WHY no crosstalk measurements in your dac measurement

[shoden]

WHY no crosstalk measurements in your dac measurement!!!!!!!!!!!!!

I find this measurement crucial, there is a reason people prefer dual mono dacs...

i cant believe you recommend the modi 3 when on their website they say Crosstalk: -80dB, 20-20kHz

even phase shift of between channel at the output should be measured and also ultra fast pulse distortion.. square wave distortion at multiple frequencies.. like 30hz 100hz 300hz 10khz

also dac latency at all input and all drivers on USB.. be it asio etc.. RME will have very low latency in asio since they are pro audio card company..

most audiophile dac have very bad usb latency because of their crappy drivers.. or because of their async reclocking module

you have to redo all your measurement with these included or else you give false hope to people and they will buy crap based on your limited measurements..

you currently encourage topping and other companies to focus on THD and distortion but they will omit other important aspect. For example they skip doing dual mono for all the path to save money.

 

[DonH56]

Crosstalk of -80 dB will be completely inaudible with a real source using real speakers in a real room. The main benefit of mono, or dual-mono, design practice is in power amplifiers when the two (or more) channels need not share the same power supply. Even that is rarely an issue.

In consumer systems latency is not really an issue (for recording or working with a DAW, yes, but not for consumer playback, except perhaps for gamers and then it is usually a driver issue rather than a fundamental HW issue).

As for the rest, this just reads like trolling, sorry...

 

[andreasmaaan]

As for the rest, this just reads like trolling, sorry...

It's at least very... forward, isn't it

 

[Soniclife]

It's at least very... forward, isn't it

Sounds like he wants a refund.

 

[stunta]

i cant believe you recommend the modi 3 when on their website they say Crosstalk: -80dB, 20-20kHz

If Schiit is going to have specs that make them look bad listed on their website, I would have to rethink the whole meaning of life, universe and everything.

 

[stunta]

In consumer systems latency is not really an issue (for recording or working with a DAW, yes, but not for consumer playback, except perhaps for gamers and then it is usually a driver issue rather than a fundamental HW issue).

Or for video if the DAC is slow enough to cause lip-sync issues. I've never heard of this though except for some heavy DSP-based systems like Devialet. For the purpose of Amir's reviews, I think the focus is on music playback. In any case, I wouldn't demand anything of Amir - he doesn't work for us

 

[RayDunzl]

The little woman is watching General Hospital, broadcast TV.

Cable Box -> TV -> Optical Switch -> miniDSP OpenDRC-DI -> DEQ2496 -> DAC ->preamp and ADC/DSP/Class D in JBL LSR 308.

So, there are some delays already baked into the system here.

I added another 100ms delay at the miniDSP (I'd consider that a lot), and found it still acceptable/almost unnoticeable for "lip sync" ( I didn't press for a limit).

Will try again with someone playing a musical instrument... I know I've been annoyed with latency there in the past. Drums might be the worst offender.

100ms at 48kHz (TV) = 4,800 samples delay, a goodly though not Chord-like amount for a digital path.

She did complain ("What happened???") when the delay was initially applied to only the left channel, on the way to being applied to the other. Sudden audible inter-channel echo occurred.

 

[amirm]

I quit!!!!

 

[mindbomb]

That crosstalk seems really high. Are they measuring at 32 ohms for some reason?

 

[Blumlein 88]

I quit!!!!

You act as if you have the option to quit.

I don't think the OP said do this or else. Just do this or you'll be giving people false hope.

 

[sergeauckland]

Crosstalk of -80 dB will be completely inaudible with a real source using real speakers in a real room. The main benefit of mono, or dual-mono, design practice is in power amplifiers when the two (or more) channels need not share the same power supply. Even that is rarely an issue.

In consumer systems latency is not really an issue (for recording or working with a DAW, yes, but not for consumer playback, except perhaps for gamers and then it is usually a driver issue rather than a fundamental HW issue).

As for the rest, this just reads like trolling, sorry...

