Master Thread: Are measurements Everything or Nothing?

[manisandher]

@Thorsten Loesch , really great to have you here.

FWIW, I'm finding that DeltaWave's 'PK Metric' correlates closely with what I hear. Particularly, the DACs I have here with the lowest THD+N don't measure best in PK Metric.

(More about PK Metric here: https://deltaw.org/pk_metric.html)

Mani.

 

[Holmz]

Please define "real 32 Bit of resolution".

Ignoring the steps in between the DAC chip and the analogue output, it is whether the chip has discrete levels that represent each of the 2^32 or 2^31 levels…
It is whether the DAC choir is 32 bit… but its spec sheet is looking like it is 24bit DAC chip.

Maybe you consider this semantics. And maybe it is.
It seems like it is a good unit, so I was mostly intrigued as I had not seen 32 bit DAC chips… but then again I live under a rock.

It either is the ability to receive a 32 Bit word and output the converted analogue equivalent at however many BIT ENOB are available.

Or it is a DAC with an ENOB of 32 Bit which is impossible, just for now we do not see DAC's with 24 Bit ENOB.

Or is your definition of "real 32 Bit of resolution" not actually the physical capabilities of the DAC itself, but the way the Chip manufacturer markets the DAC?

Thor

I agree that one cannot hit the 190ish dB, that 32 bits would yield in theory… it is whether the DAC chip is 24 or 32 bits.

I am sure a true 32 bit DAC chip would likely perform the same as the 24 bit chip in most current applications. And I would further agree that it likely doesn’t matter whether it is 120dB, 190 dB or 1000 dB… we are well past the point of hearing much nuance.

…
After reversing the distortion cancellation tweaks, the extra loop gain stage by stage and scaling the circuit back to something a little more ordinary it had better than -110dB THD & N (110dB SINAD) and it sounded much better. In fact after "making it way worse" (30dB more H2/H3 distortion) it started to actually be preferred by listeners to the other options.
…

Thor

Now ^that^ is interesting… and echos what @atmasphere has mentioned numerous times, and what many people that like tubes are likely finding is the attractive nature of them.
It also somewhat throws a wrench into the spokes of measurements of SINAD, without some look at the harmonics, Is likely not telling as much of a story of how it might sound… especially if the 110dB is higher harmonics, and not noise and H2/H3.

Was that DAC pushed directly out to headphones or amp+speakers?
And if it was run through a tube preamp with some H2/H3, should we assume that the earlier version with the higher SINAD would also sound better?

.,.
We have a DAC with the ability to receive 768kHz/32Bit Samples.
…
An FPGA is used to turn the 32Bit PCM into 45/49MHz Delay Sigma modulated 1Bit audio.
…
TI never updated their datasheet's and re-charsterised and relaunched / readvertised the part.
…

wait… ! !
Now I get it… it is not a like a DAC with single sample and hold doing the A-D conversion on the entire 24 or 32 bits…

 

[Holmz]

.,.
We have a DAC with the ability to receive 768kHz/32Bit Samples.
…
An FPGA is used to turn the 32Bit PCM into 45/49MHz Delay Sigma modulated 1Bit audio.
…
TI never updated their datasheet's and re-charsterised and relaunched / readvertised the part.
…

Now I get it… it is not a like a DAC with single sample and hold doing the A-D conversion on the entire 24 or 32 bits…

 

[Thorsten Loesch]

@Thorsten Loesch , really great to have you here.

FWIW, I'm finding that DeltaWave's 'PK Metric' correlates closely with what I hear. Particularly, the DACs I have here with the lowest THD+N don't measure best in PK Metric.

(More about PK Metric here: https://deltaw.org/pk_metric.html)

Mani.

This looks interesting.

Every few years we see someone criticism g traditional test methods for audio, proper a new well founded and researched new method that correlates better with listening experience and is ignored.

Kind of like what Feynman commented on rat running experiments in his 1974 CalTech commencement speech, otherwise known as "Cargo Cult Science".
https://calteches.library.caltech.edu/51/2/CargoCult.htm

Thor

 

[Thorsten Loesch]

it is whether the chip has discrete levels that represent each of the 2^32 or 2^31 levels…

Then the definition is 194dB SINAD @ 0dBFS for a 32 Bit DAC.

And there will never be such a thing, nor will there likely ever be a 24 Bit DAC by this definition.

So all the debate is semantics and pointless in reality.

We should judge DAC's strictly by ENOB and is the ENOB is sufficiently to handle music without compromise.

Anything else is just mater debating and cunning linguism.

Thor

 

[Thorsten Loesch]

Now I get it… it is not a like a DAC with single sample and hold doing the A-D conversion on the entire 24 or 32 bits…

No, it is irrelevant what precisely happens between the Data input and the analogue output of the system that comprises the DAC.

