New review of the iFi ZEN DAC V3

You may read the full review here posted in this thread, but if you wish to see the full review with test graphs and interpretation, you can see it at https://www.euphonicreview.com/reviews.html

 

Thanks for the excellent review 

1 hour ago, Andrew Allen Ballew said:

You may read the full review here posted in this thread, but if you wish to see the full review with test graphs and interpretation, you can see it at https://www.euphonicreview.com/reviews.html

 

When the sun dips below the horizon and the world quiets down, true audio aficionados know it’s the perfect time to indulge in their favorite tunes. Enter the iFi ZEN DAC V3, my nocturnal companion in this auditory adventure. In this review, we'll explore whether this sleek little device lives up to its promise of delivering high-fidelity sound that could make even a bat envious of its precision. Join me as I dissect the ZEN DAC V3’s performance, unravel its quirks, and see if it truly enhances those cherished nighttime listening sessions.

One of my very favorite easy entries into to the high fidelity head-fi world is the iFi ZEN DAC V2.  It has a more than capable DAC with a unique architecture that lends to its beguiling good sound at its low price.  The built-in headphone amp is also of good quality, with an impedance and gain matching function called 'Power Match' that widens the range of headphones it can drive well.  Even some typically harder to drive, high impedance headphones will respond decently well.   

The ZEN V3 is more of the same, but better.  The DAC section is the same as before, and measurements show there is not much of a new story to tell here, although subjective experience says there is just a bit more refinement to the sound.  What is truly an upgrade here is the headphone amp.  Just as with the previous V2 version, the Power Match feature does this piece of kit a great service, as it does the ear a great service.  I found no difficulty driving higher impedance headphones, and low sensitivity models such as the Hifiman HE-560 had more than adequate drive.  

Continuing on the features front (click here for more details about the iFi ZEN DAC V3 feature set in whole), three large changes I will highlight are the change to USB-C input, the upgrade of the headphone amp processing, and the addition of DXD 2x and DSD512 capability. 

The V2 had a bass boost called 'TRUEBASS', which was basically a shelf EQ that lifted the frequency response from 200hz down to 20hz.  The V3 has the 'XBass+' feature, which is like the more refined processing found in the ZEN Can Headphone Amplifier.  It is a more nuanced and natural sounding effect that really enhances the listening experience, especially if you are a basshead. Or maybe you just feel like some extra groove one day without overdoing it.  'XBass+' is your ticket.  

Another major feature change is just part of an industry standardization to universal usage of USB-C.  I don't like it.  Call me and old man resistant to change as I approach 50, but I already have a slew of USB cables and I am a bit resentful knowing eventually they will all have to be changed to these little smart phone sized connectors.  We are doing quite well, though, that this is my only nitpick worth mentioning about the iFi ZEN DAC V3.  I even like its slightly darker, multi toned color scheme.  It really looks good.  iFi aesthetics have really come a long, long way from a decade ago with the introduction of the iDAC. (Click here for a trip down legacy lane.) 

Finally, there is now an option for DXD 2x and DSD512 playback.  As there are no native DXD 2x or DSD512 recordings I can think of, this is a feature directed at the growing set of audiophiles who are using programs like HQplayer to upsample all their music to 768khz, or DSD512 and higher to achieve performance that the onboard DAC cannot achieve with its own oversampling filters.  I did do some experimentation on my own, and I can confirm that DXD 2x and DSD512 playback worked perfectly with HQPlayer via direct PC connection.  (HQPlayer also works perfectly with the iFi ZEN STREAM, which has its own dedicated HQPlayer mode.  However, the STREAM is limited to transmission of DXD 1x and DSD256.)

SO HOW DOES SHE HUM ALONG, CAPTAIN?  

Once you have a look at the measurements at the end of the review, you might conclude the ZEN V2 and ZEN V3 sound exactly the same.  This would be an incorrect assumption, especially via the headphone output.  Nor is the sound difference limited to the headphone output.  The sound is a notable step upward from the V2, which was already my go-to standard in this price range.  And still would be even if the price were somewhat higher.  

The following listening impressions were using the V3 powered by its 5v USB bus through an Intona USB 2.0 Galvanic isolator, and with a ZEN Can headphone amplifier with neither of the analog processing enhancements activated for the majority of the evaluations.  The 4.4mm Pentaconn balanced connection was used to connect the ZEN DAC V3 to the ZEN CAN amp.  Playback source was Roon connected by both direct PC connection with WireWorld USB cables, and via the iFi ZEN Stream connected via ethernet and a WireWorld USB cable.   

I started with "Django" from " European Concert, Vol. 1" by The Modern Jazz Quartet in FLAC at 192khz 24bit.  The opening setting is extremely atmospheric with an excellent sense of stage. The Vibraphone has a luscious and full tone, with vibrato swells rendered to perfection.  Bass is not too strong, not too light, and has an excellent sense of its own playing space with good separation across the soundstage.  The Upright Bass follows in parallel with the piano, but does not dominate the line.  The ZEN is good enough to render the piano slightly farther back in the soundstage that the Bass, but at the same time allows the excellent musicianship that can fool the listener into the two merging into one organic instrument.  Before the drums kick in, the tiny detail of someone moving in their seat, which adds to the charm of this excellent track, is rendered just a naturally as if one was there in the midst of the recording space.  I am not 100 percent sure which chair makes the noise, but the precise imaging suggests it’s the drummer changing his posture as he is getting ready to play.  There is resolution measured on paper, and there is ‘musical’ resolution that is important to actual listening, and this is a fine example.  As the drums make their entrance, the quartet is completed, and the ZEN V3 unfolds the soundstage as well as anything in recent memory, including many more expensive products.  The drum brushes timbre is spot on, not sounding too bright and thin, nor too flat, lifeless and undetailed.  Natural is the word of the day.  One could listen to this for hours and hours and never experience any fatigue whatsoever.  At the same time, one would not be left wanting for any sense of detail or resolution. 

