EARLY LOOK: The E1DA Cosmos Scaler prototype. As usual, Neil Young (in "World Record" album interview) remains neurotic against digital...

 

Today, let's talk about something rather unique.

In the image above, we see the E1DA Cosmos Scaler in use as I was running measurements of some DACs I have around here. As I mentioned last time, this is a prototype product which although has been announced/discussed online, as far as I am aware, is not at the stage of imminent production or public release. As such, be mindful that details I describe here might change if this were to be released as a product.

The Scaler functions as a balanced 2-channel in/out pre-amp that can apply 16 discrete levels (this is not a regular volume knob for routine volume control) of very clean amplification (using OPA1612) from 0dB to +26.7dB along with an "auto-ranger" feature that aims at a target voltage level to feed the ADC. Along with this, the Scaler's stable high 200kΩ balanced (100kΩ unbalanced) input impedance resolves one of the limitations of the Cosmos ADC; the lower input impedance for example 1.66kΩ at a commonly-used 4.5Vrms setting.

Thanks again to Ivan Khlyupin (IVX) for sending me the package to test out.

As with any pre-amp product, the question is how quietly does this device perform when feeding the output into a high-resolution measurement ADC in this case. And related to that, whether the device adds any of its own distortions.

First, let's take a look at the box itself:

The device enclosure has identical dimensions to the previously tested Cosmos ADC and APU. The black anodized aluminum shell feels robust and the white laser-etched markings provide an easy-to-read contrast. The rear plate for this prototype is exactly like the APU with just a USB-C connector for 5V DC power (a clean, low-noise, lithium battery pack is recommended). On the front, we have a nice, smooth, high quality, metal knob for volume control from 0dB to +26.7dB. There's a "click" when turned down from 0dB to "Auto" mode which is the "auto-ranger" setting. In use, whenever the pre-amp level changes, you can hear a soft click of the internal relays.

Given the small form factor, as you can see, there are two sets of stereo 4-pole balanced 3.5mm phono inputs (+/-/GND/GND) for in/out instead of XLR connectors. You can purchase 3.5mm phono-to-XLR cables (these male XLR, and female XLR should work), but I constructed some using generic 4-pole phono connectors and cut up a pair of 3' XLR cables for male and female ends:

No need for expensive cabling to demonstrate very high levels of performance... ;-)

This way I can customize the lengths I want; with a short 6" cable going from Scaler to ADC (male XLR end) to minimize potential noise and the DAC to Scaler cable can be longer at around 3' (female XLR end) for convenience. This is what the final cables look like attached to the Scaler with the Scaler-ADC units slapped together using some rubber bands for convenience.

The output impedance from the Scaler is a low 20Ω, great for feeding into the Cosmos ADC:

There are 2 LEDs to show output level for "OutL" and "OutR" with different colors to indicate levels for visual confirmation. By default (4.5Vrms auto-ranger target), LEDs are turned off when output voltage is <1.7Vrms, it turns blue between 1.7-4.5Vrms, green between 4.5-10Vrms, and red >10Vrms as a potential warning that the signal appears a bit "hot". Voltage limit of the preamp is up to 20Vrms balanced, and 10Vrms unbalanced.

When in the auto-ranger mode, the default output target is 4.5Vrms which can be reprogrammed. To change the target level, present the output target voltage you want into the left input and turn the knob to 0dB. Then very quickly, switch the knob between 0 and "auto" 5 times over 2 seconds (yes, this takes a bit of dexterity!) ending up in "auto". The LEDs turn white to indicate that it's ready to be programmed. Then click it one more time from "auto" back to 0dB and it'll remember that target voltage. You can set the target output level to between 1-10Vrms. If you ever need to go back to default, clicking the knob 5 times as above with <1V input into the left channel (eg. silence) and get back to 4.5Vrms.

Since these days many DACs will have >4.5Vrms balanced output, I've set the Cosmos ADC to 6.7Vrms and likewise the Scaler auto-ranger to the same target. 

