Wavelength Audio USB Spacelator

[bogi]

3 hours ago, CG said:

I did and posted the results in another thread.

Could you provide link to your measurements?

[CG]

Jeez, is the search function broken on this website?  I'd better contact Chris.

 

I'd have to search myself.

[CG]

10 minutes ago, bogi said:

 

I wrote "it looks Spacelator does not implement galvanic isolation". Not that I know! :)

Miska previously stated in HQPlayer thread discussion that Intona and HS02 provide galvanic isolation.

And in this thread Miska wrote "This doesn't sound like the power lines would be galvanically isolated. Like they are on Intona." as reaction to "The power supply is sent though a linear filter, then regulated to 5V/500ma before it hits the Device USB connector."

Maybe Miska could explain, where he sees the difference against Intona.

 

I would suggest that you contact the manufacturers of these devices.  They would certainly know what's inside them.

[bogi]

2 hours ago, CG said:

One person I know owns both and prefers the Wavelength Spacelator to the Intona.  (See above for this comments.)

 

I own both an HS02 and a Spacelator.  I prefer the performance of the Spacelator.  The HS02 sits in its box.  (OT:  I wish I knew of a reasonable and reliable way to sell excess stuff like this.)

 

Most of USB 'enhancers' don't provide galvanic isolation and still users report improvement.

An example device of that category is iFi iPurifier. Except of reclocking it does yet active noise cancellation and power filtering. It does not break ground loops.

I am using HS02 together with two iSilencers. I clearly hear improvement when iSilencers are plugged in.

If Spacelator does something more than breaking ground loops then that can make the difference against Intona.

Wavelength Audio  is stating "The downstream portion of the USB is reclocked with a low jitter clock." Or maybe they do better filtering.

 

From the previous communication it looks that Miska does not consider USB reclocking as something what could have positive impact on sound. He stated in this thread "USB clocking is pretty irrelevant aspect.". In December we discussed an  issue when GMG perceived difference in sound when 2 different computers were used despite of he uses fully galvanically isolated optical USB cable. The opinion I expressed that time was that inaccuracies when USB packets reach DAC USB receiver could generate noise patterns in DAC which could bring the audible difference GMG perceived. I cannot provide any proof but I see that reclockers exist and users report improvement.

 

 

[bogi]

23 minutes ago, CG said:

I'd have to search myself.

 

If you remember some specific keywords you used in that post you can use Google search this way:
CG your keywords site:audiophilestyle.com

[CG]

4 minutes ago, bogi said:

From the previous communication it looks that Miska does not consider USB reclocking as something what could have positive impact on sound.

 

You have to ask Miska himself about that.  I wouldn't want to second guess what he was saying or put words in his mouth.  I am sure that he has a well considered reason, though.

[CG]

3 minutes ago, bogi said:

 

If you remember some specific keywords you used in that post you can use Google search this way:
CG your keywords site:audiophilestyle.com

 

Guess what!

 

I just clicked on that self-portrait icon next to one of my posts.  It immediately brought me to a list of recent posts made by me.  There, in glorious color even, the plots showed up in the brief summary of the posts.  (I found your recent posts by clicking on your icon, too.)

[bogi]

5 minutes ago, CG said:

You have to ask Miska himself about that.

Yeah I could split my previous post to two posts. I did not await reaction from you to the 2nd part. Sorry.

 

7 minutes ago, CG said:

I just clicked on that self-portrait icon next to one of my posts.  It immediately brought me to a list of recent posts made by me.  There, in glorious color even, the plots showed up in the brief summary of the posts.  (I found your recent posts by clicking on your icon, too.)

 

Link is better (after few months your newer posts will appear on top in your profile).
3 posts starting here: https://audiophilestyle.com/forums/topic/27754-ayre-acoustics-qx-5-twenty-–-the-digital-hub/?do=findComment&comment=1242537
I suppose the measurements are about Spacelator (you did not mention name of the isolation device in these posts).

[Miska]

3 hours ago, CG said:

OK.  Maybe they don't work equally.

 

Measurement data would show how these rate, that's why I was asking about it. How much for example this makes difference to Intona, measured from DAC output.

 

[CG]

2 minutes ago, Miska said:

 

Measurement data would show how these rate, that's why I was asking about it. How much for example this makes difference to Intona, measured from DAC output.

 

 

I don't have an answer for you on that.

 

What I can say is that most of the very little measurement data I've seen published was made with an Audio Precision test system.  That's obviously good in some ways, but bad in another.  

 

AP gear is very well engineered to make measurements of the device under test under ideal conditions.  The CMRR is as good as it gets.  The AP power supply is very well isolated from the AC line.  And so on.  Any Heisenberg questions have been minimized.  

