Review and Measurements of the PS Audio Stellar Gain Cell DAC

[SIY]

And, please folks, I'm tired of arguing over expectation bias because, as I have explained....

"...I have no idea how it works and I'll keep stomping my foot anyway."

 

[MarcT]

"...I have no idea how it works and I'll keep stomping my foot anyway."

Yeah, I figured you'd have to chime in.

 

[Killingbeans]

And it seemed like the Stellar Gain sounded better with higher res material than with 16/44.1 tracks, which I understand from ASR is not supposed to happen.

This might be the culprit?

 

[digitalfrost]

The filter is the same at other samples rates, just higher. I would say above 48khz, filter topology doesn't really matter from an audible point of view. The only reason people fuss about these is because at 44.1 it's so close so 20khz.

 

[Meagan_02]

This is a review and detailed measurements of the PS Audio Stellar Gain Cell DAC. It is on kind loan from someone who read my review of PS Audio DirectStream DAC and wanted to know how this model of PS Audio DAC performs. The Stellar Gain Cell costs US $1,699 which as high-end audio DACs go is a "bit of a bargain."

The industrial design is good:

View attachment 35199​

There is quite a lot of heft to the unit. Not a great fan of blue OLED display with tiny font for sample rate and filtering. Then again, I take this over blinking/color LEDs you are meant to interpret yourself on some other units. Volume control and associated display value is responsive which I appreciated.

The back panel shows the true value of this unit:

View attachment 35201​

In addition to usual set of digital inputs we also have a set of analog inputs! For people like me who have Reel to Reel deck and others who have turntables, being able to integrate everything into the DAC and thereby eliminating an external pre-amp, this is great!

Inclusion of an analog pre-amp and switcher presents a problem for volume control though. Standard way of using the volume control built into the DAC chip (ESS in this case) doesn't obviously work for analog inputs. So companies resort to an external analog variable gain output stage. This brings is up to the name of this product, "Gain Cell."

The Gain Cell is the PS Audio name for a differential mixer circuit called Gilbert Cell. Using 6 transistors in a clever differential configuration, you can either mix two analog signals together, or use DC on one pair of inputs and control the gain of the other. The latter is what is used here. By controlling that DC unit using the front-panel microprocessor, you can now have control of the attenuation including the ability to have a remote control.

PS Audio contrasts this with using a potentiometer which could have longevity issues. That is true, but PS audio goes further, saying this scheme puts nothing in the signal path and hence is "pure" in execution:

View attachment 35202

Well, if anyone is able to test for "zero-loss" and "pure analog signal path is us." With instrumentation it is easy to see if the paper concept of Gilbert cell is realized in practice. While Gilbert cell is used in IC designs to good effect, there, matching of the transistors is easy. Not so in a discrete design as is used here. The transistors used to control the gain have their own non-linearities so the concept that there is no loss is quite dubious. But again, measurements will tell us all so let's get into that.

Pre-amplifier Audio Measurements
I fed a set of XLR signals to the Stellar Gain Cell from my Audio Precision APx555's analog output which has a SINAD well above 120 dB, with distortion products at vanishingly low levels. The signal is at 1 kHz so an ideal, zero-loss device under test would just produce a single sharp line and nothing else. This is what the Stellar Gain Cell (SGC) produces instead:
View attachment 35203

You are seeing what I am seeing? PS Audio, what the heck have you done to my pristine input signal?

Starting with the worst offenders, second and third harmonic distortions are way up at near 70 dB, resulting in SINAD (signal over noise and distortion) of just 68 to 71 dB. The distortion is asymmetrical between the channels showing lack of precision that way too.

But look at what else is going on. The button of our 1 kHz tone has been broadened. This indicates noise modulation of our signal much like can happen with DACs and jitter. Gilbert cell is also a modulator so any noise fed to its second input will modulate the primary (music) signal giving us what we are seeing. Not good.

Likewise, any power supply noise and spikes will also modulate our main signal creating sidebands on each side of our desired signal, creating that spray of spikes all around our 1 kHz tone we fed to the unit.

Put simply, we fed the Stellar Gain Cell one tone, and got dozens and dozens of unwanted tones and distortion product. The modulated ones are by definition not harmonic in nature so let's not have the argument that these are "good distortions."

At any rate, no way, no how this pre-amp and its core circuit can be called "zero loss" or "pure." There is nothing pure about it. Imperfect and non-matching transistors are in the circuit fed by noise and distortion that modulate them. The cure for potentiometer is far, far worse than the disease.

There are voltage controlled amplifier ICs used in Audio/Video Receivers and they perform far better than this.