 
I next turned to "Sibelius Works for Violin in Orchestra", Ondine label mastered by 2XHD in DSD64 'from original DSD recording', listening to a remaster of Pekka Kuusisto playing Sibelius' "Humoresque no. 1 in D minor" along with the Tapiola Sinfonietta.  Normally this kind of recording, with its massive chain of conversions and DSP, is not my cup of tea.  I must admit however that there is some kind of voodoo in what 2x mastering engineers did, because there is such an abundance of texture, timbre, and an intangible presence or weight in the instrument sound that one just ‘knows’ when it is heard. 

The mastering process begins with an original Ondine label DSD64 master, converted to analog by an unnamed Digital to Analog Converter.  This analog signal is sent through a “high-end tube preamplifier” before being re-recorded in DXD by a DCS 905 ADC, and is redistributed in either DSD or PCM at 24/192. 

On paper, it seems they make a mess of things.  Not so much to the ear.  Via the ZEN DAC V3, I can hear the very distinctive individual timbre of Kuusisto’s violin.  Even the best violins on most recordings take on a somewhat generic sound, but not this time.  There is a distinctive slight ‘grain’ in the timbre, which is full bodied, never thin.  It is heavy on the overtones.  Exactly what my ears like to hear, measurements or accuracy be damned.   Either the original recording started this way, or the re-mastering process is responsible, but the soundstage here is quite intimate and a bit small, however the violin centerpiece is never intimate and small.  It dominates your attention and draws you in.  The ZEN DAC V3 takes what is on this recording and paints it tonally in the best of light.  Just out of curiosity, I did some research on what kind of violin Kuusisto may have been playing that night, as its tone struck me so deeply.  It isn’t his ‘Scotta’ violin made by Stradivari in his ‘Golden Period’ ca. 1709.  Even still, my curiosity was piqued.  Even if not the Scotta Strad, it is a world-class instrument.   (I wonder if Stradivari, if alive today, would be greatly insulted if told his instruments didn’t measure well enough.  Not enough SINAD, Vecchio Uomo.)  Frankly while using this track for review, I simply got lost in the music, the player and the instrument and its collective pure beauty.  I think that says all I need to say about the role of the ZEN DAC V3 in this case.  It transmitted the music, which is all I can ask of any piece of gear and at the same time is the highest compliment I can give. 

Finally, I needed to audition something that ‘rocked’ a bit to see how the DAC handles lesser recordings with a wall of sound.  I went to Tidal and cued up Lil’ Wayne’s duet with Nicki Minaj “What’s Wrong With Them”.  There is nothing audiophile here.  However, the ZEN DAC V3 completely holds it composure.  The powerful bass track never overwhelms or affects the vocal tracks, whether main or backing track, in any way.  The ‘wall of sound’ is powerful, yet everything stays in its lane, so to speak.  Lil’ Wayne’s rapping is as clear and articulated as anyone could ask.  Often I find hip-hop mixes to be too bright.  I think these guys doing the mixing are often half-deaf from too many years grinding their way to the top in the clubs. I guess its a price to be paid. 

There is nothing bright here in this Lil' Wayne mix. It sounds about as natural as a modern processed track can sound. 

To this point in time, as mentioned earlier, I was listening via the ZEN CAN, bypassing the ZEN DAC V3 headphone amp.  The reason I began auditioning this way is to get a baseline comparison with other DACs in my collection that occupy a similar price tier in the market.  To get a handle on how good the onboard headphone amp is, I switched at this time, and re-auditioned a couple tracks.  Headphones were still the Sennheiser HD650. 

Lil’ Wayne’s track “What’s Wrong with Them” sounded essentially the same, however, it was not presented with quite the same ease.  His lead rap didn’t stay quite as precise and centered in the mix, and the tone of his voice wasn’t quite as consistent.  Bass was not quite as strong either, that is, until I took the first opportunity to try the ‘XBass+’ feature.  (I had used the original XBass feature on the iCan while auditioning this track.)  IMPRESSIVE.  Now the sound was much more like what I heard through the iCan.  Not only did the XBass+ feature increase the power of the bass without any added bloat or boom, it seemed to have the side effect of creating more precision in the track overall.  It still didn’t reach the level of refinement of the ZEN Can, but it was impressive nonetheless.

The Sibelius Humoresque still sounded quite good through the ZEN DAC V3 onboard headamp, but the violin didn’t quite jump out at me like it did via the ZEN iCan.  It was still a very, very sweet sound with most of the same good qualities I heard through the iCan. 

There is a rumor that a new ZEN ICan is possibly on its way to us as well.  If it matches and even slightly exceeds the pairing of the ZEN DAC V3 with the current iCan, then that possibly will be the head-fi value of the year. 


CONCLUSION

Don't be fooled into thinking my criticism of the ZEN V3 onboard headamp compared to the dedicated ZEN headamp is a negative critique.  The reality is the onboard headamp is as good as in any integrated piece I have heard at this price.  It is an upgrade from the ZEN V2, and the 'XBass+' processing is a big part of that.  Its contribution to the overall presentation is very impressive.  Overall the iFi ZEN DAC V3, in spite of its many similarities with its predecessors, is a worthwhile upgrade.  If you already own the V2, you know as I do how already very, very good it is.  Normally I would suggest waiting another generation before you will find any significant upgrades.  In THIS Case, I would suggest just go ahead and upgrade now.  I find the ZEN V3 to simply be an exceptional player of music, and at this price range I can only think of products with similar fidelity that DO NOT have a built-in headphone amp of any kind, no to mention one that might have one as high quality as this!  Well done iFi.  

EASY PICK FOR EDITOR'S CHOICE STATUS.  

OUR RATING:  (0-10 scale.  8 or higher numbers represent best of industry.  For the final score calculation, each category is given a (undisclosed) weighting and is tallied for final score.   Although perfect 'fairness' is unlikely, our system tries to be as unbiased as possible.  

sound quality: 8.9
build quality: 9.0
ease of use: 9.5
measurements: 8.8
value: 10.0

TOTAL RATING:   91.7


 

---

 

---

 

MEASUREMENTS 

As is the usual case with iFi products, measurements are consistent, very good, and have no 'red flags' to indicate any issues. iFi creates very well thought out and engineered products.  