[I used the RME ADI-2 Pro FS to present the correct voltage to program the auto-ranger. 6.7Vrms is 18.75dBu so I set the output of the RME to +19dBu 0dBFS level in the menu, and played a -0.25dB 1kHz tone to teach the Scaler.]

I. A look at the amplification structure & Auto-Ranging:

As noted above, this is not a typical digital volume control which might increase the amplification by regular 0.5dB steps across the dial. Rather, the 16 levels are derived from internal relays and have the following nonlinear gain structure (starting with "Level 0" for no amplification, unity gain):

Level 0 = 0dB        Level 1 = +7.0dB     Level 2 = +10.8dB

Level 3 = +13.5dB    Level 4 = +16.4dB    Level 5 = +17.9dB

Level 6 = +19.2dB    Level 7 = +20.3dB    Level 8 = +21.8dB

Level 9 = +22.6dB    Level 10 = +23.3dB   Level 11 = +24.0dB

Level 12 = +25.0dB   Level 13 = +25.6dB   Level 14 = +26.1dB

Level 15 = +26.7dB

To demonstrate how this looks like, we can run a 1kHz amplitude stepped sine measurement with the auto-ranger turned on to show the level transitions as we increase the output from -30dBFS to 0dBFS (using small 0.1dB steps), note that this is with output target set to 6.7Vrms, not the default 4.5Vrms:

More commonly, you'll see saw-tooth graphs like the following to show that the auto-ranger switches every few dBs to maintain the approximate target output:

I'm mindful of the -7dB dip around 3Vrms when measuring devices; this changes depending on the target voltage of course.

Assuming the Scaler doesn't add any noise or distortion, using the auto-ranger, we can leave the Cosmos ADC at a set level (like 4.5Vrms default, or 6.7Vrms as my preferred) and have the Scaler amplify the output of the test device to take advantage of the dynamic range afforded by the ADC at whatever output level.

You might be wondering: How fast is the auto-ranger switching?

I took a 1kHz tone and gradually reduced the output level to see how quickly the auto-range picked up the reduction and jumped to the next amplification level, recording this process so I can measure the time it took; here's what a simple single step change looked like:

That's fast. The device detected the threshold and made the switch in less than 0.1 seconds, this is fast enough that the distortion introduced by the switch will generally not affect something like a level-step measurement in REW. The only times I've seen anomalies are when the output level is right at the margins, triggering the auto-ranger switch with some hesitation. (The solution is to increase averaging so the software has a little more time to grab an accurate result.)

At the extreme, what happens if I go from silence to say a 5Vrms signal? This will basically ask the auto-ranger to go from a full +26.7dB amplification (Level 15) down to 0dB (Level 0):

1.3 seconds is the longest I've seen the Cosmos Scaler take to adjust. Ivan mentioned that his Audio Precision SYS-2522 can take 2-3 seconds.

Of course, all this isn't important if the Scaler adds noise or distortions to the signal! Which brings us to the next important section of this evaluation...

II. Noise level & distortion:

The easiest way to have a look at whether connecting the Scaler adds any extra noise is by examining the noise level measured from the Cosmos ADC either directly or with the Scaler feeding into the ADC:

Above, we see the dB(A) noise level and FFT from 15Hz to 192kHz for my Cosmos ADC (left channel, running in mono mode for lowest noise level) either alone or with the Scaler slapped on top with the home-hacked 6" 3.5mm-to-XLR cables attached.

These are excellent results. The Cosmos ADC uses the ESS ES9822Pro ADC chip which maintains a flat noise floor to 100kHz with only a small rise above that up to 192kHz when running at 384kHz sample rate (along with very low jitter). As you can see, the noise floor is clean with no evidence of any spurious tones introduced by the Scaler. The measured noise floor is basically unchanged with Scaler amplification set to 0dB; an insignificant -126.3dB(A) to -126.0dB(A). Even when we maximize the Scaler amplification to +26.7dB, A-weighted noise level only moved up 4.4dB. This suggests that the residual noise level of the Scaler is more than 20dB below what's measurable by the ADC without amplification, somewhere on the order of -145dB(A)!