 

If you want to introduce imperfections, like how an amplifier's distortion might change when not driving a purely restive load, you have to add the imperfections yourself.  There's no standards for these, generally, and very few testers even bother.

 

So it is with a lot of the digital measurements.  When you try to replicate the environment of what a real system in a home is like, you often get flaky and unpredictable results.

 

I got especially interested in this when I bought a pretty good A-D converter for testing along with a DAC that I wanted to use as a signal source.  The A-D arrived first and I tested it with a really low distortion (like -150 dBc) oscillator that was battery powered.  With a notch filter to reduce the fundamental frequency into the A-D, indeed the measured distortion was about that.  With no notch, the worst distortion product was shown to be the third harmonic at about -135 dBc, if I remember right.  That demonstrated the limits of the A-D.  Can't much complain about that performance.

 

When the DAC arrived, I plugged it into the same test system and the measured noise floor bounced and there was crap all over the place.  I tried fiddling with various measurement software settings, the device drivers, and everything I could think of.  I was convinced the sample I had was broken and was getting ready to return it.  

 

In desperation, I tried attaching the A-D to a second, separate computer which was plugged into a different AC mains socket across the room.  So, the DAC was running on one computer generating a tone and the A-D on another measuring the output of the DAC.  The performance still was quantitatively bad but was entirely different qualitatively.  

 

Hmmm...

 

So I then tried attaching the A-D to a battery powered laptop.  I got almost exactly the same results as what was published in the DAC reviews I found online for the product I'd bought.  The crap was gone entirely.

 

Must be a common mode current loop, I thought.  When the DAC was connected back to the test system computer, adding a USB isolator fixed all of the bad performance, which mostly confirmed my thinking.

 

Now, an obvious conclusion might be that the A-D is deficient.  Well, I measured the common mode rejection at its input to be somewhat better than -60 dBc.  That's not as good as an AP test system (according to its published specs.)  But, that level of common mode performance is almost certainly better than all but a few preamps or power amplifiers that people use to play music.

 

My thinking, which everybody is welcome to disagree with, is that real systems in actual use have lots of noise currents everywhere, and a lot of it is transmitted as a common mode signal.  Improving that condition seems really desirable.  After all, I could measure it.

 

To be fair, what I measured was just a part of a system.  The computer running the software probably generated the tones and played them out from memory.  I have no idea what the analysis part of the software did in terms of toggling hardware bits on the computer board, generating noise.  

 

A music system is somewhat different in its task, for somewhat obvious reasons.  How that affects noise generation, I can only speculate.  

 

Plus, there was no preamp or power amp attached in my test.  Those would only complicate the common mode current mesh in ways that are hard to predict.

 

I suppose I could try to measure the noise currents directly with current probes, spectrum analyzers, and so on.  But, why?  I'm not selling anything.  I'm not publishing a paper.  I only do it for my own purposes.  If I come across a seemingly good solution that I like and seems to conform to some idea I have (that doesn't happen all the time...) that's good enough for me.  I'm even reluctant to share anything I might find with anybody but a couple close friends who are interested.  I don't wish to spend my hobby time in a online debate where the debate will become the thing.  I only responded here because somebody asked what I thought was a very reasonable question that I have some information about.

 

Sorry for going on like that.  

 

 

[Miska]

8 minutes ago, CG said:

To be fair, what I measured was just a part of a system.  The computer running the software probably generated the tones and played them out from memory.  I have no idea what the analysis part of the software did in terms of toggling hardware bits on the computer board, generating noise.  

 

A music system is somewhat different in its task, for somewhat obvious reasons.  How that affects noise generation, I can only speculate.  

 

I just perform standard set of measurements with isolator and without isolator and compare the results. Test signals are played same way as music would be played. Files just contain test signals instead of music.

 

[CG]

6 minutes ago, Miska said:

 

I just perform standard set of measurements with isolator and without isolator and compare the results. Test signals are played same way as music would be played. Files just contain test signals instead of music.

 

 

Well, I'm not sure I agree entirely.

 

Test signal files tend to be short.  They get loaded into memory and get repeated.  Music files are longer.  They often get loaded from disk to memory. (Some software does allow you to load the music into memory - a sort of RAM disk)  The SSD or hard drive has different current demands, often from different regulators because of the voltages, if I recall right.  That has to change the generated noise in some way.  By how much, I can't say.  But, I have observed in two entirely distinct playback software solutions on two different computer architectures that playing from memory sounds different.  Better, to me anyway.  That suggests that something is different.  What, I don't know.  Nor have I even tried to measure that.

 

Anyway, the measurements I made were exactly as you described.  None were of an Intona product.

[Miska]

39 minutes ago, CG said:

Test signal files tend to be short.

 

My test files are 10 minutes long.

 

39 minutes ago, CG said:

That has to change the generated noise in some way.