Frequency Response (20hz to 20khz): dB -0.35, +0.01
SNR:  -117.8db A-weighted
THD:  -110.3db / 0.000305%
THD+N: -100.7db / 0.000923%
IMD: -108db / 0.000398%
JITTER: 40.4ps Peak / 13.6ps RMS
LINEARITY at -100db: <0.1db error
LINEARITY at -111db: <0.5db error
LINEARITY at -115db: <2db error
THD+N Multitone 32 Test: -95.3db / 0.001718%
________________
DSD MEASUREMENTS

DSD64 THD+N: -100.7db / 0.000923%
DSD128 THD+N: -100.5db / 0.000944%
DSD256 THD+N: -94.8db / 0.00182%
DSD512 THD+N: -87.5db / 0.004217%

DSD64 JITTER: 74ps Peak / 35.4ps RMS

 

Thank you,  Andrew for the lik and the provide iformation.
Could you elaborate THD-N loss between DSD64 and DSD512 ?
thanks a lot, Tom

25 minutes ago, Andrew Allen Ballew said:

YEAH..

 

There are a couple things.  You first have to compensate in the design of the analog output.  The iFi DSD1793 chip has a discrete analog FIR filter for all DSD playback.  I won't go into that in depth.  What is important is, as you double the DSD bitstream rate, the filter cutoff doubles too.  The higher the filter cutoff, the more likely that ultrasonic noise passes into the analog stage, creating intermodulation distortion in the audible frequencies.  As far as the ZEN series of iFi products, based on measurements and educated guessing, the FIR filter is set at just under 80khz at DSD64.  Double the speed to 128, its corner frequency is at 160khz.  At DSD256, we are at 320khz, but by DSD512, its a very, very high 640khz.  Its just how these CIC type Analog FIR filters work.  Even though the noise is shaped to higher frequencies at DSD512---- at some point, something has to give.  I used a 7th order modulator in that test.  It was done in multitone.  I have DSD512 test tones made in HQPlayer that I have not used to test yet.  I am writing a followup article about this specific phenomenon.  

Oh, it is not unique to iFi.  Not at all.  As we are now measuring for these things, drops in both SNR and SINAD/THD+N are common in highspeed DSD.  

There is another reason too, that Andreas Koch alluded too awhile ago about how once the bitstream hits a certain speed, the logic in the DAC loses precision.  It just can't accurately do its job as well at those high speeds.  The tolerances are not tight enough.  At the time he stated this, it was disputed.  But again, our measurements are bearing this out now we are getting access to more NATIVE DSD devices.  

Here is his article on the issue:
https://positive-feedback.com/audio-discourse/raising-the-sample-rate-of-dsd-is-there-a-sweet-spot/



Thorsten Loesch has also made mention of this with the iDSD PRO.  Turns out it is one of the best at keeping consistent THD+N numbers, but at DSD1024 there is a measurable noise drop in compared to DSD512.  Other iFi products has always been very consistent as well up to DSD256.  More so than some others I have measured.  But when DSD512 was added as a feature to the more affordable stuff, a noteable drop is seen.  

I know this all sounds bad, but as for iFi products, they have been among the most consistent in this regard.  The have always shown very little change from DSD64 up to DSD256, and on the pro little change all the way up to DSD512.. but some degradation at DSD1024.  

This is just my opinion, but simply adding on the ability to convert DSD512 in the ZEN without addressing the stress it adds to the analog output is a bit meh, but I can't be too critical because its more a greater technological issue than any negligence, and beside, it would add more cost to make changes to the analog stage that might help address it somewhat.  


P.S

This is something you won't see when DSD is digitally filtered and remodulated internally by the DAC, such as with ESS, or with AKM when the bypass mode isn't engaged.  The measurements will be basically the same at all rates, PCM and DSD, because you are now measuring the characteristics of the multi-bit delta sigma modulator, and the onboard fully digital filters that can be perhaps more flexible and capable of dealing with ultrasonic noise.  Also, I don't think you will find any multi-bit DSM's running at 45MHZ bitstream speeds either!  They have no need to, because they can achieve significantly high levels of SNR running at no more than 256x speeds.

Thanks Andrew this is exactly what I noticed as well previously with the ifi neo that dsd256 was definitely the sweet spot. To be honest, I even found the ifi pro sound cleaner at dsd256.

interestingly, the dsc2 seem to actually sounded better at dsd512 than dsd256, not sure if that has anything to do with the transformer output stage in place of analog filters 

Just now, dericchan1 said:

Thanks Andrew this is exactly what I noticed as well previously with the ifi neo that dsd256 was definitely the sweet spot. To be honest, I even found the ifi pro sound cleaner at dsd256.

 

interestingly, the dsc2 seem to actually sounded better at dsd512 than dsd256, not sure if that has anything to do with the transformer output stage in place of analog filters 

your thought about the transformers mirrors mine.  I keep dragging my feet... I have so many things to measure.  Requests, and I want to get back to the DSC2 and do a thorough job checking its performance out.

2 minutes ago, Andrew Allen Ballew said:

your thought about the transformers mirrors mine.  I keep dragging my feet... I have so many things to measure.  Requests, and I want to get back to the DSC2 and do a thorough job checking its performance out.

This one I am not exactly certain but I also have a feeling with the dsc2 regular 5EC or 7EC modulators are cleaner than the 512fs variants possibly due to the 512fs being cut down from 200khz to 100khz… 

5 hours ago, Andrew Allen Ballew said:

YEAH..

 

There are a couple things.  You first have to compensate in the design of the analog output.  The iFi DSD1793 chip has a discrete analog FIR filter for all DSD playback.  I won't go into that in depth.  What is important is, as you double the DSD bitstream rate, the filter cutoff doubles too.  The higher the filter cutoff, the more likely that ultrasonic noise passes into the analog stage, creating intermodulation distortion in the audible frequencies.  As far as the ZEN series of iFi products, based on measurements and educated guessing, the FIR filter is set at just under 80khz at DSD64.  Double the speed to 128, its corner frequency is at 160khz.  At DSD256, we are at 320khz, but by DSD512, its a very, very high 640khz.  Its just how these CIC type Analog FIR filters work.  Even though the noise is shaped to higher frequencies at DSD512---- at some point, something has to give.  I used a 7th order modulator in that test.  It was done in multitone.  I have DSD512 test tones made in HQPlayer that I have not used to test yet.  I am writing a followup article about this specific phenomenon.  