While not shown, I did take a few more measurements of the noise floor at intermediate amplification levels between 0dB and +26.7dB. They all looked good so I won't bother showing them here. At no point did I see any unusual spurious tones creep in.

That's an excellent start. So let's see if we can confirm this with an actual DAC measurement to make sure the results look similar whether we use the Scaler or not.

Notice that even without gain added, there is a slight difference in things like the level of the 1kHz peak (-4.1dBFS ADC direct compared to -3.3dBFS through the Scaler), and some variation in the relative harmonic levels. These are likely related to the change in impedance (Cosmos ADC without Scaler 2.46kΩ, with Scaler 200kΩ - 6.7Vrms ADC input, balanced). Otherwise, we don't see any unusual excess noise added by the Scaler, the THD+N results calculated a little better with the Scaler.

For best measurement of THD+N as we approach -120dB and better, we should be using the Cosmos APU notch filter whether with the ADC alone or Scaler → ADC. Let's have a peek at results both ways:

Notice that with the APU, THD+N results look identical with the same relative harmonic levels. Both set-ups show that the Sabaj A20d 2022 DAC, based on the ESS ES9038Pro DAC chip is capable of a very impressive -124dB THD+N 1kHz @0dBFS. A couple details you'll see in the graphs are that the APU in/outs are unbalanced hence we see the small 60Hz hum picked up. Also, notice with this measurement I turned on the Scaler's auto-ranger function which pushed the input level up to -2.67dBFS as opposed to the ADC-direct measurement without the extra gain at -13.45dBFS. This is an example of the Scaler's unbalanced performance, still maintaining low noise level while applying amplification.

[Perhaps of interest, notice that I'm achieving -124dB THD+N with the Sabaj A20d 2022 here whereas when I published my review a few months ago, it was around -123.5dB. Not a big deal of course, and no, it's not because it's now "broken in" over the last few months. ;-) Rather, this is just a reflection of the fact that ambient temperature has dropped a few degrees between summer and autumn where I'm testing. Cooler temperatures get us measurably lower noise especially at this ultra-high level of resolution.]

If you're curious, here's what the test bench looked like when I chained the Cosmos APU → Scaler → ADC together; notice the lithium batteries used to power the APU and Scaler:

Before leaving this section, let me show you the Cosmos Scaler in action with Auto-Ranger turned on performing the REW level step with pure 1kHz sine (24/96) from -30dBFS to 0dBFS, in small 0.1dB steps using a couple DACs:

Notice the auto-ranger steps reflected in the noise floor plot.

By using the auto-ranger, we can get a good look at distortion across the highest 30dB output levels and see that the RME ADI-2 Pro FS is able to maintain better than -100dB THD+N (see cursor) from -18dBFS to 0dBFS. Let's compare this to the higher resolution Sabaj A20d 2022:

The Sabaj is able to maintain better than -100dB THD+N from -27dBFS to 0dBFS. And we can see in the graph that even without using the APU notch filter, the Sabaj can be seen to approach -120dB THD+N.

One other thing to notice as well is the nature of the harmonic levels comparing the AKM-based RME ADI-2 Pro FS and the ESS-based Sabaj. With the current-generation ESS DACs, the harmonics fluctuate a lot more than the AKM as previously discussed in greater detail. [Subjective audibility of this objective finding undetermined and believed to be insignificant due to the very low level nature of these harmonics.]

III. A few other things...

To round up the measurements for this preview, let's have a quick peek at how the Cosmos Scaler handles passing impulse responses (let's turn the Sabaj DAC level down, and use the +20.3dB Scaler volume setting so it's not just passing without amplification):

Looks totally normal. The rationale for doing this is to double check that a quick transient would not be distorted going through the device. Nothing unexpected here and basically other than the 16-level control mechanism, at its heart, the Scaler is a very low-noise analogue amplifier.

