt, the Emotiva XSP-1 is a fabulous preamp.  And in every way except looks competitive with everything on up, including the $30,000 preamps like the Levinson ML52.  It has now, after far too long, taken it's rightful place in my Living Room System.

The Moving Coil preamp must be about as quiet as it gets.  I'm finding it so quiet I can not hear any noise at all, even with my ear to the speaker, except under rare circumstances*.

(*No refrigerator or AC running, in the deepest night, and comparing the sound ear to speaker with the sound immediately to the side of the speaker after playing a chamber concerto at full loudness**, I could finally hear the faintest of pale high woosh.  This has to be consonant with at least 80dB S/N.  Yes this also shows how hum and noise free the rest of my system is.  **It was +1dB gain on the Emotiva .  I listen to as high as about +4dB so far--that's about as high as I need for the loudest I can stand.  0dB gain on the Emotiva seems a higher output level than the XPS-1 I was using before, for which I was always so gain starved I dialed in 4-6dB of digital gain in my Tact, which can't restore the missing information from a too low level digitization. )

So this is far quieter than even the XPS-1, which was far quieter than the (well known to be quiet) dB systems DB1HG, which is an order of magnitude quieter than my old tube preamp with the Jensen transformer I was using in the 80's.  (My tube preamp, which took two years to build, blew away the Audible Illusions 3 that a friend of mine had.  Sadly, I ultimately dumped the old power supply, though not as well built as I would have liked it was still working perfectly.)

With the tube preamp, I could hear the hiss from the listening position after any record ended.

Hiss might make music seem more expansive, increased depth, etc., but it's all fake.  Quiet reveal the real depth and dynamics that are actually in the recording.

So far I can't fault the phono sound.  I think it's better yet than the Emotiva XPS-1, which was better than the dB Systems and so on.  Records are revealed as never before, without anything added or exaggerated.  The first record I played, formerly believed to be worse than mediocre was stunning in fidelity at  (Rick Wakeman's Journey to the Center of the Earth).

It seems fully immune to turning the tensor light on, but not entirely immune to turning it off, a very tiny pop.  With the XPS-1 it was more like an explosion, which might max out the Lavry inputs (set to about 3V).

My biggest concern was this: as a result of running the XPS-1 balanced to my Lavry ADC (balanced is by far the best way to use the XPS-1, btw, according to my measurements) I can no longer switch single ended signals (from other disk players than the balanced Oppo BDP-205) passively, as I was doing before.  Now, I MUST run all the line signals through an amplifying device.  How much would other things loose.

THAT's actually what I spent most of the first weekend investigating, and my ultimately conclusion was high level inputs lose very little, if anything.  Though at first I thought otherwise, I finally figured out why it was sounding strained--it was clipping the input of my Lavry ADC set to 3V for 0dB full scale.  One always has to be careful about the gain settings.  I wasn't paying attention to the fact that the balanced outputs were putting out twice as much as a single ended input of the same level (and, as it turns out, the Emotiva doubles the level of single ended inputs when it outputs them balanced, that is the way it does the conversion, which is what one would expect but I hadn't been fully thinking it through).  I somehow had been thinking the Lavry would automatically adjust for this, so a 2V "pseudobalanced" (through an RCA to XLR adapter, mine was a Cardas) input might be equivalent to a 4V balanced.  That's the way that SE and balanced outputs often work--the balanced has double the voltage because that's what you would get with a equal differential signal.  But, no, from the Lavry's perspective 2V is 2V and 4V is 4V, whether balanced or unbalanced.  SO that is actually adding about 6dB gain to everything, including the phono, compared with no preamp, at the supposed "0dB" level.

But since I have an 0.5dB resolution gain control, I can now easily fine tune that for any input.

I try to adjust all inputs so that the peak reaches the -1dB level, but NOT the 0dB level.  Reaching the 0dB level on the Lavry is likely to indicate clipping.  If the peak never gets past -1dB, you are perfectly safe.

I have determined so far, I am safe with the Emotiva set to -2.5dB for the Oppo BDP-205 and -3dB for the Sony 9000ES.  This is with the Lavry at -10db setting (this makes something like 3.8V the 0dB level).  This puzzled me because with the balanced outputs of the Oppo, it is putting out nearly twice as much differential voltage as the Sony.  But I have to remember, the Oppo is going balanced straight through the Emotiva, wheras the Sony is being converted to balanced by the Emotiva, which is also doubling the output.  I've been meaning to compare these two players on SACD.  I have always believe the 9000ES to have a special place in SACD reproduction (when it works...) because it has the same true DSD converters as the original Sony SCD-1.  Everything after that went to sigma delta, which might actually be better, but isn't truly native DSD anymore.  Well the 9000ES isn't DSDx1, it's actually DSDx10 !  Sony was trying to see what they could accomplish with a purist approach to DSD at first.  So the 9000ES may be magic, but it's also possible, not yet verified by me, that a Good Enough sigma delta will be even better.  So I need to do this test.  From relatively casual listening so far, I can't tell, my FUD is telling me I need to keep the 9000ES.  Of course I am also keeping my Denon DVD-9000 for HDCD.  And the DVD-9000 has a special place for HDCD.  It is one of the VERY few players that implements HDCD as Keith Johnson specified--it almost doubles the peak for the "peak" mode of HDCD.  So that means that HDCD's actually reach almost 4V, which was why I've had my Lavry set to -10dB for some time--that seems to work for most HDCD's on the Denon to the Lavry.  Now, however, with balanced conversion rather than adapting, I'll need to cut the gain level to -5dB or so.  So far I've determined I need about -4dB but lower might be needed.

