[Johnfg465vd]

As an overall design, despite the age (2013/14) the iDSD micro platform is much more "high end" from the word go. So I'd most certainly expect it to be much better. But it lacks Bluetooth and is rather large, so I long ago stopped carrying my black label even in luggage for travel.

The large size is the only reason I sold mine, Bluetooth is not something I care about so that wasn't an issue. If iFi can make a well implemented Dual DAC Chip DAC (similar to Signature) the size of Gryphon with similar power (~1W) that would be pretty awesome.

 

[BartekR]

Just to be clear, when I said Mojo 2 is better than the Gryphon. I'm talking specifically about it's technical capabilities like resolution/detail, separation and clarity. When it comes to tuning, I also prefer the sound signature of Gryphon. I had both Mojo & Mojo 2, they had better technical ability with almost 3D like soundstage and detail with denser and textured Bass but something about the sound was not natural to me, it could be a characteristic of the FPGA Chip.

If we are talking about tuning, Gryphon is better than Mojo 2 for me but other devices like the H7, R6 Pro II... sound more natural and correct without sacrificing and details or texture.

I actually sold my first Gryphon as I was a bit disappointed by it initially, especially coming from the micro iDSD Signature which to me sounded better. I later on bought it again to see if maybe I was being too harsh on the little AIO. I ended up not using it cause to me, the other DACs I have sounded better. I think it's a good AIO but compared to N7, N8ii... No way it stands a chance. In it's own price range though, it is a very good product. You will get better performance for just a bit more though with the likes of H7, Angel, Q7...

Mojo2 sounds more "studio" - it's not 'sexy' for pleasure listening.
Even tried to eq it to a very simillar sound like Gryphon with X-Bass+Presence,
but it lacks that something special Gryphon delivers.
Clinical is not something I am looking for, so Gryphon suits me more.
I also really like Questyle M15 - very enjoyable, especially with IEMs like Fir Xe6 or Dorado2020.

 

[Ferdinando1968]

@Thorsten Loesch

Hi Thor.
I have a question related to SPDIF/COAX and MQA.

What makes Gryphon able to receive MQA via SPDIF ?
Why don't other devices (even IFI) do it?
Is a particular circuit necessary?

 

[Johnfg465vd]

I didn't hear any improvement when using Optical with Gryphon but there was a noticeable improvement when using the same with the Mojo 2 compared to USB. I think it's a good thing that there is no noticeable improvement, it means what ever filteration Gryphon is using for USB is pretty good.

 

[Thorsten Loesch]

@Thorsten Loesch

Hi Thor.
I have a question related to SPDIF/COAX and MQA.

What makes Gryphon able to receive MQA via SPDIF ?

Have you confirmed it can do so? If so it would use the 16-Core XMOS platform similar to iDSD Neo.

Why don't other devices (even IFI) do it?
Is a particular circuit necessary?

There are two generations of XMOS devices and two "sizes".

All iFi products up to around 2018 used the XU100 series 8-Core device. They all literally used the same actual hardware and shared one unified firmware.

An 8-Core device is not powerful enough to act as MQA decoder, it can only act as render (not that this really matters) and while in theory it is possible to get a decoded MQA stream via SPDIF, there was simply no way to make this run on 8-Core hardware. It was actually quite an effort together with the MQA software team in the UK to get something that worked.

Starting with the iDSD Pro iFi started deploying the XU200 series and a 16-Core device. This device can handle full MQA decoding and also can be used as SPDIF decoder with an external clock synthesiser and a lot more functionality.

The unit that became later known as iDSD Neo was actually the first unit of what was meant to be the next generation of iFI products, using a lot of the tech from the x Series (which the iDSD Neo has) and a 16-Core XU200 Series XMOS including BT and SPDIF across the board.

This would have included a replacement for the iDSD micro as a smaller, more portable version with improved performance and a "micro" level desktop unit (the Neo and xDSD were designed as portable & desktop replacements for the nano range) to fill the huge ga below the iDSD Pro.