Nowadays multibit DAC's (that is for example R2R ladders with N taps where N is the number of bit's) and that apply no digital filtering are functionally extinct.

ALL DAC's use any amount of digital processing and a mix of Delta Sigma and Multibit (again, detailed implementations diverge widely past that) to convert whatever input theuy receive to whatever output they produce.

I went into this in some shallow "popular science" level detail in a pirce written for defunked"AudioStreams":

https://web.archive.org/web/2016030...loesch-amrifi-audiostream-addendum-pcm-vs-dsd

The general class of TI's "advanced segment DAC" has the least amount of such manipulation of any DAC Chip available off the shelf (by bypassing the Digital filter and directly inputting signals into the core) and this one of the reasons I like to use them.

Wolfson Micro used to have some interesting tech, but that mostly was lost/dumbed down after they went bust betting on digital headphones too early and were taken over by Cirrus Logic.

Thor

 

[solderdude]

After reversing the distortion cancellation tweaks, the extra loop gain stage by stage and scaling the circuit back to something a little more ordinary it had better than -110dB THD & N (110dB SINAD) and it sounded much better. In fact after "making it way worse" (30dB more H2/H3 distortion) it started to actually be preferred by listeners to the other options.

What would be the reason for preference for inaudible amounts of distortion and inaudible amounts of distortion if that is the only difference ?

What ENOB did you get from a 140 SINAD DAC and what was the maximum output voltage of that DAC ?

 

[pma]

Please define "real 32 Bit of resolution".

Maybe computational resolution .

Hi Thorsten, haven't seen you in the web space for long.

 

[manisandher]

This looks interesting.

Every few years we see someone criticism g traditional test methods for audio, proper a new well founded and researched new method that correlates better with listening experience and is ignored.

Kind of like what Feynman commented on rat running experiments in his 1974 CalTech commencement speech, otherwise known as "Cargo Cult Science".
https://calteches.library.caltech.edu/51/2/CargoCult.htm

Thor

Yes, I tend to love all things Feynman. I'm currently really enjoying McGilchrist's 'The Matter With Things'.

And your earlier Vonnegut link was a gem.

Mani.

 

[Thorsten Loesch]

What would be the reason for preference for inaudible amounts of distortion and inaudible amounts of distortion if that is the only difference ?

What ENOB did you get from a 140 SINAD DAC and what was the maximum output voltage of that DAC ?

First, it was not a DAC, but a Headphone Amplifier, as remarked elsewhere, the iFi iCAN Pro.

Secondly, clearly the differences that were able to be confirmed using traditional instrumentation and traditional measurements, e.g. using an AP System 2322, were ones that should be reliably inaudible. And no, the classic dismissive line of "well, some people just like distortion" cannot be used if we are debating -140dB vs. -110dB.

I'd also like to know what is the underlying cause to cause a polarised and reliable in blind conditions reaction among many listeners.

What I would posit is that there clearly a further factor at play not covered by traditional measurements and specifications.

Some previously discovered ones like TIM/TID/SID etc. can be pretty much excluded as the circuitry is high current discrete and current feedback, which imparts very high slew rates etc. plus the use of bandwidth limiting filters ahead of active circuitry to ensure the Amplifier works in the most linear range. Such a system is highly resistant to TIM/TID/SID etc.

If you have a proposal as to what factor is at play, I'm all ears.

FWIW, the tweaks concerned specifically two areas, one a compensation scheme that while delivering identical stability to the conventional one allowed greater loop gain at high frequencies. Tweaks that increased low frequency loop gain by around 20dB. Tweaks that deliberately used predistortion in previous stages to compensate distortion in later stages.

With all these tweaks pulled out, it did not change too much of the basic design but things like cascoding of stages and current sources to increase open loop gain and specific tuning of degeneration and load for harmonic cancellation between stages.

The subjective improvement was about equal for removing interstage cancellation, using conventional compensation and reducing open loop gain and were approximately compounding in subjective impact.

Thor

 

[solderdude]

If you have a proposal as to what factor is at play, I'm all ears.

That's what I was asking you.

IMO finally something that might be really interesting to look at but not researched further.
Assumed inaudible vs also assumed inaudible but some specific testing and how rigid ? Are there numbers (amount of test subjects and listening tests I mean) ?
Scientifically sound testing ?
Or was it some more informal testing ?
Any nulling that showed any differences ?
Could it be load interaction with feedback loops ?

Genuinely curious. I would assume this would be interesting.
F.w.i.w. I am on the same page as you are on what you previously posted.

frohe Weihnachten übrigens

 

[Thorsten Loesch]

That's what I was asking you.

IMO finally something that might be really interesting to look at but not researched further.
Assumed inaudible vs also assumed inaudible but some specific testing and how rigid ? Are there numbers (amount of test subjects and listening tests I mean) ?
Scientifically sound testing ?
Or was it some more informal testing ?
Any nulling that showed any differences ?
Could it be load interaction with feedback loops ?