The 6dB of "bonus" gain provided by the unbalanced-to-balanced conversion would not apply to amplifiers with truly separate balanced and unbalanced inputs that have different sensitivities.  However I suspect most power amps, including my Krell FPB, have the same sensitivity for both balanced and unbalanced inputs, just like my Lavry ADC.

Convenience

It is also extremely convenient for me to have the volume control AND muting switch (another plus) right underneath the turntable.  Previously I had to jump 10 feet away to mute the system on my Tact digital preamp, or I had to flip to an unused input--which was very risky because if there were CD player signal on that input it would be 10dB higher in level than the XPS-1 I was using.

One exception is that it was much easier to set the loading level.  With the XSP-1 the loading switches are on the back.  Of course, on the XSP-1 this is good because ALL the analog circuitry is in the back, and shielded from the power supply AND digital controller and display.

It's also nice to have the push-button selector switches, and to have lighting around all the controls so you can see where they are in the dark.  It would have been nicer to have lighted labels on all the buttons as well, but only McIntosh with their glass panels has that.

One friend of mine is very upset that Emotiva did NOT include a polarity switch.  I think polarity is of marginal importance, but I would have liked to have had a polarity switch also--for convenience.  Most people can change polarity by reversing speaker connections, but it takes a lot of work for many audiophile systems (and mine, with 3 speakers on each channel).  For me, this is not an issue because I can change polarity in my downstream Tact digital preamp, which is easier from the listening position anyway.   If I were into the Polarity Obsession, I would not buy this Preamp on the principle that it doesn't have a polarity control, and such products deserve boycott.  Solid state high end big dogs like Levinson and Krell generally do have polarity switches.

Meanwhile there are many features that are completely unimportant to me, such as the built-in crossover.  Nothing like this provides close to the flexibility of my Behringer DSP's, and my next thing will be the use of 3 miniDSP units which, using digital FIR filters, can correct phase response and produce a linear phase Linkwitz Riley.

On similar grounds, I doubt I'll be much using the tone controls, but I think digitally switched tone controls are generally a good idea.

Having two balanced inputs is excellent.  Many historic balanced preamps only had one balanced input.  All in all, having 4 single ended and 2 balanced inputs is excellent, and I'm finding I do not even need to cascade and additional output selector switch as I was doing before even to include my two tuners in the lineup (I don't need to include the tuners because they are directly encoded in digital by the Sonos Connect units that see their fixed outputs).  Actually I have observed that the variable output of the Pioneer F-26 is sufficient inferior to the fixed output (depending on level setting) that I may prefer it going through Sonos anyway, and I decided to remove the F-26 variable output connection to the analog preamp so that it doesn't load down the fixed output any because there is no buffering in the F-26 output.  So that actually leaves me with one spare unused input!  (That sort of thing never seems to last long.)

Performance

Audiophiles have unfortunately been trained to believe measurements are meaningless.  I don't believe that for a moment.  Although I suspect I could not easily hear distortion differences below 1%, I take it as a given that, generally speaking, distortion is one thing that needs to be eliminated to achieve perfection.  And likewise noise should be eliminated.

And the Emotiva specified distortion levels, and also noise levels, are as good as I have seen anywhere.  My own measurements yield the same or better numbers (which confounded me, as the distortion level I measure is lower than the residual of the Juli@ soundcard I am using for the measurment).  This is especially true for the balanced inputs, where I have measured as low as 0.0003% distortion.

Meanwhile, Levinson specifies 0.03% even for their $30,000 532 preamp, as well as their others.  It appears, however, that the Levinson numbers are extremely conservative for the top of the line 532, but not very conservative for the "inexpensive" $10,000 model.  Stereophile easurements of the 532 preamp show noise and distortion about as low as the Emotiva.  Measurements of other Levinson models show higher levels of noise and distortion than the Emotiva.

The spectrum of distortion is also excellent, being confined generally to only the first two harmonics being visible above the noise floor.

Comparisons


The two chassis behemoth Mark Levinson Model 52 preamp is certainly fabulous.  However, these detailed measurments at Secrets does not show superior distortion performance (it shows virtually identical performance as the XSP-1 at 0.0003% THD, same as the Emotiva specifies for balanced to balanced, and that it pretty much mostly what the Levinson measures in the review at 4V output.   (Levinson specifes the 32 conservatively with the same 0.03% spec as used for their "cheapest" model, at $10,000, which does not exceed the spec quite as vastly.)