We can test the non-linearities by feeding the unit dual tones: one low frequency and one high and compute the "intermodulation" distortion relative to the level of input signal:
View attachment 35204

Up to about 0.3 volt input, the Stellar Gain Cell is dominated by high level of noise. After that, distortion starts to take over. At full output level, one channel goes nuts far worse than the other (light green). At the nominal 4 volt output used in the dashboard and most of these tests, that doesn't happen but we suffer from higher levels of distortion than full output! This is backward of what typically happens in amplifiers.

Changing our tones to 19 and 20 kHz, we get high intermodulation distortion there as well:

View attachment 35205

Only the two tall spikes are the signal tones fed to the unit. Everything else is "non-pure" contributions from the Gain Cell giving us a usable distortion-free range of 70 dB (SINAD) which agrees with our dashboard.

Fortunately in some other areas things are not as broken. Frequency response is very flat and good:

View attachment 35206

Dynamic range is good too especially at full output although I am not sure how often you would use that:

View attachment 35207

Notice that both here and in our THD+N in the dashboard we are getting almost the same specs PS Audio advertises. This means our testing very much matches theirs and they are well aware of the poor measured performance of this unit.

Crosstalk audibly is fine as well:
View attachment 35208

DAC Audio Measurements
I was surprised to find out that the PS Audio chose a very low-end offering from ESS to use in this DAC: the ES9010K2M mobile DAC:

View attachment 35209

THD+N translates into a SINAD of 106 dB which is way, way better than the pre-amplifier in the PS Audio Stellar Gain Cell. So I thought we would get the same dashboard view as the pre-amplifier. I was wrong:

View attachment 35210

What? What??? How did the digital path cost us another 10 or more dB of distortion? What have you done PS Audio?

This is nothing short of disaster, placing the SGC DAC third from the bottom of over 175 DACs tested:
View attachment 35211

The SGC DAC uses an FPGA to supposedly clean up jitter. It is a wasted effort:

View attachment 35212

Noise floor is very high (in best DACs it goes under the AudioScienceReview tag). And we see countless sources of noise, jitter and unwanted spurious tones. No doubt many of these are contributed by the output gain stage but why bother with any digital jitter clean up when you are analog stage sprays a good bit on it?

Linearity is poor graphically:
View attachment 35213

But looking at the spectrum as the test was running showed that the actual tones produced by the DAC were fine but high amplitude spike at 180 Hz shown in FFTs was interfering with it. Linearity test uses a very sharp bandpass filter to exclude almost all noise and distortions but is wide enough to let the 180 Hz in. In other words, the little ESS DAC is trying to do right but the actual implementation steps on its output.

Multitone test simulate "music" with 32 tones shows the problem with this DAC vividly:

View attachment 35214

Since music has even more tones than 32, this means that no detail below -70 dB or so will be resolvable. Converting that to bits, you have 12 of them. Yes, 12 bits. Anyone who thinks they are hearing new detail they had not heard before with this DAC, should know that it is not the DAC that is doing that. It is your brain listening better than before and hearing detail that is well above -70 dB. Don't confuse perception with what your ears pick up.

There are only 3 filters for the DAC unlike the usual 7 or so:
View attachment 35215

The F1 filter is basically no filter. F2 and F3 are similar. F3 was recommended for best measurements and that is what I used for my testing. And would be my choice if I were to use this DAC.

The one benefit of digitally controlling gain is that channel matching should be good and is:

View attachment 35216

Listening Tests
I have to run out to our Audio Society meeting in half hour so no time for this. Will perform tests and update the review later. Owner was anxious to see the measurements so here they are.

Conclusions
The words on PS Audio website for Stellar Gain Cell DAC paint a super seductive picture of this DAC. The words "pure," "zero loss," etc. are music to everyone's ears, mine included. Going into this review I didn't think it would be exciting. Due to use of ESS DAC I thought the performance would be good. Nothing prepared me for the disaster that unfolded in front of me in test after test.

This DAC and pre-amp spray your beautiful recordings with a ton of noise, and distortion caused by non-linearities. The pre-amp/output stage is so bad that it gives you digital-like distortions while using a completely analog stage!

Then we have a DAC stage whose performance has been degraded by whopping 50 dB! Engineers kill themselves to get 3 dB improvement in noise and distortion at times. Yet, PS Audio destroys the purity of what comes out the ESS DAC by incredible 50 dB. Is there any hope that companies respec the word "high-fidelity" anymore?

I hear it: "but it sounds great to my ears." I am sure it does because a) audiophiles are terrible when it comes to critical listening abilities and hearing non-linearities and b) you don't know how to do proper listening tests.

So no, don't help bastardize this hobby any more than it already is by paying a fortune for a DAC and pre-amp which destroy any hope of fidelity to what the talent and engineers produced for us.