Oh, it is not unique to iFi.  Not at all.  As we are now measuring for these things, drops in both SNR and SINAD/THD+N are common in highspeed DSD converted 'natively' with discrete analog filters.  

There is another reason too, that Andreas Koch alluded too awhile ago about how once the bitstream hits a certain speed, the logic in the DAC loses precision.  It just can't accurately do its job as well at those high speeds.  The tolerances are not tight enough.  At the time he stated this, it was disputed.  And he also addressed what I mentioned already about distortion starting to creep down into the audible band because of the nature of analog filtering.   But again, our measurements are bearing this out now we are getting access to more NATIVE DSD devices.  

Here is his article on the issue:
https://positive-feedback.com/audio-discourse/raising-the-sample-rate-of-dsd-is-there-a-sweet-spot/



Thorsten Loesch has also made mention of this with the iDSD PRO.  Turns out it is one of the best at keeping consistent THD+N numbers, but at DSD1024 there is a measurable noise drop in compared to DSD512.  Other iFi products has always been very consistent as well up to DSD256.  More so than some others I have measured.  But when DSD512 was added as a feature to the more affordable stuff, a noteable drop is seen.  

I know this all sounds bad, but as for iFi products, they have been among the most consistent in this regard.  The have always shown very little change from DSD64 up to DSD256, and on the pro little change all the way up to DSD512.. but some degradation at DSD1024.  

This is just my opinion, but simply adding on the ability to convert DSD512 in the ZEN without addressing the stress it adds to the analog output is a bit meh, but I can't be too critical because its more a greater technological issue than any negligence, and beside, it would add more cost to make changes to the analog stage that might help address it somewhat.  


P.S

This is something you won't see when DSD is digitally filtered and remodulated internally by the DAC, such as with ESS, or with AKM when the bypass mode isn't engaged.  The measurements will be basically the same at all rates, PCM and DSD, because you are now measuring the characteristics of the multi-bit delta sigma modulator, and the onboard fully digital filters that can be perhaps more flexible and capable of dealing with ultrasonic noise.  Also, I don't think you will find any multi-bit DSM's running at 45MHZ bitstream speeds either!  They have no need to, because they can achieve significantly high levels of SNR running at no more than 256x speeds.

My deepest appreciation for your elaborate answer, @Andrew Allen Ballew. Thank you for taking the time to respond to the question at length.

Thus, it is
A) a chip thing
B) an analog filter thing
I would assume, until DSD256, the difference in fidelity may be inaudible for human hearing, but the loss of fidelity through noise with DSD512 could have an audible effect?
regards, Tom

I'm  an iFi fan, and I enjoyed the review, but I'm always disappointed when a review doesn't include a comparison to a competitive product.

22 hours ago, Andrew Allen Ballew said:

You may read the full review here posted in this thread, but if you wish to see the full review with test graphs and interpretation, you can see it at https://www.euphonicreview.com/reviews.html

 

When the sun dips below the horizon and the world quiets down, true audio aficionados know it’s the perfect time to indulge in their favorite tunes. Enter the iFi ZEN DAC V3, my nocturnal companion in this auditory adventure. In this review, we'll explore whether this sleek little device lives up to its promise of delivering high-fidelity sound that could make even a bat envious of its precision. Join me as I dissect the ZEN DAC V3’s performance, unravel its quirks, and see if it truly enhances those cherished nighttime listening sessions.

One of my very favorite easy entries into to the high fidelity head-fi world is the iFi ZEN DAC V2.  It has a more than capable DAC with a unique architecture that lends to its beguiling good sound at its low price.  The built-in headphone amp is also of good quality, with an impedance and gain matching function called 'Power Match' that widens the range of headphones it can drive well.  Even some typically harder to drive, high impedance headphones will respond decently well.   

The ZEN V3 is more of the same, but better.  The DAC section is the same as before, and measurements show there is not much of a new story to tell here, although subjective experience says there is just a bit more refinement to the sound.  What is truly an upgrade here is the headphone amp.  Just as with the previous V2 version, the Power Match feature does this piece of kit a great service, as it does the ear a great service.  I found no difficulty driving higher impedance headphones, and low sensitivity models such as the Hifiman HE-560 had more than adequate drive.  

Continuing on the features front (click here for more details about the iFi ZEN DAC V3 feature set in whole), three large changes I will highlight are the change to USB-C input, the upgrade of the headphone amp processing, and the addition of DXD 2x and DSD512 capability. 

The V2 had a bass boost called 'TRUEBASS', which was basically a shelf EQ that lifted the frequency response from 200hz down to 20hz.  The V3 has the 'XBass+' feature, which is like the more refined processing found in the ZEN Can Headphone Amplifier.  It is a more nuanced and natural sounding effect that really enhances the listening experience, especially if you are a basshead. Or maybe you just feel like some extra groove one day without overdoing it.  'XBass+' is your ticket.  

Another major feature change is just part of an industry standardization to universal usage of USB-C.  I don't like it.  Call me and old man resistant to change as I approach 50, but I already have a slew of USB cables and I am a bit resentful knowing eventually they will all have to be changed to these little smart phone sized connectors.  We are doing quite well, though, that this is my only nitpick worth mentioning about the iFi ZEN DAC V3.  I even like its slightly darker, multi toned color scheme.  It really looks good.  iFi aesthetics have really come a long, long way from a decade ago with the introduction of the iDAC. (Click here for a trip down legacy lane.) 

Finally, there is now an option for DXD 2x and DSD512 playback.  As there are no native DXD 2x or DSD512 recordings I can think of, this is a feature directed at the growing set of audiophiles who are using programs like HQplayer to upsample all their music to 768khz, or DSD512 and higher to achieve performance that the onboard DAC cannot achieve with its own oversampling filters.  I did do some experimentation on my own, and I can confirm that DXD 2x and DSD512 playback worked perfectly with HQPlayer via direct PC connection.  (HQPlayer also works perfectly with the iFi ZEN STREAM, which has its own dedicated HQPlayer mode.  However, the STREAM is limited to transmission of DXD 1x and DSD256.)

SO HOW DOES SHE HUM ALONG, CAPTAIN?  