The older tech was recycled in a low cost range embodied by the Zen Range and the "Hip DAC", which were designed using most of the existing micro and nano tech. The Zen DAC is in effect the iDAC 2 micro with minor improvements and the "Hip DAC" is the same but with a battery.

As it happened, iFi decided not to invest what was needed to get "The Next Generation" range developed and into production. And XMOS decided to cancel the 8 Core XU100 unit.

This meant iFi had somehow keep selling the existing "micro" unit(s) so they got a facelift into the "Diablo" and "Signature/Finale" and had to get a backwards compatible hardware replacement for the XU100 hardware circuit board using a XU200 device including the required firmware rewrite.

By that time I had already left iFi and was only working as external consultant on very limited hours, which I eventually also stopped. I saw the beginning of the XU200 replacement hardware on my final way out and this was painful to watch.

The 16-Core XMOS was also retrofitted on low cost platforms like Zen DAC V2 (and Zen ONE) and Hip DAC V2 as it was actually cheaper than the 8-Core it replaced. However for whatever reasons these platforms did not receive the necessary hardware changes (not very many TBH) to allow SPDIF input int the XMOS.

So even though these devices have a 16-Core XU-200 XMOS, they do not have the hardware to support SPDIF.

Thor

 

[Ferdinando1968]

Have you confirmed it can do so? If so it would use the 16-Core XMOS platform similar to iDSD Neo.



There are two generations of XMOS devices and two "sizes".

All iFi products up to around 2018 used the XU100 series 8-Core device. They all literally used the same actual hardware and shared one unified firmware.

An 8-Core device is not powerful enough to act as MQA decoder, it can only act as render (not that this really matters) and while in theory it is possible to get a decoded MQA stream via SPDIF, there was simply no way to make this run on 8-Core hardware. It was actually quite an effort together with the MQA software team in the UK to get something that worked.

Starting with the iDSD Pro iFi started deploying the XU200 series and a 16-Core device. This device can handle full MQA decoding and also can be used as SPDIF decoder with an external clock synthesiser and a lot more functionality.

The unit that became later known as iDSD Neo was actually the first unit of what was meant to be the next generation of iFI products, using a lot of the tech from the x Series (which the iDSD Neo has) and a 16-Core XU200 Series XMOS including BT and SPDIF across the board.

This would have included a replacement for the iDSD micro as a smaller, more portable version with improved performance and a "micro" level desktop unit (the Neo and xDSD were designed as portable & desktop replacements for the nano range) to fill the huge ga below the iDSD Pro.

The older tech was recycled in a low cost range embodied by the Zen Range and the "Hip DAC", which were designed using most of the existing micro and nano tech. The Zen DAC is in effect the iDAC 2 micro with minor improvements and the "Hip DAC" is the same but with a battery.

As it happened, iFi decided not to invest what was needed to get "The Next Generation" range developed and into production. And XMOS decided to cancel the 8 Core XU100 unit.

This meant iFi had somehow keep selling the existing "micro" unit(s) so they got a facelift into the "Diablo" and "Signature/Finale" and had to get a backwards compatible hardware replacement for the XU100 hardware circuit board using a XU200 device including the required firmware rewrite.

By that time I had already left iFi and was only working as external consultant on very limited hours, which I eventually also stopped. I saw the beginning of the XU200 replacement hardware on my final way out and this was painful to watch.

The 16-Core XMOS was also retrofitted on low cost platforms like Zen DAC V2 (and Zen ONE) and Hip DAC V2 as it was actually cheaper than the 8-Core it replaced. However for whatever reasons these platforms did not receive the necessary hardware changes (not very many TBH) to allow SPDIF input int the XMOS.

So even though these devices have a 16-Core XU-200 XMOS, they do not have the hardware to support SPDIF.

Thor

Oh yes, I'm sure.
I show you the optical signal from a streamer configured as MQA pass-through going into Gryphon:

I won't show it to you, but if I configure the streamer to perform the first unfold, instead of MQA, OFS would appear.

If I do the same test on the Signature (which only supports the render function, so I have to perform the first unfold outside it), again with the streamer configured to perform the first unfold, instead of the purple MQA light, the LED remains green (PCM signal ):

Thanks for the explanations, but I'm confused.