Crosstalk hasn't been an issue for over 50 years. Anything over 20dB is quite adequate on programme material, as evidenced by good old vinyl, which rarely manages more except at mid frequencies, and yet provides perfectly good stereo. FM radio rarely did better than perhaps 40dB, and yet digital sources manage over 80dB routinely, so is hardly worth a mention unless for some bizarre reason it should be much worse.

Distortion in the crosstalk is more of an issue, and I've measured over 1% distortion in the crosstalk. So, if the crosstalk is only, say, 20dB, and distortion of that crosstalk is 10% (i.e. -20dB), then it's possible that the distortion in the main channel is as much as 1% which could just be audible. With digital sources, however distorted the crosstalk is, it becomes irrelevant when crosstalk is -80dB or better as inaudible whatever it is.

If a DAC should have unexpectedly poor crosstalk, it might be worth knowing, as it could be indicative of poor design, but if it's as expected, then not worth mentioning.

S

 

[FrantzM]

Amir

THis time you got caught red handed... You need to focus on the essential You are a shill to the Audio Precision people when you could and should have asked them to devise a PRATT-0-meter, of course a soundstage width and depth meter and an Organic-meter for Digital devices. Of course you need to measure crosstalk, you are routinely measuring THD and screams when it hits -80 dB yet you make no mention of crosstalk!!. Some have mentioned Vinyl but it is because most do not listen to mono with monoblock and one speaker preferably full range then they would understand how such systems have no crosstalk ...

So measure or shut up!!!

 

[restorer-john]

Crosstalk of -80 dB will be completely inaudible with a real source using real speakers in a real room.

Whilst I agree with Don, channel separation/crosstalk testing was always a parameter tested on CD players. It sought to pursue the absolute best in terms of layout and design and interchannel interference.

80dB is terrible for digital and such a number should be called out for what it is.

One could easily and successfully argue there is little point digging down to 0.0001% from 0.001% THD as those numbers are also "completely inaudible".

Focusing on THD is an obsession here, when other parameters such as absolute phase, square wave response, impulse response, channel separation and 16 bit performance are all ignored.

I don't think jumping on the OP is really fair in this case.

 

[DonH56]

Note -80 dB is only 0.01% (still inaudible, but in the context of other measurements is "large"). I have a vague memory of Amir running some crosstalk tests early on but deciding it was so small in most cases that it was overshadowed by other things so he reduced scope to fit his test time.

My comment was spurred more by the unexpected attack, the unreasonable (to me) demand for more tests, and claim that all previous tests must be redone due to these omissions.

I have advocated for other tests, primarily impulse/step response and multitone, but realize Amir has finite time and is not a test lab (though he is doing a pretty good job of it). That said, when someone has questioned something or asked for more, he has been good at providing additional measurements. Again, I was taken aback by the the OP's tone, and seemingly "all or nothing" approach.

As far crosstalk, 80 dB is pretty good, and expecting 130 - 140 dB as the OP suggested in another thread seems a bit unreasonable to me. But I have not looked at crosstalk for some time. Taking a quick gander at a few things I have in my files:

DAC: Benchmark DAC3: -137 dB at 20 Hz; -130 dB at 1 kHz, -116 dB at 20 kHz
Player: Oppo UDP-205: analog outputs >118 dB (no frequency range provided)
Player: Sony UBP-X800 -- not specified
DAC: Topping D50: -116 dB @ 1 kHz
AVR: Anthem AVM 60 -- not specified
AVR: Denon X4500 -- not specified
Processor: Emotiva XMC-1: >100 dB
Preamp: Mark Levinson No 523: >107 dB @ 1 kHz
Amplifier: Benchmark AHB2: >115 dB @ 1 kHz, >92 dB @ 20 kHz
Amplifier: Crown xls 1502 power amp: >85 dB @ 1 kHz, >55 dB @ 20 kHz
Amplifier: Emotiva XPA-series -- not specified
Amplifier: Outlaw Model 5000 (7-ch amp): >70 dB from 20 Hz to 20 kHz
Amplifier: Mark Levinson No 532H -- not specified

So, I am very wrong about crosstalk for current SOTA DACs and amplifiers, but not too far off that of amplifiers, which is mainly where I look at crosstalk. Whilst I still contend 80 dB is inaudible for any realistic environment, it is obviously well below that of other low-level components.