Once you have a look at the measurements at the end of the review, you might conclude the ZEN V2 and ZEN V3 sound exactly the same.  This would be an incorrect assumption, especially via the headphone output.  Nor is the sound difference limited to the headphone output.  The sound is a notable step upward from the V2, which was already my go-to standard in this price range.  And still would be even if the price were somewhat higher.  

The following listening impressions were using the V3 powered by its 5v USB bus through an Intona USB 2.0 Galvanic isolator, and with a ZEN Can headphone amplifier with neither of the analog processing enhancements activated for the majority of the evaluations.  The 4.4mm Pentaconn balanced connection was used to connect the ZEN DAC V3 to the ZEN CAN amp.  Playback source was Roon connected by both direct PC connection with WireWorld USB cables, and via the iFi ZEN Stream connected via ethernet and a WireWorld USB cable.   

I started with "Django" from " European Concert, Vol. 1" by The Modern Jazz Quartet in FLAC at 192khz 24bit.  The opening setting is extremely atmospheric with an excellent sense of stage. The Vibraphone has a luscious and full tone, with vibrato swells rendered to perfection.  Bass is not too strong, not too light, and has an excellent sense of its own playing space with good separation across the soundstage.  The Upright Bass follows in parallel with the piano, but does not dominate the line.  The ZEN is good enough to render the piano slightly farther back in the soundstage that the Bass, but at the same time allows the excellent musicianship that can fool the listener into the two merging into one organic instrument.  Before the drums kick in, the tiny detail of someone moving in their seat, which adds to the charm of this excellent track, is rendered just a naturally as if one was there in the midst of the recording space.  I am not 100 percent sure which chair makes the noise, but the precise imaging suggests it’s the drummer changing his posture as he is getting ready to play.  There is resolution measured on paper, and there is ‘musical’ resolution that is important to actual listening, and this is a fine example.  As the drums make their entrance, the quartet is completed, and the ZEN V3 unfolds the soundstage as well as anything in recent memory, including many more expensive products.  The drum brushes timbre is spot on, not sounding too bright and thin, nor too flat, lifeless and undetailed.  Natural is the word of the day.  One could listen to this for hours and hours and never experience any fatigue whatsoever.  At the same time, one would not be left wanting for any sense of detail or resolution. 

 
I next turned to "Sibelius Works for Violin in Orchestra", Ondine label mastered by 2XHD in DSD64 'from original DSD recording', listening to a remaster of Pekka Kuusisto playing Sibelius' "Humoresque no. 1 in D minor" along with the Tapiola Sinfonietta.  Normally this kind of recording, with its massive chain of conversions and DSP, is not my cup of tea.  I must admit however that there is some kind of voodoo in what 2x mastering engineers did, because there is such an abundance of texture, timbre, and an intangible presence or weight in the instrument sound that one just ‘knows’ when it is heard. 

The mastering process begins with an original Ondine label DSD64 master, converted to analog by an unnamed Digital to Analog Converter.  This analog signal is sent through a “high-end tube preamplifier” before being re-recorded in DXD by a DCS 905 ADC, and is redistributed in either DSD or PCM at 24/192. 

On paper, it seems they make a mess of things.  Not so much to the ear.  Via the ZEN DAC V3, I can hear the very distinctive individual timbre of Kuusisto’s violin.  Even the best violins on most recordings take on a somewhat generic sound, but not this time.  There is a distinctive slight ‘grain’ in the timbre, which is full bodied, never thin.  It is heavy on the overtones.  Exactly what my ears like to hear, measurements or accuracy be damned.   Either the original recording started this way, or the re-mastering process is responsible, but the soundstage here is quite intimate and a bit small, however the violin centerpiece is never intimate and small.  It dominates your attention and draws you in.  The ZEN DAC V3 takes what is on this recording and paints it tonally in the best of light.  Just out of curiosity, I did some research on what kind of violin Kuusisto may have been playing that night, as its tone struck me so deeply.  It isn’t his ‘Scotta’ violin made by Stradivari in his ‘Golden Period’ ca. 1709.  Even still, my curiosity was piqued.  Even if not the Scotta Strad, it is a world-class instrument.   (I wonder if Stradivari, if alive today, would be greatly insulted if told his instruments didn’t measure well enough.  Not enough SINAD, Vecchio Uomo.)  Frankly while using this track for review, I simply got lost in the music, the player and the instrument and its collective pure beauty.  I think that says all I need to say about the role of the ZEN DAC V3 in this case.  It transmitted the music, which is all I can ask of any piece of gear and at the same time is the highest compliment I can give. 

Finally, I needed to audition something that ‘rocked’ a bit to see how the DAC handles lesser recordings with a wall of sound.  I went to Tidal and cued up Lil’ Wayne’s duet with Nicki Minaj “What’s Wrong With Them”.  There is nothing audiophile here.  However, the ZEN DAC V3 completely holds it composure.  The powerful bass track never overwhelms or affects the vocal tracks, whether main or backing track, in any way.  The ‘wall of sound’ is powerful, yet everything stays in its lane, so to speak.  Lil’ Wayne’s rapping is as clear and articulated as anyone could ask.  Often I find hip-hop mixes to be too bright.  I think these guys doing the mixing are often half-deaf from too many years grinding their way to the top in the clubs. I guess its a price to be paid. 

There is nothing bright here in this Lil' Wayne mix. It sounds about as natural as a modern processed track can sound. 

To this point in time, as mentioned earlier, I was listening via the ZEN CAN, bypassing the ZEN DAC V3 headphone amp.  The reason I began auditioning this way is to get a baseline comparison with other DACs in my collection that occupy a similar price tier in the market.  To get a handle on how good the onboard headphone amp is, I switched at this time, and re-auditioned a couple tracks.  Headphones were still the Sennheiser HD650. 

Lil’ Wayne’s track “What’s Wrong with Them” sounded essentially the same, however, it was not presented with quite the same ease.  His lead rap didn’t stay quite as precise and centered in the mix, and the tone of his voice wasn’t quite as consistent.  Bass was not quite as strong either, that is, until I took the first opportunity to try the ‘XBass+’ feature.  (I had used the original XBass feature on the iCan while auditioning this track.)  IMPRESSIVE.  Now the sound was much more like what I heard through the iCan.  Not only did the XBass+ feature increase the power of the bass without any added bloat or boom, it seemed to have the side effect of creating more precision in the track overall.  It still didn’t reach the level of refinement of the ZEN Can, but it was impressive nonetheless.