Genuinely curious. I would assume this would be interesting.
F.w.i.w. I am on the same page as you are on what you previously posted.

frohe Weihnachten übrigens

Many questions.

Objective testing with AP2 (Sys2322) fully loaded, plus 500MHz Scope with high speed signal generator, 5GHz RF spectrum Analyzer with tracking generator, LISN'S etc and shielded test chamber. Pretty serious test kit if you ask me and many tested outside traditional audio.

Listening as pointed out earlier, blind, but using preference ratings. Additionally also a lot sighted listening.

The final listening tests with different levels of modification were systematic, level matched, and blind.

Test subjects, may be dozen in the formal tests, many more on informal testing on preview units supplied to retailers and distributors.

Inaudibility of the harmonic distortion differences is assumed, based on extant scientific work and the general assumption that if listening at less than 110dB SPL and with HD at -110dB or lower audility should be eliminated as the levels are below hearing thresholds.

Could be load/loop feedback interactions. Maybe, but how? Output stage has 0.5 Ohm open loop output impedance, proper build out networks etc.so I am inclined to suggest "no".

One thing I have observed before on audio gear, aggressively minimising THD / HD in a circuit beyond what "wants to do naturally" has this tendency to cause a stressed/boring sound.

Like that time I made a zero feedback 300B SE amplifier that had 0.05% THD @ 1Watt (driving 97dB/2.83V 16 Ohm 15"Coax monitors) instead of 1% THD for a conventional design.

Without knowing details, everyone who ever listened to the system preferred the straight 6SL7 Mu-Follower to 300B to my "Low THD" 300B amp. And yes, level matched comparisons, amplifiers build on the same chassis and looking pretty much identical unless you opened them up.

I'd love for someone to tell why without arguing "well all your friends just like distortion".

Frohes Fest, guten Rutsch, Hals und Beinbruch.

Thor

 

[svtcontour]

For me I think it depends on if it the measurement is to the level when it becomes audible or not. Like will amp with harmonic distortion of 0.1% vs 0.001% be audible? I'm debating not given that a loudspeaker's distortion figures are much greater so I take measurements as a guide but I dont base my purchase solely on great figures. Warranty, service, aesthetics may play a greater role for me once the specs are decent enough for me to deem it inaudible at least to me.

 

[solderdude]

Many questions.

Have some more...

We all know recordings usually have more distortion than reproduction gear. Many cheap op-amps are commonly used with not so great PCB layout, imperfect anti-alias filters before the ADC, signal processing, instruments with distortion by themselves, compression, limiters and whatnot.

When this is reproduced near perfect (say -140dB noise and THD) and that sounds boring does that mean the distortion in recordings is boring and becomes less boring when an inaudible amount of extra distortion is added ?
Is it the distortion spectrum that is important even at those (inaudible) levels or the lack of noise ?
Have you ever done some tests when you found that or just took it for what it is and decided chasing lowest noise/distortion is not beneficial anyway.

Listening as pointed out earlier, blind, but using preference ratings. Additionally also a lot sighted listening.

Then I would suggest leave the sighted out of the results which narrows down the testing.
What was compared exactly 2 similar designs with only 1 difference or compared to a certain reference amp ?
How many listeners with what headphones and what music ?
How many attempts per listener ?
AB or ABX (in case of blind) ?
Or were all listening tests kind of informal and only about preference with a minimal amount of switching ?

The final listening tests with different levels of modification were systematic, level matched, and blind.

Were there records made of these tests. It seems to me this would yield very important info.
Certainly in the light of the mentioned tube amps, with their many forms of distortion, that would make the -110 vs -140dB something that could end all debates and invalidate the reasons for forums like this. (SINAD chasers) where beyond a certain point (my opinion) the benefits are only numerical and not of any practical concern.
What else was different other than technical performance (noise + THD) ? It seems like fundamental design differences between the topologies but maybe not in the output stage.
Was it the distortion component, noise component or the combination that was responsible for the 'boring-ness' or was that not specifically tested for ?
What was the maximum output level the -140dB was referenced to ?
Even with 20V (which is quite loud in headphones) 2uVrms noise is quite a feat and when considering headphones with say 100dB/V sensitivity that be 13dB below the audible threshold which is impossible to reach with a headphone as self noise of blood would be many, many dB's above that level.

It is really interesting (to me) and if I ever encountered such a phenomenon I would pursue it till funds dried out.
I can understand that for a business this would be really not that interesting though and making a profit and getting lots of sales is of more importance.

Test subjects, may be dozen in the formal tests, many more on informal testing on preview units supplied to retailers and distributors.

We can safely discard the retailers and distributors opinions. From a business standpoint that feedback (regardless if obtained in a scientifically sound manner or not) is far more valuable than some academic style research.