Since I've been using simple 1kHz sine waves thus far, let's look at the complex multi-tone FFT - here's the 1/10-Decade Multitone 32, with the Scaler volume set to +16.4dB, fed by the balanced Sabaj DAC XLR output for evaluation of noise and distortions:

Left channel picked up some low-level "grass"-like noise above 10kHz. Appears to be related to running DAC at the very high 384kHz sample rate (192kHz BW) and digital volume reduction. Not an issue with the Cosmos Scaler or ADC.

Nothing anomalous within 20Hz-20kHz and more than 120dB "clean" range, excellent especially considering +16.4dB of amplification was applied by the Scaler.

And we can also take a Triple-Tone TD+N reading with my typical tones at 48/960/5472Hz:

TD+N of -112.5dB is very good considering I dropped the volume of the Sabaj DAC down using the digital volume knob (at 73/99 on the display), while having the Scaler amplify the signal by almost +18dB. Previous testing shows that the Sabaj DAC achieves -118dB TD+N at 0dBFS.

The last thing I want to mention is the convenient "Auto-MONO" function. Basically, if you connect an input to the left Scaler channel, the signal will be duplicated to both outputs feeding the ADC. As discussed with the Cosmos ADC, if we change the volume level to 100% on the computer, the inputs will be processed in mono mode into the left ADC channel, achieving greater dynamic range.

To demonstrate, without using the APU, let's capture the 1kHz THD+N of the Sabaj DAC again in Stereo mode and then "Auto-MONO" to ADC left channel:

Stereo mode is convenient to capture both channels at the same time. Notice that in "Auto-MONO", the noise floor dropped 3dB and we achieve down to around THD+N -120dB which is again phenomenal resolution! Obviously, if you want the most accurate result for 1kHz THD+N/SINAD, use the Cosmos APU. (Needless to say, using APU + MONO mode doesn't add anything.)

Notice blue LED on the Cosmos ADC - picture taken when doing "Auto-MONO" measurement.

IV. Summary

While designed for the E1DA Cosmos ADC, the E1DA Cosmos Scaler works as an excellent front end to feed a measurement ADC with balanced or unbalanced signals while providing low-noise and low-distortion amplification along with a programmable auto-ranger function (target output 1-10Vrms). The input impedance of 200kΩ balanced and 100kΩ unbalanced is like Audio Precision measurement devices. As intended, the low 20Ω output impedance from the pre-amp makes it an excellent match with basically any ADC.

My results are certainly consistent with what I would expect of a device that's performing even quieter than the ADC without contributing its own distortion characteristics. Ivan's measurements of residual noise from the Scaler is down at 200nVrms(A), or about -147dB below a 4.5Vrms signal (consistent with my estimated noise level in Part II above, I could also use the APU's +60dB gain to measure the noise level).

E1DA has a nice collection of very high performance Cosmos devices now - great for audio hobbyists like myself looking at making measurements of products; even with THD+N lower than -120dB (SINAD >120dB) without difficulty. These days, when we see many DACs achieving this level of resolution, it's easy to lose touch of the realization that less than 0.0001% distortion and noise is a really really low number. Phenomenal work, Ivan!

Beyond the home hobbyists, this "trinity" of products with the Cosmos Scaler + APU + ADC would I believe also be very useful as high quality portable test kits for audio hardware designers looking to prototype devices when professional Audio Precision gear might not be available. Heck, given the affordable prices of these tools, I would hope every audio company, "High End" or otherwise, can do objective measurements to confirm high performance levels (or at least the intended level of performance) with new product designs even if they don't want to invest in AP machines! (This should also mean all companies must have objective data to show prospective buyers if they ask?)

In the 2020's, IMO, a company can still claim all kinds of things about how fantastic their products subjectively sound, but there is no reason to lack high-quality data to demonstrate the fidelity of the product, especially electronics gear like DACs, music players (CD, SACD players, streamers, etc...), amplifiers, and pre-amps.