The Sibelius Humoresque still sounded quite good through the ZEN DAC V3 onboard headamp, but the violin didn’t quite jump out at me like it did via the ZEN iCan.  It was still a very, very sweet sound with most of the same good qualities I heard through the iCan. 

There is a rumor that a new ZEN ICan is possibly on its way to us as well.  If it matches and even slightly exceeds the pairing of the ZEN DAC V3 with the current iCan, then that possibly will be the head-fi value of the year. 


CONCLUSION

Don't be fooled into thinking my criticism of the ZEN V3 onboard headamp compared to the dedicated ZEN headamp is a negative critique.  The reality is the onboard headamp is as good as in any integrated piece I have heard at this price.  It is an upgrade from the ZEN V2, and the 'XBass+' processing is a big part of that.  Its contribution to the overall presentation is very impressive.  Overall the iFi ZEN DAC V3, in spite of its many similarities with its predecessors, is a worthwhile upgrade.  If you already own the V2, you know as I do how already very, very good it is.  Normally I would suggest waiting another generation before you will find any significant upgrades.  In THIS Case, I would suggest just go ahead and upgrade now.  I find the ZEN V3 to simply be an exceptional player of music, and at this price range I can only think of products with similar fidelity that DO NOT have a built-in headphone amp of any kind, no to mention one that might have one as high quality as this!  Well done iFi.  

EASY PICK FOR EDITOR'S CHOICE STATUS.  

OUR RATING:  (0-10 scale.  8 or higher numbers represent best of industry.  For the final score calculation, each category is given a (undisclosed) weighting and is tallied for final score.   Although perfect 'fairness' is unlikely, our system tries to be as unbiased as possible.  

sound quality: 8.9
build quality: 9.0
ease of use: 9.5
measurements: 8.8
value: 10.0

TOTAL RATING:   91.7


 

---

 

---

 

MEASUREMENTS 

As is the usual case with iFi products, measurements are consistent, very good, and have no 'red flags' to indicate any issues. iFi creates very well thought out and engineered products.  

Frequency Response (20hz to 20khz): dB -0.35, +0.01
SNR:  -117.8db A-weighted
THD:  -110.3db / 0.000305%
THD+N: -100.7db / 0.000923%
IMD: -108db / 0.000398%
JITTER: 40.4ps Peak / 13.6ps RMS
LINEARITY at -100db: <0.1db error
LINEARITY at -111db: <0.5db error
LINEARITY at -115db: <2db error
THD+N Multitone 32 Test: -95.3db / 0.001718%
________________
DSD MEASUREMENTS

DSD64 THD+N: -100.7db / 0.000923%
DSD128 THD+N: -100.5db / 0.000944%
DSD256 THD+N: -94.8db / 0.00182%
DSD512 THD+N: -87.5db / 0.004217%

DSD64 JITTER: 74ps Peak / 35.4ps RMS

 

Hi Andrew, would you say for the zen v3, dsd128 is likely the sweet spot as you gain effectively 6db in audible band noise compared to dsd256?

2 hours ago, dericchan1 said:

Hi Andrew, would you say for the zen v3, dsd128 is likely the sweet spot as you gain effectively 6db in audible band noise compared to dsd256?

By the measurements, yes.  

But the source of the THD+N drop in the iFi up to DSD512 isn't so much just an increase in random noise in the audible range.  It's more from an increase in harmonic distortion.  Primarily second order.  

There are resolution advantages in the higher rates still.  I personally don't mind a little bit more touch of low level harmonic distortion.  I listen via a tube amp lol.  So the levels of harmonic distortion even at DSD512 in this case would still be lower than what I am listening to.  

I have seen other DACs where the increase in THD+N is definitely noise from intermodulation, however, which is not good.  






 

8 hours ago, DuckToller said:

My deepest appreciation for your elaborate answer, @Andrew Allen Ballew. Thank you for taking the time to respond to the question at length.

Thus, it is
A) a chip thing
B) an analog filter thing
I would assume, until DSD256, the difference in fidelity may be inaudible for human hearing, but the loss of fidelity through noise with DSD512 could have an audible effect?
regards, Tom

not just the analog filter by itself.  Its the fact even at the higher DSD rates, their rise in ultrasonic noise, even with a high order modulator, starts at a lower frequency.  So the filter does its job well, its just it starts working above the point where the ultrasonic noise starts its increase, meaning it can't cut it out and it's passed into the actual analog amplification stage, which reacts poorly to the noise, creating various types of distortion and noise.  The better then analog stage, however, the better it can deal with these problems on its own.  

Yes, the reason the iFi products by default peak at around -100db THD+N is the chip used is not capable of better.  There are those of us that think 90 to 100 decibels down is plenty good enough.  I am not sure the so called SINAD wars are very fruitful in actually providing us with better, high quality, long lasting, truly unique (and subjectively good sounding) products.  

what is actually audible is a good question.  I don't have a good answer lol.  Compared to the history of audio reproduction, even these levels are way higher fidelity than in the past.  I personally didn't notice an audible loss of fidelity at DSD512.  It noise increase is primarily from higher low order harmonic distortion, which some people actually find pleasant.  Its one of the factors believed to make tube audio so seductive to many.  

Im not big on the SINAD/THD+N chasing bandwagon.  

I have reviewed for instance the Topping E70.  With THD+N that is lower than -120db.  But I would still listen to the iFi ZEN DAC anyday, even though its actual chipset (DSD1793 from late 1990's) has a maximum performance capabilty 20db worse than that. 

My advice is always listen first.  Don't let measurements get in your head unless they really scream something is badly wrong.   At the end of the day what you hear is what actually matters.  

40 minutes ago, Andrew Allen Ballew said:

By the measurements, yes.  

But the source of the THD+N drop in the iFi up to DSD512 isn't so much just an increase in random noise in the audible range.  It's more from an increase in harmonic distortion.  Primarily second order.  