Inaudibility of the harmonic distortion differences is assumed, based on extant scientific work and the general assumption that if listening at less than 110dB SPL and with HD at -110dB or lower audility should be eliminated as the levels are below hearing thresholds.

Yep, my thoughts as well + I have not met any people that have a dynamic range, during music listening, exceeding 70dB and do not know of recordings with a low enough noise floor.

Could be load/loop feedback interactions. Maybe, but how? Output stage has 0.5 Ohm open loop output impedance, proper build out networks etc.so I am inclined to suggest "no".

That would be my question. An unlikely possibility would be RF coupling and demodulation in the amp but we can safely assume this is not the case seeing LISN's and EMC gear was used to check performance/immunity in the RF range.

One thing I have observed before on audio gear, aggressively minimising THD / HD in a circuit beyond what "wants to do naturally" has this tendency to cause a stressed/boring sound.

Is this also the case if one uses a very high open loop gain (say nested feedback) which is highly effective in minimizing THD and noise but the feedback of course is 'natural' (just resistors and Miller caps).

Without knowing details, everyone who ever listened to the system preferred the straight 6SL7 Mu-Follower to 300B to my "Low THD" 300B amp. And yes, level matched comparisons, amplifiers build on the same chassis and looking pretty much identical unless you opened them up.

I'd love for someone to tell why without arguing "well all your friends just like distortion".

I wasn't there so could only guess as to what the reasons were. Preference is preference though 0.1% or 0.00001% isn't that an important difference depending on the recording, used transducers and playback level.

The 'becomes boring' is interesting though. I usually relate that to recordings and transducers, time of day and mood mostly. Hardly ever to gear.
Admittedly I never played with any amps better than -110 THD+N (certainly not loaded) so have no reference frame there.

Frohes Fest, guten Rutsch, Hals und Beinbruch.

Thor

Danke... und ein gutees neues Jahr

 

[Thorsten Loesch]

For me I think it depends on if it the measurement is to the level when it becomes audible or not. Like will amp with harmonic distortion of 0.1% vs 0.001% be audible? I'm debating not given that a loudspeaker's distortion figures are much greater so I take measurements as a guide but I dont base my purchase solely on great figures. Warranty, service, aesthetics may play a greater role for me once the specs are decent enough for me to deem it inaudible at least to me.

Ok, I think it's time give a few reality check numbers.

Let's take a really well designed Loudspeaker of recent vintage measured by the NRC for Soundstage:

NRC Measurements: Paradigm Founder Series 100F Loudspeakers

Paradigm Founder Series 100F Loudspeakers Review

This a well designed and fairly priced in my view (5,200 USD) large floor standing tower speaker. It measures in term of distortion and dynamic compression much better than average. In fact, I'd venture that this speaker is pretty close to SOTA at this point outside of large format studio monitors.

To set a baseline I would add the THX replay level recommendation of 85dB SPL at the listening position with -20dBFS, or 105dB SPL with 0dBFS.

Much music these days has a crest factor (peak to average ratio) much below 20dB, so using a "typical" Music signal (see AES Standard AES2-2012 for background) we are looking at 92dB average playback level. In my youth "average playback level" was known as 0dB VU (usually around -14dB of permissible peaks). And yes - East and West German Radio & TV studios used 90dB/0dBVU replay levels and featured a frequency compensated level control for those who desired lower levels.

So overall the 92dB/-12dBFS figure matches studio practice from my studio days and the THX playback level recommendations for THX certified Movie Theaters, I think it is a reasonable for listening LOUD but not DEAFENING LOUD, or perhaps (based on my measurements in purely acousticalconcerts of classical music what we may call "realistic" for center row 10.

Let's set a 3m listening distance and assume a pair of speakers. Increasing distance from the speaker from 1m to 3m we loose around 8-9dB SPL. Using two speakers with correlated (mono) signal adds 6dB. So for at 3m we need 108dB peak SPL from each speaker at 0dBFS to reach such playback levels and it would give the average SPL of 92dB for modern music with 95dB from each speaker.

All these notes are simply to "keep it real". This is what I would recommend in the real world as requirements for a "high quality music playback system".

The NRC and Soundstage sadly only test at 95dB SPL, which represents ~92dB at the listening position for a speaker pair and matches our -12dBFS SPL, but not higher. We can however estimate that 12dB greater SPL will increase the THD of a Speaker with H2 dominat by around 6dB IF the speaker still can perform within the linear range.

Let's have a look what the Speaker I lauded as close to the "State of the art" provides:

Ok, if we take the section between 200Hz-500Hz, there we have ~43dB SINAD, at 100Hz and 10kHz we have < 40dB SINAD. As we do not see anything below 45dB SINAD we must guess, from experience I would expect 50-55dB SINAD maximum in the midband (500Hz-5kHz) marred by a few spikes.

Note that is with a -12dB signal, a 0dB signal would further reduce SINAD by 6dB or so.