As I mentioned at the start, this Cosmos Scaler I'm testing is a prototype from Ivan. I would not be surprised if some of what I'm discussing above changes if/when there is a final product release. For example, in one of our E-mails, there was discussion about fine tuning the amplification structure to see if that first step (Level 0 to 1) can be smaller, around 3-4dB rather than the current 6-7dB. I have no idea about the price point E1DA might want to aim at for the device. First things first, I suspect there needs to be a decision made as to whether this device will even be made available to the public, depending on the interest and demand out there as well as complications like supply chain hassles and prices of high quality components for the time being.

I think it's best to hop onto the E1DA Discord server to join the discussion and monitor if there are further up-to-date details if you're interested in the Scaler. I'll certainly be adding this tool to my measurement bench for use in future reviews as my test regimen evolves over time.

Yet again, fantastic work Ivan, on these affordable very high quality tools that in many ways perform at and even exceed way more expensive AP equipment for measurements when paired with excellent software available like REW! ;-)

--------------------

As for my music listening this past week, thanks to Gilles for suggesting Jean-Michel Jarre's newest work Oxymore (2022, CD DR7, binaural DR10) a couple weeks ago. Some very cool "surround" effects even in the CD release. The binaural headphone mix is notably more dynamic (can be heard on YouTube with compromised sound quality comparatively) and there's also the multichannel stream available on Apple Music (not sure about other services). Like with Revolver, I hope at some point we can get a lossless Bluray TrueHD-Atmos version rather than streamed Dolby Digital Plus (EAC3) multichannel lossy.

Jarre has previous works available in multichannel mixes. For example, check out his compilation AERO from 2004 as a good retrospective. Even though an older lossy DTS encode, it can be used as a nice demo comparing how much difference multichannel makes A/B'ing between the CD and 5.1! On my system, there's not just depth and envelopment of the music from the sides and behind, but also portions sound like they're "inside" my head as if listening with headphones. Very cool demonstration of the creative potential for artists, even at just 5.1 non-Atmos.

Supposedly, Oxymore is Jarre's first album fully conceived as a "surround" piece of work - here's an interview from c't, and the promo video:

--------------------

Finally, let's talk about Neil Young, yet again expressing opinions about digital (even hi-res digital) as if there's some horrific sound-destroying effect that happens the moment one (gasp!) digitizes audio in the promotion of his new album. Here's an interview the other day from Apple Music:

From what I can tell in that interview, it looks like the album was produced with at least 2 analogue ⇄ digital conversion steps if you buy it on vinyl (assuming they kept it analogue from final tape master to LP, regardless, the added distortions from tape to LP would be even greater), and at least 3 if you're listening to it digitally.

They recorded it to tape (even if for a "split second"), then immediately digitizes the audio, then after digital studio manipulation, "We mixed to analogue again after it went into the digital realm. We mixed it to analogue tape." according to Rick Rubin as per 5:40 in the video. Which also means there would have been another ADC step to go from tape back to digital for all the final streaming/download/CD releases.

 

Is this in any way special or desirable? Why does anyone care?

So what does going through the analogue steps get you? Check out some FFTs from the 24/192 hi-res download version of World Record (2022, DR9):

As you can see, probably more noise got introduced into the chain with the analogue steps. If we look at the quiet portion of "Walkin' On The Road (To The Future)", notice ultrasonic noise spikes at 29/54/57.5kHz. Notice that the noise floor is around -105dB across the audible frequencies (65k FFT) which isn't particularly good; proper hi-res albums typically will show levels down at -120dB or better.

The "This Old Planet (Changing Days)" average FFT of the whole song shows us there's a small amount of content out to near 40kHz. But if you watch the frequency analysis in real-time during playback, there's little actual music-correlated material above 25kHz.