There are resolution advantages in the higher rates still.  I personally don't mind a little bit more touch of low level harmonic distortion.  I listen via a tube amp lol.  So the levels of harmonic distortion even at DSD512 in this case would still be lower than what I am listening to.  

I have seen other DACs where the increase in THD+N is definitely noise from intermodulation, however, which is not good.  






 

As I am also using a tube preamp this is what i noticed. When I went higher dsd rates in my ifi neo say dsd512, you are now adding more harmonic distortion (pleasing or not) to the chain. So my tube preamp is providing tube distortion on already distorted signal from the neo. It sounds great for music that are easy to listen to - jazz, classical, old days love songs…

However so much details were masked by the distortion when you listened to stuff that are complicated - pop/rock, punk, hard rock, metal… often times you can hard hear what the lyrics were…

switching back to dsd256 was obviously better and seem to have brought the vocal a bit cleaner/more forward…

3 hours ago, audiobomber said:

I'm  an iFi fan, and I enjoyed the review, but I'm always disappointed when a review doesn't include a comparison to a competitive product.

Thank you.  I will take your criticism into account.  The entire review site is basically that, though.  These are all meant to be compared against one another directly through a consistent as possible rating system, an consistent measurement system (which is still unfortunately evolving and I have not been able to finalize a permanent format).  

Of course, I did make a comparison to another product in this review.  The previous iFI ZEN DAC V2.  But I understand what you mean.  At this point, I could make general observations, but I only keep on hand the data, notes, and actual review of competitive products.  I only hold onto very few actual products for a direct comparison.  Going by memory is extremely difficult if not actually impossible.  Our sound quality memories seem to last a matter of minutes at most, maybe even seconds. 

In the future, I will try and incorporate more references to other brands in similar price ranges that I have direct knowledge of, and contrast and compare.  But I have to have thoroughly reviewed it.  I can't ethically make comments about products I may have read about but have never heard for a considerable review period.  

11 minutes ago, dericchan1 said:

As I am also using a tube preamp this is what i noticed. When I went higher dsd rates in my ifi neo say dsd512, you are now adding more harmonic distortion (pleasing or not) to the chain. So my tube preamp is providing tube distortion on already distorted signal from the neo. It sounds great for music that are easy to listen to - jazz, classical, old days love songs…

 

However so much details were masked by the distortion when you listened to stuff that are complicated - pop/rock, punk, hard rock, metal… often times you can hard hear what the lyrics were…

 

switching back to dsd256 was obviously better and seem to have brought the vocal a bit cleaner/more forward…

 

so, you are finding the harmonic distortion quantitative, not masking?  I have not really put too much thought into it. 

7 minutes ago, Andrew Allen Ballew said:

so, you are finding the harmonic distortion quantitative, not masking?  I have not really put too much thought into it. 

In fact I think so. For example with the same tube preamp, my cyan 2 which is measured to have less impact on distortion at higher dsd rate than the ifi is much cleaner than the ifi. However the same rule as you pointed out applies, with the cyan 2, dsd256 is the “sweet spot”. Going to dsd512 you lose about 3db of THD+N and even that is audible. Dsd256 just sound cleaner more forward compared to dsd512

4 minutes ago, dericchan1 said:

In fact I think so. For example with the same tube preamp, my cyan 2 which is measured to have less impact on distortion at higher dsd rate than the ifi is much cleaner than the ifi. However the same rule as you pointed out applies, with the cyan 2, dsd256 is the “sweet spot”. Going to dsd512 you lose about 3db of THD+N and even that is audible. Dsd256 just sound cleaner more forward compared to dsd512

One thing I should do is actually measure the intermodulation distortion as well.  I have not done that.  It could shed even more light onto the phenomenon.  

On my iDSD PRO, I never used the oversampling to DSD1024.  It measured worse and did indeed seem to be audible.  DSD512 was what I used when I engaged the DSD oversampler.  IMO its best performance, though, was just direct conversion of the signal.  Bit-perfect.  

2 minutes ago, Andrew Allen Ballew said:

One thing I should do is actually measure the intermodulation distortion as well.  I have not done that.  It could shed even more light onto the phenomenon.  

On my iDSD PRO, I never used the oversampling to DSD1024.  It measured worse and did indeed seem to be audible.  DSD512 was what I used when I engaged the DSD oversampler.  IMO its best performance, though, was just direct conversion of the signal.  Bit-perfect.  

With the ifi pro, have you compared hqplayer upsampling to 512 vs ifi internal remastering to 512? Or when you say bit perfect, you basically just feed it pcm?

35 minutes ago, dericchan1 said:

In fact I think so. For example with the same tube preamp, my cyan 2 which is measured to have less impact on distortion at higher dsd rate than the ifi is much cleaner than the ifi. However the same rule as you pointed out applies, with the cyan 2, dsd256 is the “sweet spot”. Going to dsd512 you lose about 3db of THD+N and even that is audible. Dsd256 just sound cleaner more forward compared to dsd512

I didn't want to include this part, because it really is a slight tangent, could muck things up for people unaware of it, and quite frankly, I am not sure what I think about it.  

But it is another reason why higher DSD rates may be important, beyond simply the perceived SNR from noise shaping.

 

"I have to quality this next statement as controversial in the DSD 'world', but this comes from Thorsten Loesch, and I believe Lynn Olson was onto something similar.  I really hope I don't butcher their thoughts on this.  TL or Lynn, please feel free to publicly correct me. 

I do think there are some advantages to the increased speed that make it not a black and white choice.  One thing that is true about higher rates, since bitstream is an averaging format (essentially talking about first order noise shaped 1-bit DSM now I believe) you add 1 bit of actual resolution (independent of the resolution gain you get simply from oversampling) because it is a time splicing format. 

If you compare to 44.1hkz PCM, 2.8mhz DSD has in the same duration of a 44.1khz sample 64 pulses to represent the amplitude of that sample.  The real complication here is that higher frequencies will get much fewer pulses to render their amplitude, lower frequencies will get more pulses, therefore more resolution averaged over time.  

But to keep it simple, compared to a 44.1khz PCM signal at X bit-depth, each second DSD64 gives the average equivalent of 6bits binary resolution.  128 will increase to 7 bits, 256 to 8 bits, 512 is 9, and 1024 is 10bits.  