How about dynamic compression?

So going from 70dB SPL to 95dB causes an across the board compression of around 1dB.

Now by Speaker standards this great engineering and very, very solid performance.

So this what a good Speaker can do.

How about headphones? Amir likes the Dan Clark ones as do I. What do we get?

So 81dB SINAD is amazing... In a headphone.

How about something a bit more down to earth, like a Sennheiser HD600?

Not stated, but at 1kHz we get around 65dB SINAD, 75dB at 10kHz and 45dB at 100Hz.

Now I NEVER EVER heard anyone call the HD-600 being heavily distorted. Bass-shy, with a bit of midrange prominence and a bit of a lack of HF, yes, but heavily distorted? Nope.

So this gives a little context.

If headphones or speakers with 40dB...60dB...80dB SINAD do not sound distorted, is there any meaning to SINAD (or indeed THD&N) for subjectively audible distortion, beyond (perhaps) stating that better -60dB at rated power with 0dB FS will likely be good enough and if we can get 90dB we are likely well past "godd or bad".?

And if this is the answer, which I propose it is from the evidence presented, why does anyone accept THD&N / SINAD as a measure of quality, goodness and perhaps a reason for a purchasing decision, except in Speakers or Headphones?

Thor

 

[Thorsten Loesch]

Have some more...

Hell yeah, you do!

I will try my best.

> We all know recordings usually have more distortion than reproduction gear. Many cheap op-amps are commonly
> used with not so great PCB layout, imperfect anti-alias filters before the ADC, signal processing, instruments with
> distortion by themselves, compression, limiters and whatnot.

I think that is not always true and a oversimplification. You can get gear with PGA2500 Mic-Pre from TI into a really goo AD converter Chip (TI PCM4222) for a recording chain that I think has very low noise and distortion.

Now the microphone may be another issue, though usually microphones are not harmed (yet) buy adding op-amp's to them.

Many studios now work strictly "in the box" and purely in the digital domain. Others do not.

> When this is reproduced near perfect (say -140dB noise and THD) and that sounds boring does that mean the
> distortion in recordings is boring and becomes less boring when an inaudible amount of extra distortion is added ?

Well, if we took, for example, an audiophile recording by Keith Johnson (Reference Recordings) and still get this effect we may consider that perhaps this is not a valid hypothesis.

I mainly object because it is a variation of "Oh, people just like distortion and they are wrong for liking it and they should learn to listen to pure audio instead". It'stoo pat, easy and cynical and kinda on the same level as the average Karen's fashaming her chubby neighbour in the Gym.
> Is it the distortion spectrum that is important even at those (inaudible) levels or the lack of noise ?

NO IDEA. The Amplifier in question is H3 dominant even after pulling all the tweaks and this would suggest it's"bad distortion".
> Have you ever done some tests when you found that or just took it for what it is and decided
> chasing lowest noise/distortion is not beneficial anyway.

Functionally I test in order to have a sellable product. I don't have time/budget for scientific research.

> Then I would suggest leave the sighted out of the results which narrows down the testing.

Well, many of these "sighted tests" were trusted retailers and/or distributors and other industry professionals receiving "Alpha Versions" (me included) and giving their reactions. The expectation for all of us was to listen to something great and we were disappointed. We actually all wanted this product in the original condition to sound great and admitting it does not is not easy.

> What was compared exactly 2 similar designs with only 1 difference or compared to a certain reference amp ?

Comparison was between multiple units of the same production in pre-production, externally visually identical. Internally some where as originally designed others modified. The listeners were unaware which unit was which.

> How many listeners with what headphones and what music ?

I'd say over a dozen or so listeners, headphones wide ranging, including Sennheiser 600, 800, AKG701, AKG K1000, Audeze, Abyss, Focal, Final. Various IEM's including Campfire Andromeda (original), music generally listeners choice.
> How many attempts per listener ?

No limit.
> AB or ABX (in case of blind) ?

ABCDEF
> Or were all listening tests kind of informal and only about preference with a minimal amount of switching ?

No switching, unplugging headphones and plugging them into another unit.

> Were there records made of these tests. It seems to me this would yield very important info.

Yes. I have no access now and no idea where there are now. It's been years.
> Certainly in the light of the mentioned tube amps, with their many forms of distortion, that would make the
> -110 vs -140dB something that could end all debates and invalidate the reasons for forums like this.

I think basic common sense and looking at transducers distortion and compression should do that. If that does not do it, nothing will and any evidence will be picked apart, a nit found and even if this nit does not invalidate most of the whole test and the result will be used to rubbish the results and to keep going as always.

I still remember the whole story with Earl Geddes, his (very well meaning) distortion metric and his listening tests and the fact that for ONE (and only one) of the test cases a minor issue was discovered in the math used, but it was used to wholesale reject all of Mr. Geddes work.