Now, the Levinson might be superior in other ways, but the only apparent one is the immense output levels, up to 20V RMS.  The Emotiva is only capable of 12V RMS.  I can't imagine needing as much as 12V.

I'd like the polarity switch, assignable input names, and 0.1dB assignable default levels of all Levinson's since the 90's.  As John Atkinson said, praising the 38, it is an audio reviewer's dream, and aren't all audiophiles reviewers?  I've long lusted for a 380s, not to mention a 32, but I wonder if I'd actually find their performance equal or even inferior to the brand new Emotiva.  I saw my beloved Classe CP35 fail, and before than I needed to have my Aragon 28k repaired.  20 year old gear is pushing it.  And the price and gamble aspects of it.

I have seen inadequate noise performance on the less expensive models and earlier generation models.  For measured performance, probably little before the current Model 52 will compare.  Old classics from Levinson and Krell won't make the grade--as the latest and greatest Levinson is a mere equal.  Spectrai??

The famous Vendetta SCP-1 claimed some S/N in the 90's IIRC.  However, that was at a very high output (taking advantage of the enormous headroom, as compared with the "standard" level for performing such a test.  Stereophile measured 82dB, A weighted.  The spec for the the Emotiva is >39dB, which is <3dB different.  My experience suggests the Emotiva is an equal in noise level, but I have not made a measurement (which...I do have equipment for...).

The Volume Control


One of the things that sold me on the XSP-1 in the first place was the "R2R attenuator."

I envisioned something like the relay switched resistors in the very expensive Krell KRC series of preamps from the 1990's, and the Placette passive and active preamps, as well as brand new hifi attenuators made in China and sold on eBay nowadays for about $129.  I also imagined this kind of attenuation in the top of the line Mark Levinson model 32 Preamplifier and all the preamplifiers Mark Levinson has made ever since.

Mark Levinson actually says "discrete" resistors and says they are switched by "analog switches" not relays.  But what are "analog switches"?  One intrepid DIYAudio'er enlarged the original ML 32 interior photo and matched the volume control "analog switches" to the these CMOS devices--which their manufacturer Vishay calls analog switches.

In 1995 even Rowland Research was praising the virtues of the (now outdated for high end) Crystal Semiconductor (now Cirrus Logic) CS3310.  The CS3310 is rated at typical distortion level of 0.001%.

Both that chip and the more updated Muses 72320 and the 2005-era Cirrus Logic CS3308 and CS3318, internally use precision resistor ladders (R2R).  I suspect the Emotiva uses the CS3318...notably that chip has the same 117dB noise and 0.0003% distortion specs as the entire Emotiva XSP-1 itself.  The preamp itself can avoid being cumulatively worse than it's weakest part because the volume chips are run differential balanced--along with everything else, which increases S/N.

I don't recall that the 90's era Krell KRC preamps with discrete resistors AND relay switching, came close to the noise or distortion of the 532 and XSP-1, despite using relay switching.  (Notably, David Rich was critical of Krell Preamp circuitry in the 1990's.)  It's hard to make discrete circuits as clean as the best chips available now, such as the LM4562 opamps that Emotiva uses (though, the OPA 211 could be even better, no audio gear uses that).  The best chips I just mentioned are the process of three decades of intense competition to build the best linear amp.  Older chips are passe (especially the one sadly used in my Denon DVD-5000's which I was hoping to use as DACs).

Meanwhile, it also appears, that electronic switching in attenuation has been tamed, though it is not perfect, and I'd pay a premium for relay switching given the same quality electronics, but it appears that right now the only way to have it all would be to build my own preamp that does everything right.

Now, that master of the universe Paul McGowan of PS Audio has very little nice to say about any kind of chip volume attenuator in this thread which started out as a damning discussion of LDR (light dependant resistor) volume controls.  A perfect illustration of typical High End Audio's stupid dismissal of the very high level engineering that goes into the best analog chips.  However, later on, McGowan is forced to recant slightly because it is revealed that the well regarded Pass Labs XP-30, designed by another master of the universe, actually uses the Muses 72320 chips.  McGowan then admits that a good enough designer can use them "properly."

And it turns out there is something very special about the Muses 72320.  At face value, the Muses has no better performance than the plain old CS3310.  However, uniquely, the Muses can be used in bypass mode, where only the internal switched resistors are used, and not the internal (and not quite the best) opamp.

None of the Cirrus Logic chips can be used in that bypass mode (I determined by reading the datasheets linked above).  However, the CS3318 has an internal opamp that is almost as good as the best opamps anyway, and it uses it in a clever way that can use gain as well as attenuation, thereby maximizing the performance by reducing the need for actual attenuation.

So, as you can see, little is lost by using one of the best chip volume controls nowadays.  And that is how the XSP-1 can be about as good as it gets.  The best IC preamps are better than mortals can design discrete, because IC's are so well perfected now.  If an IC preamp used the SOTA chip, the OPA 211, it would probably be better than anything that could be made discrete.

[I might add more detail to this review later.]










No comments:

Post a Comment