So DSD64 with no noise shaping (well its actually first order by nature) has 36.1db of resolution.   128 increases to 42.1db.   You get the idea.  This is the simple resolution of the format from just the time-splicing pulse averaging.  So this is the ACTUAL resolution of DSD.  Even though each 1bit pulse is only 6db, you can't look at it that way, because its a system of averaging those pulses over a set period of time to create more accurate amplitude.  And the increase in resolution over 6 decibels due to time-splicing pulse averaging is the very basis of noise shaping as the massive quantization noise of the 1-bit pulse is a a higher frequency than we can hear. 

We just keep pushing the perceived resolution higher and higher with higher orders of noise shaping.  But all that distortion?  It is still there.  

If you look at a 20khz sine wave encoded in DSD64, it will look like a complete mess.  Now, all that mess we see is at higher frequencies, but the question becomes, how accurate then is the waveform we actually hear?  The standard answer is it makes no difference because all the distortion and quantization error isn't audible.  But the error is still there.  Is the actual amplitude of the signal represented at a much higher degree of accuracy just because we put all the error noise in a place where it isn't heard?  What effect does that have on the signal left behind that we still hear?  We don't hear its error noise, but, it technically still only has a maximum of 64 pulses relative to a 44.1khz PCM file.  

I believe the answer most experts will give you is the amplitude error is the noise.  get rid of or make the noise inaudible, the amplitude is rendered correctly (or as correct as a higher quantization level would be).  

But there is still this nagging idea that, no, even if you get rid of the resulting noise in pushing it higher and higher, whats left still only has 64 pulses to get that voltage correct.  Again, just because you can't hear the error doesn't mean it isn't there.  It kind of like a slight-of-hand trick.  

And as I mentioned before, it goes even deeper than that, because higher and lower frequencies, which we standardize to the unit of a second, get different amounts of pulses!  a 20hz bass signal period is much much longer than the 44.1hz sample period.  So it will get a full 64 pulses in any 44.1khz period.  On the other hand, a 20,000khz signal has a much much shorter period (50 microseconds compared to 50,000 microseconds).  Therefore it only receives 29 pulses in any 44,1khz period!  Much less pulses to average the correct amplitude, therefore much higher quantization noise at the higher frequency.  

Now, this comes from people much, much smarter than me.  I think my math is correct.  I am no mathematician, I am a musician.  So I am usually working out of a totally different hemisphere of my brain :)


The question becomes, once the error signal is filtered out by noise shaping to higher and higher frequencies (in the end making our ears the final filter) Is what is left behind in the audible range, in spite of the massive drop in noise, still as accurate a signal as it would seem to be?

I think the answer is yes, but what does remain is a real drop in linearity as frequencies go higher.  Higher order modulators mitigate this.  

But there are others, like I said much smarter than me, who still see this as an issue.  The more 'real' resolution vs noise-shaped resolution, the better. 


This got all extremely theoretical and my conclusion doesn't quite match me giving this possible phenomenon as a potential positive reason to stay at higher rates.  But I do think it is a fact that higher speeds give better actual resolution, starting from basic 1st order averaging, before any noise shaping is needed to further improve it. "

5 hours ago, Andrew Allen Ballew said:

 

But to keep it simple, compared to a 44.1khz PCM signal at X bit-depth, each second DSD64 gives the average equivalent of 6bits binary resolution.  128 will increase to 7 bits, 256 to 8 bits, 512 is 9, and 1024 is 10bits.  

 

 

I need to correct this.  I didn't mean every second.  I meant every 44,100khz period or 22.68 microseconds,  which has 64 pulses in it in DSD64.  So its about 6 bits of actual resolution in simple averaging before higher level noise shaping expands the perceived resolution.  In a second order noise shaping, I think it would expand the perceived resolution to around 8 or 10 bits in the same period.  

This would all be moot if we could have system that had 65,536 pulses per every 22.68 microseconds.  That would mean 16 full bits of real resolution with simple averaging, no higher level noise shaping at all required.   But the bitstream would be 2,890,137,600, or approx 2.89 GIGAHERTZ, not megahertz.  So noise shaping comes to save the day.  

But the question still remains out there for me, about perceived resolution vs actual resolution.  

If you remove the quantization noise out of the audible range, you still can't increase the basic number of pulses used to represent the signal.  The pulses in a bitstream time-splicing system in a set period of time are equal to the levels available in a PCM multibit sample.  

Again, if I can't hear the noise, but its actually still there, its means the source of the noise still exists, which is amplitude error.  If its at the wrong amplitude, exactly how accurate is a noise-shaped system? 

At this point in time after studying DSD as much as possible, I ask the question as like a DEVIL'S ADVOCATE.  I am looking for a great answer.  I haven't found it yet, although I do personally believe that indeed once the amplitude error/quantization noise is removed from perception, what is left to be perceived is an accurate representation of the amplitude as the SNR ratio says it is.   

 

i would also clarify this whole thing is a thought experiment only.  DSD is not measured or divided into time periods in actuality,   But it serves to illustrate an issue that I think is worth discussion and am interested in comparative and contrasting thought

the best answer I have at this time, is this

once you can't hear the amplitude error/quantization noise, the overshoot or undershoot error is eliminated from hearing.  Therefore, what you DO hear is a correct amplitude, as correct as the SNR says it should be in the audible area.  

On 6/11/2024 at 12:36 PM, Andrew Allen Ballew said:

 

not just the analog filter by itself.  Its the fact even at the higher DSD rates, their rise in ultrasonic noise, even with a high order modulator, starts at a lower frequency.  So the filter does its job well, its just it starts working above the point where the ultrasonic noise starts its increase, meaning it can't cut it out and it's passed into the actual analog amplification stage, which reacts poorly to the noise, creating various types of distortion and noise.  The better then analog stage, however, the better it can deal with these problems on its own.  

Yes, the reason the iFi products by default peak at around -100db THD+N is the chip used is not capable of better.  There are those of us that think 90 to 100 decibels down is plenty good enough.  I am not sure the so called SINAD wars are very fruitful in actually providing us with better, high quality, long lasting, truly unique (and subjectively good sounding) products.  

They claim the pro idsd sig has lower noise, would that improve the dsd1024 performance over the original?