Mea culpa - Don

 

[restorer-john]

So, I am very wrong about crosstalk for current SOTA DACs and amplifiers, but not too far off that of amplifiers, which is mainly where I look at crosstalk. Whilst I still contend 80 dB is inaudible for any realistic environment, it is obviously well below that of other low-level components.

Mea culpa - Don

I don't think you a wrong at all Don, I just think people have no historical perspective (well a few of us do ) (not that it means we are old or anything)

I think people need to be reminded that these separation figures for so-called SOTA D/As aren't really state of the art (except perhaps the Benchmark).

Put into perspective, these types of numbers were regularly significantly exceeded 30 years ago on single box CD players. THD on 16/44 was regularly better, especially at high frequencies and high levels than many of the D/As I see reviewed here and elsewhere.

As for S/N ratio, it seems things have gone backwards significantly- the machine reviewed below achieved dithered S/N with respect to 2.0V of 118.2dB (left) and 117.6dB (right) and that was in 1989 on a USD$2000 TOTL product. (another review tested it at 117.7dB and 117.8dB so there is good correlation between testing regimes and individual products)

Separation over 110dB was common in the CD era. Here is a scan of Len Feldman's review of a TOTL Sony CDP-X7esD:

Here's the AP (system one plot)

Even 20KHz 0dBfs THD on 16/44 material was 0.0015% on that machine in 1989.

It's half price brother (CDP-338esD) achieved virtually as good results. I've got reviews where they raised their spectrum analzyer FFT trace 30dB and still couldn't find any crosstalk. (>120dB).

A lot of this modern D/A in-a-can stuff is rubbish, with ordinary performance and it needs to be said. I don't agree with the way the OP introduced himself, but there is some truth buried in his post. Also, if nobody is testing it, reporting it and calling it out, why would manufacturers try to improve these key parameters?

 

[March Audio]

WHY no crosstalk measurements in your dac measurement!!!!!!!!!!!!!

I find this measurement crucial, there is a reason people prefer dual mono dacs...

i cant believe you recommend the modi 3 when on their website they say Crosstalk: -80dB, 20-20kHz

even phase shift of between channel at the output should be measured and also ultra fast pulse distortion.. square wave distortion at multiple frequencies.. like 30hz 100hz 300hz 10khz

also dac latency at all input and all drivers on USB.. be it asio etc.. RME will have very low latency in asio since they are pro audio card company..

most audiophile dac have very bad usb latency because of their crappy drivers.. or because of their async reclocking module

you have to redo all your measurement with these included or else you give false hope to people and they will buy crap based on your limited measurements..

you currently encourage topping and other companies to focus on THD and distortion but they will omit other important aspect. For example they skip doing dual mono for all the path to save money.

I don't think crosstalk is crucial at all. The levels in most kit is below the point where it will cause any problem. However, the same can be said about thd, so I have to say OK perhaps we should measure it for completeness of information. Phase shift is practically non exist ant in modern dacs. Do you have any evidence to show otherwise or data to show people are sensitive to the very low differences found?

Latency, what's your application? Again latency is fine in most dacs. Unless you have a specific application such as fold back for musicians it's a non issue. In which case I would have thought you would be looking for a professional interface anyway.

If you want to talk about latency issues I am currently struggling to get my active dsp speaker processing below 70ms so I can use it for home theatre. My AVR won't cope if it thinks the speaker distances are above 24m!

Now I am a techy geek who wants to see all of this info and more, however it has to be balanced with effective communication in the reviews. Much of the audience is not of technical level to find it useful so if a review is 10 pages of graphs then the message gets lost and people turn off. Then of course the extra time to test and write reports when I know Amir is already overloaded.