 

If your isolator is working as intended, any possibly generated noise has no effect whatsoever.

 

In addition, if you use a NAA, what ever you play always looks the same from NAA perspective. It is just constant network data flow headed to the DAC.

 

Quote

Anyway, the measurements I made were exactly as you described.  None were of an Intona product.

 

Yeah, I measured Intona. And it worked exactly as described. I've had some other USB gadgets over time which didn't...

 

[CG]

1 hour ago, Miska said:

If your isolator is working as intended, any possibly generated noise has no effect whatsoever.

 

In addition, if you use a NAA, what ever you play always looks the same from NAA perspective. It is just constant network data flow headed to the DAC.

 

 

Yeah, I measured Intona. And it worked exactly as described. I've had some other USB gadgets over time which didn't...

 

 

Big if.  Nothing is perfect.  I wish that wasn't true.  Any is a very exact and absolute word and no effect whatsoever is a very exact and absolute phrase.

 

I've seen estimates that the best isolation devices, except for optical fiber which isn't used in any the devices we're discussing here, have a coupling capacitance of around 5 pF or so.  Think of what impedance that is at, say, just 1 MHz.

 

The best 50/60 Hz power transformers I've measured come in at about 45 pF or so.  Admittedly, I have not measured every available transformer.  That 5 pF for the isolator is effectively in series with the 45 pF of the power transformer for that part of the common mode current mesh.  You could use an open loop DC-DC converter to improve on that, and those transformers often are also  in the 5-10 pF coupling range.  That all helps.

 

In my (former) day job, I learned over time that even a small amount of noise will degrade mixed signal conversions.  That could come from transients on the power supply plane, common mode signals converted into differential mode signals for a bunch of reasons, and so on.  Those can quickly reduce the MER/SNR/BER pretty quickly.  Which is why most digitally encoded transmission systems employ some sort of error correction scheme.  Or, more than one.

 

Would these same effects be audible in a home audio system?  I never got involved with that, so I'd only be guessing here.  These devices aren't perfect black boxes that look nice in a block diagram.  At least, they never behave that way.

 

~~~

 

I don't know.  I don't use a NAA.  But, I certainly believe what you're saying there.  So, does that mean it's always degraded?

 

~~~

 

I believe you and you probably have measurements you are willing to share, right?  