It is a typical cargo cult science reaction. When I was still in a mood to fight windmills I did with a gun in the field and I had my belly full of that a few times over.

Anyway, too many people listen to stuff like this, so what are we even debating sound quality?

> (SINAD chasers) where beyond a certain point (my opinion) the benefits are only numerical and not of any practical concern.

We both agree.
> What else was different other than technical performance (noise + THD) ?
> It seems like fundamental design differences between the topologies but maybe not in the output stage.

No, the overall schematic remained the same. I hope you can accept that I am playing my cards a bit close to the cheat about exact details. In a lot of ways it was not dissimilar to what was done in the famous Carver/Stereophile affair:

The Carver Challenge J. Gordon Holt | May 11, 2009 | First Published: Oct 11, 1985 Is it possible to make a $700 "mainstream-audio" power amplifier sound exactly like a high-priced perfectionist amplifier?

> Was it the distortion component, noise component or the combination that was
> responsible for the 'boring-ness' or was that not specifically tested for ?

No idea. HD was pretty much pure H2/H3 with maybe a smidgen of higher order at much lower levels. Like so:

Audio Science Review: iFi Pro iCAN Headphone Amplifier Review

> What was the maximum output level the -140dB was referenced to ?

Probably 4V balanced like Amir did, could have been a bit higher.

In balanced mode and with 0dB Gain Ein is that of a pair of high GM JFet's (~ 2nV|/Hz or around 0.3uV) and some of the feedback loop resistors. Amir got 88dB re 50mV and 124/125dB re 2V I think SE.

Calculating reverse from 50mV and 88dB I get - 114dBV Ein the way Amir tested or ~ 2uV, which I think is unweighted.

> Even with 20V (which is quite loud in headphones) 2uVrms noise is quite a feat

Thank's. Obviously this HP Amp does manage that.
> and when considering headphones with say 100dB/V sensitivity that be 13dB below the audible threshold which is impossible
> to reach with a headphone as self noise of blood would be many, many dB's above that level.

Absolutely.

> It is really interesting (to me) and if I ever encountered such a phenomenon I would pursue it till funds dried out.

I had a schedule to keep and a product to deliver to the market.
> That would be my question. An unlikely possibility would be RF coupling and demodulation
> in the amp but we can safely assume this is not the case seeing LISN's and EMC gear was used
> to check performance/immunity in the RF range.

That should still be present even with all the little THD tweaks pulled out.

> Is this also the case if one uses a very high open loop gain (say nested feedback) which is highly
> effective in minimizing THD and noise but the feedback of course is 'natural' (just resistors and Miller caps).

Generally speaking, yes. As what you describe is in effect the fundamental structure of the iCAN Pro.

All stages are degenerated (generally at least 20dB) with local feedback, including the input stage that is in effect "current feedback").

High open loop gain is achieved by avoiding loading effects from the finite input impedance of the discrete stages on previous stages. Each stage is "broadbanded" by minimising capacitive loading. This means relying on very low parasitic capacitance FET's for all gainstages and making the current sources loading stages as linear as possible (cascoding).

So for each each stage (well, there are only two for gain and a BiMos power follower) linearity is maximised inherently. Compensation can be simple (Miller) or Linsley-Hood style.

> The 'becomes boring' is interesting though.

Another way is to say reduced emotional involvement.

Now another one for fun. To me (and most listeners) switching from this:

to this:

Improved emotional involvement and preference even is the gain difference was not compensated (that was a whole round of tests), while Amir observed:

"I switched to Ether CX with balanced cable with its 25 ohm and inefficient design. The iCAN drove these to maddeningly high SPL levels in solid state mode. This is one powerful headphone amplifier.

I toggled the mode switch to Tube and the difference was very subtle. Due to long switchover time, direct comparison was difficult. I thought there was some difference but I could not quantify it, or give points for either mode.

No such problem with Tube+ mode. Performance dropped massively, with bass taking a big hit, becoming soft. Overall sound was dull and unexciting even after I turned up the volume."

Mind you, Amir was listening sighted, without rigidly matching levels and I would suggest that Amir has a strong expectation bias. The Ether CX is 108dB/1V so listening at 105dB peak SPL for 0dBFS signals from the source would require less than 1V, improving THD&N/SINAD by at least another 6dB.

The (uncompensated) level difference was 0.62dB incidentally, from Solid State to Tube+ and 0.47dB from Tube to Tube+. These level differences are large enough that should be detected in an ABX test, incidentally. So they should be compensated if performing listening tests.

So perhaps Amir heard what he expected to hear?

Not having a dig at Amir BTW. It happens to me a lot, that I hear what I expect to hear if I am not very, very careful to avoid that.

Then again, when not listening "professionally" I am happy I hear what I expect a lot...

Thor

 

[solderdude]

I think that is not always true and a oversimplification. You can get gear with PGA2500 Mic-Pre from TI into a really goo AD converter Chip (TI PCM4222) for a recording chain that I think has very low noise and distortion.