Perhaps a solution is the further automation of extensive tests, which I believe the AP is adept at doing, the reviews stay in current format picking out pertinent measurements, with a link to the extended data which can simply be in the report form directly from the AP software. People who are interested can delve into the detail without much/any additional work for Amir.

 

[maxxevv]

If its in the name of completeness, perhaps but if its for general consumer consideration probably not as it would mostly be moot in the results consequences as already pointed out. It has to be a really terrible piece of engineering if crosstalk is below 80 dB.

But that being said, I have always wondered regarding the impulse response of amplifier output and their effects on sound output as a direct consequence.

It goes back to the often cited catch phrase " The HD800 is amp picky... " .

I always thought that it was just a lot of hot air from users but turns out there is an element of possibility due to the extremely high transient impedance values of the HD800. That the impulse response of some amps may not be sufficient to keep up with the transient demands of the headphone and resulting in clipping, thus altering its output.

Not sure how it would be adequately measured in this regard, but I thing it would help debunk and demystify some of these motherhood statements on the various forums.

 

[amirm]

I don't know why Schiit says crosstalk is -80 dB. This is their measurements:

And in the report it says:

 

[March Audio]

If its in the name of completeness, perhaps but if its for general consumer consideration probably not as it would mostly be moot in the results consequences as already pointed out. It has to be a really terrible piece of engineering if crosstalk is below 80 dB.

But that being said, I have always wondered regarding the impulse response of amplifier output and their effects on sound output as a direct consequence.

It goes back to the often cited catch phrase " The HD800 is amp picky... " .

I always thought that it was just a lot of hot air from users but turns out there is an element of possibility due to the extremely high transient impedance values of the HD800. That the impulse response of some amps may not be sufficient to keep up with the transient demands of the headphone and resulting in clipping, thus altering its output.

Not sure how it would be adequately measured in this regard, but I thing it would help debunk and demystify some of these motherhood statements on the various forums.

the extremely high transient impedance values of the HD800.

I think you have to explain this, I for one have no idea what it means, sorry. I think you may be confusing some different technical areas

FWIW I pay no attention whatsoever to the "difficult to drive" statements that are banded about. It usually betrays a lack of understanding of the issues involved.

The main issue is that different headphones have a very wide range of impedance and sensitivities. They are not difficult to drive in the slightest, generally only requiring a few tens of mW to get them to stupid volume levels. They generally (there are exceptions like iems) dont have crossovers which create horribly complex reactive loads.

So you need an amp that will provide an adequate output voltage to drive the higher impedance/lower sensitivity cans. Lower impedance cans shouldnt be too much of a problem. If an amp is running out of steam and cant provide the small currents and power to drive the lower impedance cans to decent level without distortion, then it really is poor. Why so many headphone amps have high output impedance is beyond me. No reason for it at all.

So really ignore most of the subjective rubbish talked in forums about the subject. This is where measurements rule. Amir shows the amp behavior into different loads, output voltage etc.

Just a note on impulse response of amps. Ask yourself a question. How fast does an amp need to respond to a change in signal? Whats the fastest changing signal that is going to be presented to an amp?

Well unless you are a bat probably 20kHz. So if the frequency response is flat to well beyond 20kHz is this "impulse response" an issue? Probably not. Its a bit more complex and more to it than that but people imagine that a transient signal like a drum hit is really fast / high frequency. Its not. There may be harmonic content that sprays up into higher areas but if your microphone or CD recording only goes up to 22.05 kHz then there is not much reason for your subsequent amp to do anything spectacularly beyond this in terms of FR or therefore impulse response. Your signal is already band limited. So square wave responses which you sometimes see are very misleading IMO.

 

[maxxevv]

The impedance values in most output devices be it headphones / earphones / speakers are not flat across the full spectrum of audio frequencies they are supposed to output sound. As such when a device is supposed to produce certain frequencies, it will result in varying levels of power demands on the input devices.

Here's a measurement from Innerfidelity:

HD800

Focal Utopia

HD660 / HD700

I have seem some impedance curves from speakers too. But they generally don't look as extreme as the above.