 

~~~

 

Anyway, this has gotten way off the track of the original subject matter of this thread.  I apologize for my role in that.  As I suggested  before, I'm certainly not going to convince anybody of anything or help answer their questions.  Besides, at least in audio, whatever people like is all good.  It's a hobby.

[Miska]

40 minutes ago, CG said:

I've seen estimates that the best isolation devices, except for optical fiber which isn't used in any the devices we're discussing here

 

One reason I developed the NAA concept was because optical Ethernet provides absolute isolation and plenty of bandwidth. It avoided all the shortcomings of Toslink. Quite a lot of people use NAA over optical networking.

 

41 minutes ago, CG said:

Would these same effects be audible in a home audio system?

 

On my tests, Intona 7054 already cleaned up enough that all the conducted noise peaks fell below noise floor, meaning well below -150 dBFS. That can be probably considered enough for home audio.

 

[CG]

7 minutes ago, Miska said:

 

On my tests, Intona 7054 already cleaned up enough that all the conducted noise peaks fell below noise floor, meaning well below -150 dBFS.

 

 

How did you determine that?

 

(Note - I violated my promise to stop posting here, but I want to understand this.  So, no response from me after this.  No need to cheer.)

[Miska]

8 hours ago, CG said:

How did you determine that?

 

By measuring? While playing test signal, everything was clean at least to -150 dBFS.

 

One of the common offenders is 8 kHz USB packet ticking. This got cleaned up as well.

 

[CG]

19 minutes ago, Miska said:

 

By measuring? While playing test signal, everything was clean at least to -150 dBFS.

 

One of the common offenders is 8 kHz USB packet ticking. This got cleaned up as well.

 

 

With 20 ENOB, maybe, doesn't that level of measurement imply a lot of averaging?  So, for example, if any noise was more or less randomly distributed, it would show as a modest change in noise floor.  And, how do you know the A-D contribution?

 

I'm only asking to understand.  If it's above my level or a secret of some kind, no need to reply.

 

 

[Miska]

5 minutes ago, CG said:

With 20 ENOB, maybe, doesn't that level of measurement imply a lot of averaging?  So, for example, if any noise was more or less randomly distributed, it would show as a modest change in noise floor.

 

256k point FFT with 8 averages, 100 kHz bandwidth. Gain ranging is especially useful for this kind of measurement.

 

I compare with/without the isolator. And these kind of computer noises are generally not white noise, but instead various tones with their harmonics. If it would be white noise and below the DAC's analog noise floor, then it doesn't really matter because it would be totally masked by the DAC's self noise. Modest changes in noise floor are easy to see by using overlay masks.

 

5 minutes ago, CG said:

 And, how do you know the A-D contribution?

 

I know it quite well, since it's properly calibrated audio analyzer I've been using for years with tens of different DACs. Well, actually I have four different analyzers I use for different purposes, depending on what kind of analysis I'm precisely doing.

 

 

I've been working professionally on DSP and signal analysis systems since early 90's. So not totally newbie on this.

 

[CG]

24 minutes ago, Miska said:

 

I've been working professionally on DSP and signal analysis systems since early 90's. So not totally newbie on this.

 

 

Didn't suggest or imply that you were a newbie or even not an expert.

 

I will say that my own experience, not having to do with audio, has been than non-synchronous and uncorrelated imperfections tend to get lost in the noise of measurement, so to speak.  Since they tend to be random, sort of, short bursts get averaged out because they appear once during a sweep and not again for a while.

 

An example of that is watching the constellation of a QAM or ODFM signal on a spectrum analyzer.  You can occasionally see a collapse or explosion of the displayed constellation.  Since it only occurs during that one frame, it barely affects the averaged MER.   But, if you were to be keeping track of the BER, you'd see errors.  (One solution to that problem is to chart the MAX or peak level of the noise floor over a longer period in order to capture those randomized bursts.  I've seen bursts like that caused by resonances in the power distribution system that just happen to get excited by a particular code combination getting processed.  Lots of other reasons, too.)

 

The test equipment can fool you, too.  Really good test equipment puts the DUT in an ideal environment where flaws don't necessarily show up as they would in actual use cases.

 

But, this is audio.  Some aspects get judged by people listening.  Other aspects get judged by people making measurements.

[Miska]

1 hour ago, CG said:

I will say that my own experience, not having to do with audio, has been than non-synchronous and uncorrelated imperfections tend to get lost in the noise of measurement, so to speak.  Since they tend to be random, sort of, short bursts get averaged out because they appear once during a sweep and not again for a while.

 

This is what I use peak hold mode for. Another way is to play suitable silence pattern and enable trigger capture just above noise floor. I can have all the different things running in parallel.

 

Remember though, that you cannot use plain digital silence, such as 0's as test signal, as that will trigger output mute on the DAC!

 

For example lost USB packet (125 µs gap), which usually produces faint tick in the sound is very obvious in the measurement, because it causes waveform discontinuity which in turn explodes the FFT result by tens of dB.

 

But one such harder to catch thing is ESS DAC chip fingerprint which causes slowly wandering noise floor. It's like waves in ocean and the cycle takes tens of seconds.

 

1 hour ago, CG said:

The test equipment can fool you, too.  Really good test equipment puts the DUT in an ideal environment where flaws don't necessarily show up as they would in actual use cases.

 

Yes, that's what I'm doing. That's why I said the proof is in measuring from the analog domain. At very least from DAC's analog outputs. I can also easily measure for example from power amp speaker terminals. Since the analyzer has maximum input voltage of 300 VAC.

 

Are you trying to say that Intona doesn't work (they also provide some measurement results for their products)? Those are also used in industrial test and measurement environments. They are not audiophile products as such, but serious tools for serious environments. Or are you trying to say that the Wavelength is somehow magically better? Not sure what kind of conclusion you are trying to reach?

 

1 hour ago, CG said:

An example of that is watching the constellation of a QAM or ODFM signal on a spectrum analyzer.  You can occasionally see a collapse or explosion of the displayed constellation.

 

My example is catching submarines, submersibles and divers from sea background noise in passive sonar system, that's what I used to work on, back in the days...

 

[vgrubb]

Will a lps 1.2 adequately power the Spacelator? Also, those who have used the Spacelator or Intona, would you spend your budget on a really good USB cable or split your budget between a lessor cable and one of these?

[Mercman]

 

2 hours ago, vgrubb said:

Will a lps 1.2 adequately power the Spacelator? Also, those who have used the Spacelator or Intona, would you spend your budget on a really good USB cable or split your budget between a lessor cable and one of these?

The LPS 1.2 should work fine with the 7v setting.

 

The Curious Evolved USB cables ( input and output) made a big difference for me with the Spacelator. 

 

I guess I would go first with the cables for either the Intona or Spacelator, but I heard a definite improvement when using a linear power supply  (Uptone Audio JS-2) with the Spacelator.

[vgrubb]

Thanks Mercman

[ittaly]

On 6/18/2023 at 2:10 AM, bogi said:

No specs on their page ... very unusual.

It's okay for the manuf. to play their marketing strategy in any way they choose on their own site (or via the std. reviewer channels, Twittering Machine, etc).

What I find unusual is that in this AS thread, there are no formal measurements (yet) or disassembly/inside photos.