My head is down on behalf of the industry I participate in. Needless to say, I cannot recommend the PS Audio Stellar Gain Cell in any form or fashion. There is nothing that is stellar about its measured performance.

------------
As always, questions, comments, recommendations, etc. are welcome.

Please consider supporting reviews like this (I have to pay to ship this back) by donating using: https://www.audiosciencereview.com/forum/index.php?threads/how-to-support-audio-science-review.8150/

Every time I read this review, and many others by Amirm, I'm astounded by two things: 1) His unearned arrogance, and 2) His shocking ignorance. He worships at the altar of his almightly Audio Precision dowhacky as if measurements are all that matters in the world of audio. His disdain for audiophiles is obvious, most likely because he has no understanding of what makes for great audio, much less how to listen. He hears with his eyes only. He's the L. Ron Hubbard of audio. Simply put, he is an engineer, not an audiophile. The magic of audio escapes him entirely.

 

[MaxBuck]

Every time I read this review, and many others by Amirm, I'm astounded by two things: 1) His unearned arrogance, and 2) His shocking ignorance. He worships at the altar of his almightly Audio Precision dowhacky as if measurements are all that matters in the world of audio. His disdain for audiophiles is obvious, most likely because he has no understanding of what makes for great audio, much less how to listen. He hears with his eyes only. He's the L. Ron Hubbard of audio. Simply put, he is an engineer, not an audiophile. The magic of audio escapes him entirely.

Magic is fine, but remember that magic is nothing more than illusion.

 

[Jimbob54]

Every time I read this review, and many others by Amirm, I'm astounded by two things: 1) His unearned arrogance, and 2) His shocking ignorance. He worships at the altar of his almightly Audio Precision dowhacky as if measurements are all that matters in the world of audio. His disdain for audiophiles is obvious, most likely because he has no understanding of what makes for great audio, much less how to listen. He hears with his eyes only. He's the L. Ron Hubbard of audio. Simply put, he is an engineer, not an audiophile. The magic of audio escapes him entirely.

I'll bite. What does make for great audio then?

 

[BDWoody]

I'll bite. What does make for great audio then?

You bored this weekend?

Try to avoid feeding the trolls...

 

[Jimbob54]

You bored this weekend?

Try to avoid feeding the trolls...

I've got a sweep running on the answer though.

 

[SIY]

I'll bite. What does make for great audio then?

Paying a premium for ****** performance so Paul can buy a newer BMW.

edit: wow, we have a new auto-censor!

 

[Jimbob54]

Paying a premium for ****** performance so Paul can buy a newer BMW.

edit: wow, we have a new auto-censor!

Well fuck me pink!

Edit. Not a very good one.

 

[SIY]

Well fuck me pink!

Edit. Not a very good one.

Needs adjustment.

 

[Bruce Morgen]

1. Calls an instrument that is relied upon by pretty much every audio manufacturer with the budget to afford one a "dowhacky." Accuses an expert -- who's also a lifelong audiophile and well known as the "golden eared" subjective listener at two companies that have been among the major influences on the audio state of the art for multiple decades -- of having "no understanding."

2. Apparently believes in "magic."

Enough said...

 

[Purité Audio]

Bugger
Bollox
Wank
Who is in charge of the auto-sensor settings, we need to know.

Keith

 

[Jimbob54]

Bugger
Bollox
Wank
Who is in charge of the auto-sensor settings, we need to know.

Keith

Maybe It wax set up by Thomas and only just launched? Such lingo would not register on that ****'s radar.

 

[Purité Audio]

Yes far too tame, I went to Sunday school so have an extremely limited expletive library.
Keith

 

[Bruce Morgen]

Magic is fine, but remember that magic is nothing more than illusion.

...and so is marketing.

 

[Killingbeans]

Magic is fine, but remember that magic is nothing more than illusion.

Exactly, and I can't imagine a magician just making random changes to his props until a trick emerges.

Don't see why audio reproduction should be any different.

 

[Bruce Morgen]

He's the L. Ron Hubbard of audio.

That aptly describes the amiable, grandfatherly Paul McGowan -- who's currently celebrating the arrival of his new ($30k/pair!) speakers....from China.

 

[ahofer]

Every time I read this review, and many others by Amirm, I'm astounded by two things: 1) His unearned arrogance, and 2) His shocking ignorance. He worships at the altar of his almightly Audio Precision dowhacky as if measurements are all that matters in the world of audio. His disdain for audiophiles is obvious, most likely because he has no understanding of what makes for great audio, much less how to listen. He hears with his eyes only. He's the L. Ron Hubbard of audio. Simply put, he is an engineer, not an audiophile. The magic of audio escapes him entirely.

Thats two ignores in 24 hours. And this one might be Paul McGowan.