To me, this 24/192 album should easily and safely be downsampled and dithered to 16/48 without hesitation as it's obviously not hi-res audio with such an elevated noise floor, containing anomalous ultrasonic tones (which we should filter out), and lack of actual content above 25kHz (assuming you want to keep ultrasonic content). It looks like they basically used the analogue tape as a low-level "noise" generator for the album which I guess is fine if that's what Young wants to do. It's just kind of silly to be talking about what's "real" music (as opposed to fake sounds I guess, as if there was ever an "absolute sound" anyways), and along the same lines, there's some weird idealization of "analogue" - shades of the hysterical "MoFiGate scandal" a few months back.

There's even a portion in the video talking about how quickly the tape sound changes (according to these guys within a year, see 8:30) and that's why it "degrades" fast and "you got to do a mix the night of", "it's always changing", and "it's alive" (Rubin). Come on guys, if master tapes are that prone to sonic change/degradation, then what's the point of music lovers getting excited about remasters using original tapes from 30+ years ago? This isn't even just anti-digital neurosis (like at 9:30 describing digital as "already stepped down"), this is some next-level "drama queen" stuff - featuring old men.

Maybe Young thinks that saying crazy things like this will result in positive hype to draw attention to him and the album? It comes across as very silly in the 2020's.

Oh yeah, of course we need to talk about vinyl starting around 42:45! Neil Young justifies the carbon footprint for LPs as "a dainty little peck on the f*cking Earth". That's some beautiful rationalization there, Neil. Do I sense at least even a little bit of hypocrisy while you're singing "Love Earth" and "This Old Planet (Changing Days)". Maybe he thinks he has offset enough carbon credits with that Lincvolt. ;-)

I trust that as educated members of the public, we know it's easy for artists to say all kinds of things, and sing sweet words in songs they want people to buy. When faced with a serious, simple, real-time question, apparently it's harder for Young to answer with a bit more sweetness, wisdom and grace. I would venture to say that the answer he gave to that question is probably more reflective of the heart of Neil Young, the man, rather than his idealistic lyrics. Interestingly, Rick Rubin doesn't collect LPs any more in the discussion at 49:40 for various reasons. Nice that he likes having "no stuff", as opposed to Young the "huge collector".

Discussions at 48:10 touches on "opinion vs. fact" and Young not liking the "back and forth" of social media. Well, when he can't really speak with facts about even the audio technology he works with, are we sure Mr. Young knows that difference when it comes to other domains? When it comes to "facts", he has created a lot of nonsense and confusion over the years among hi-fi enthusiasts so he has to take some responsibility I think for some of the unhappiness on social media. As for Rick Rubin, he produced Californication (1999, DR4) and Death Magnetic (2008, DR3), two of the nastiest, loudest, compressed popular albums of all time - so even with him, I think we audiophiles need to be a little cautious about his standards for sound quality and hope that he has changed over the years. At least DR9 for World Record is decent.

I noticed Young/Rubin mused about starting an AM radio station and playing "vinyl quality" over the air. Sure, go for it... Other than "lo-fi" curiosity, or in a sentimental mood thinking about the mid-20th Century, I doubt many high-fidelity audiophiles would be interested. I wonder how Young feels about multichannel and Spatial Audio remixing of his old material. Harvest was available in 5.1 as a DVD-A; it has been awhile so I'll need to have another listen to check if it was done well.

"Old Man".

Anyhow, after all these years, coming across Young's claims about audio technology based on fanciful beliefs (or claims he hears), whether it be stuff like this or back in the Pono era, I guess it's fun showing just how distorted ideas about audio can become, projected now to the general music consumer through interviews from Apple Music. I hope we audiophiles are not like general music consumers. ;-) It's well past the time for audiophiles to be smarter about understanding hi-fi audio, making sure we speak in truth far beyond Neil Young ever seems to be capable of. This should not be hard based on what's on display in the interview. Well, at least Young's against MQA, so that's alright.

Hopefully I'll get some time in the next week to listen to all of World Record (rating on Metacritic looks alright). As usual, I have respect for Young as an artist and appreciate his decades of creativity even if I believe most of what he says about hi-fi / hi-res / analogue vs. digital is extremely bogus.

With that, ladies and gentlemen audiophiles, I wish you and yours many wonderful music listening sessions!