Yep, there are some pretty high end studios with very competent engineers around that make astoundingly good recordings.

Now the microphone may be another issue, though usually microphones are not harmed (yet) buy adding op-amp's to them.
Well, if we took, for example, an audiophile recording by Keith Johnson (Reference Recordings) and still get this effect we may consider that perhaps this is not a valid hypothesis.

I mainly object because it is a variation of "Oh, people just like distortion and they are wrong for liking it and they should learn to listen to pure audio instead". It'stoo pat, easy and cynical and kinda on the same level as the average Karen's fashaming her chubby neighbour in the Gym.

Most responses are a bit shortsighted indeed as it usually is more than just an elevated amount of distortion (profile).
The gist is that the original waveform is changed (lets not use the word distortion even though it technically is) in an audible way and/or the perception of the listener is 'modified' by being impressed or a belief.

Functionally I test in order to have a sellable product. I don't have time/budget for scientific research.

Figured as much. I think all commercial products are, or at least, should be tested for that.
It's what I like about Harman research where they also publish (partly) what they found.
In the end it is all about making money and not the research for certain aspects. That would need to be funded by someone.

Well, many of these "sighted tests" were trusted retailers and/or distributors and other industry professionals receiving "Alpha Versions" (me included) and giving their reactions. The expectation for all of us was to listen to something great and we were disappointed. We actually all wanted this product in the original condition to sound great and admitting it does not is not easy.

Understood. I have been in that kind of business from the repair-side of retail and to be honest I do not think highly of most importers/retailers when it comes to understanding what they are dealing with. All of them are fully convinced being 'authorities' while lacking all kind of (what I believe to be) basic knowledge.

> AB or ABX (in case of blind) ?

ABCDEF

hmmmm... I find comparing A to B and or X already exhausting. I could not throw more into the mix unless it is about switching between a bunch of speakers and quickly assessing which are worth listening and not. And even then I prefer to switch between a known reference and C, D etc.

No switching, unplugging headphones and plugging them into another unit.

Ah....

> Were there records made of these tests. It seems to me this would yield very important info.

Yes. I have no access now and no idea where there are now. It's been years.

Was probably not allowed to make public then anyway, certainly in development stages.

When I was still in a mood to fight windmills I did with a gun in the field and I had my belly full of that a few times over.

I used to fight windmills as well in the past. Even when I thought I had clearly proven a point to someone I always found it did not change their minds.
These days I just post my opinion with some kind of explanation which I hope isn't too technical and think I am talking in 'layman language' but always still seems to technical.

> Was it the distortion component, noise component or the combination that was
> responsible for the 'boring-ness' or was that not specifically tested for ?

No idea.

Have no idea either, well.... maybe some ideas but they are the usual ones.

> Even with 20V (which is quite loud in headphones) 2uVrms noise is quite a feat

Thank's. Obviously this HP Amp does manage that.

Overkill but as long as that doesn't hurt longevity and the price I am all for that.

Thanks for expanding on the design things.

> The 'becomes boring' is interesting though.

Another way is to say reduced emotional involvement.

Now another one for fun. To me (and most listeners) switching from this:

to this:

Improved emotional involvement and preference even is the gain difference was not compensated (that was a whole round of tests), while Amir observed:

"I switched to Ether CX with balanced cable with its 25 ohm and inefficient design. The iCAN drove these to maddeningly high SPL levels in solid state mode. This is one powerful headphone amplifier.

I toggled the mode switch to Tube and the difference was very subtle. Due to long switchover time, direct comparison was difficult. I thought there was some difference but I could not quantify it, or give points for either mode.

No such problem with Tube+ mode. Performance dropped massively, with bass taking a big hit, becoming soft. Overall sound was dull and unexciting even after I turned up the volume."

When one would be looking at multitones of both modes it would probably more clear as to the differences in performance between the modes where tube+ approaches potentially audible levels (music dependent).

It happens to me a lot, that I hear what I expect to hear if I am not very, very careful to avoid that.

Guilty as charged, all fine as long as one dares to admit it.

Then again, when not listening "professionally" I am happy I hear what I expect a lot...

I am only looking for enjoyment and care less about technical performance.
Prefer to know that the gear is at least capable enough.


Fun exchange which is refreshing...

 

[pma]

Have some more...

We all know recordings usually have more distortion than reproduction gear. Many cheap op-amps are commonly used with not so great PCB layout, imperfect anti-alias filters before the ADC, signal processing, instruments with distortion by themselves, compression, limiters and whatnot.

When this is reproduced near perfect (say -140dB noise and THD) and that sounds boring does that mean the distortion in recordings is boring and becomes less boring when an inaudible amount of extra distortion is added ?
Is it the distortion spectrum that is important even at those (inaudible) levels or the lack of noise ?
Have you ever done some tests when you found that or just took it for what it is and decided chasing lowest noise/distortion is not beneficial anyway.



Then I would suggest leave the sighted out of the results which narrows down the testing.
What was compared exactly 2 similar designs with only 1 difference or compared to a certain reference amp ?
How many listeners with what headphones and what music ?
How many attempts per listener ?
AB or ABX (in case of blind) ?
Or were all listening tests kind of informal and only about preference with a minimal amount of switching ?


Were there records made of these tests. It seems to me this would yield very important info.
Certainly in the light of the mentioned tube amps, with their many forms of distortion, that would make the -110 vs -140dB something that could end all debates and invalidate the reasons for forums like this. (SINAD chasers) where beyond a certain point (my opinion) the benefits are only numerical and not of any practical concern.
What else was different other than technical performance (noise + THD) ? It seems like fundamental design differences between the topologies but maybe not in the output stage.
Was it the distortion component, noise component or the combination that was responsible for the 'boring-ness' or was that not specifically tested for ?
What was the maximum output level the -140dB was referenced to ?
Even with 20V (which is quite loud in headphones) 2uVrms noise is quite a feat and when considering headphones with say 100dB/V sensitivity that be 13dB below the audible threshold which is impossible to reach with a headphone as self noise of blood would be many, many dB's above that level.

It is really interesting (to me) and if I ever encountered such a phenomenon I would pursue it till funds dried out.
I can understand that for a business this would be really not that interesting though and making a profit and getting lots of sales is of more importance.


We can safely discard the retailers and distributors opinions. From a business standpoint that feedback (regardless if obtained in a scientifically sound manner or not) is far more valuable than some academic style research.


Yep, my thoughts as well + I have not met any people that have a dynamic range, during music listening, exceeding 70dB and do not know of recordings with a low enough noise floor.


That would be my question. An unlikely possibility would be RF coupling and demodulation in the amp but we can safely assume this is not the case seeing LISN's and EMC gear was used to check performance/immunity in the RF range.



Is this also the case if one uses a very high open loop gain (say nested feedback) which is highly effective in minimizing THD and noise but the feedback of course is 'natural' (just resistors and Miller caps).



I wasn't there so could only guess as to what the reasons were. Preference is preference though 0.1% or 0.00001% isn't that an important difference depending on the recording, used transducers and playback level.

The 'becomes boring' is interesting though. I usually relate that to recordings and transducers, time of day and mood mostly. Hardly ever to gear.
Admittedly I never played with any amps better than -110 THD+N (certainly not loaded) so have no reference frame there.



Danke... und ein gutees neues Jahr

About 20 years ago, I was asking myself similar questions you now have with Thorsten Loesch aka Kuei Yang Wang (none of these are his real names). I do not think there are simple answers and as opposite of TL, I have stopped discussions about such things in a public forum. It goes to nowhere. Yes, I have AB test records, with individuals or small groups up to three people. And I will not go into any deeper discussion anymore. It is just wasting time. Just to mention, living under conditions of a DDR/CZ engineers gave us something, even if we are happy that those times have ended.

 

[krabapple]

About 20 years ago, I was asking myself similar questions you now have with Thorsten Loesch aka Kuei Yang Wang (none of these are his real names).

strange

Thorsten Loesch

Thorsten Loesch - contributor from the UK - TNT - Internet HiFi magazine

www.tnt-audio.com

Thorsten Loesch of AMR Abbingdon | Atelier 13 Audio

About Thorsten Loesch - Chief Designer at AMR Abbingdon Music Research

www.atelier13-usa.com

https://hk.linkedin.com/in/thorsten-loesch-studio-raumklangg

etc.

Whoever he is, I don't find his diatribes against ABX (which he later modifies to 'ABX as usually practiced' ) novel or at all convincing, nor his anecdotes from 'tests' conducted way back when.

But it's Christmas and life is short.

 

[solderdude]

About 20 years ago, I was asking myself similar questions you now have with Thorsten Loesch aka Kuei Yang Wang (none of these are his real names). I do not think there are simple answers and as opposite of TL, I have stopped discussions about such things in a public forum. It goes to nowhere. Yes, I have AB test records, with individuals or small groups up to three people. And I will not go into any deeper discussion anymore. It is just wasting time. Just to mention, living under conditions of a DDR/CZ engineers gave us something, even if we are happy that those times have ended.

I was wrestling with similar questions over 30 years ago while working as a repair guy and found the answers back then already.
Of course, -140dB S/N etc was not a thing in those days and digital for consumers was CD only. No PC audio yet, no SD cards, no network players nor internet as we know it.

Have stopped asking those questions but it was fun to pick Thorsten's brain. I don't think Thorsten minds if he was just talking to a penguin with the name 'solderdude' either.

Audio is a fun hobby, the engineering behind is what I take serious. Opinions and the discussions around it is sometimes interesting.