In addition to the motivation linked to the different power of the X8 and X16 processors (which obviously produces differences in the processing capacity), I am unable to understand what is the technical obstacle of the failure to decode via SPDIF.

I mean, basically MQA we know to be a PCM FLAC with an internal subcoding that concerns certain bits (LSB, if I remember correctly).

What could possibly change on SPDIF between receiving a PCM with certain bits encoded or not?
After all, these are bits in all respects the same as PCM ones.

No ?

 

[Thorsten Loesch]

Oh yes, I'm sure.

Ok. It likely shares the XMOS circuitry with the iDSD Neo which was designed to include this and shares a lot of circuitry with the xDSD.

I am unable to understand what is the technical obstacle of the failure to decode via SPDIF.

MQA decoding / rendering (and DoP DSD decoding) are a function of the XMOS circuit. The signal must pass "through" XMOS for this decoding. You need to make the necessary connections between the SPDIF section and XMOS and the necessary connections between XMOS and the clock system.

In the older platform the SPDIF section and Clock are stand alone and are programmed from the MCU that controls the rest of the system. The decoded SPDIF signal is never routed anywhere near XMOS but passes directly to the DAC.

As there is no direct connection to the XMOS section and thus the signal cannot get to XMOS to be detected and the clock cannot be managed by XMOS to allow XMOS to decode SPDIF.

It was not even considered in 2013 when the first generation (XU100) ifi XMOS USB design was implemented (nor was MQA even considered). By 2018 it had been considered and was a solved problem as I had always seen this as an unwanted limitation.

In the new platform the 16 Core XMOS actually decodes SPDIF (or rather it can) and the clock is controlled directly by XMOS. It is a more advanced system which however needs dedicated software support. But this can not be used if "retrofitted" into the older platform, necessary hardware does not exist.

It would not be hard (at least for me) to make the necessary changes, but it needs to be designed correctly and there are limits to "copy/paste" engineering in this case, you cannot wholesale take a block from some other product and dump it in.

Thor

 

[Ferdinando1968]

As there is no direct connection to the XMOS section and thus the signal cannot get to XMOS to be detected and the clock cannot be managed by XMOS to allow XMOS to decode SPDIF.

Thor

Now is clear.
Thanks Thor.

 

[Centauri]

What's the verdict on using EQ in concert with Xbass/space? I've recently discovered the wonders of Auto EQ, but I'm worried that combining these things is... bad?

 

[Thorsten Loesch]

I've recently discovered the wonders of Auto EQ, but I'm worried that combining these things is... bad?

XBass is an analogue bass boost. It's effect / curve has been documented.

The big advantage of doing boost in the analogue domain is that a digital EQ cannot actually boost levels, as a 0dBFS signal will clip heavily. So to boost a given frequency range by (say) 12 dB a digital EQ must lower the whole level by 12dB to avoid clipping, or 2 Bits worth of resolution. So ideally digital EQ is left to "cut" use.

Analogue EQ can boost however, without such consequences at sensible volume settings. If you subtract the Bass boost from X-Bass from the desired EQ you will have less or no boost in the digital EQ which is a good thing.

XSpace is a crossfeed, based not on the traditional (Bauer/Linkwitz/Meyer) coefficients which are wrong (hence usually most people do not really like this crossfeed) but instead based on a generalised HRTF derived using dummy heads and multiple live subjects (sound engineers), so it is not like the common (and not correct) crossfeed options. It can be used with digital eq etc. - no problems.

I've recently discovered the wonders of Auto EQ

Take with a bit of salt. And remember to use a target that suits your taste (e.g. Harman, diffuse field etc.) and to subtract the XBass boost.

Thor

 

[Jake2]

Make sure to match levels, charge batteries, run the xDSD for a few days (no need to listen) and clear your mind of expectations.

Make sure Firmware loaded is the "non-MQA" version and that the filter selection is identical.

The switch on the xDSD is mislabelled, listen actually activates the "sharp" filter while "measure" activates the soft filter.

So:

Listen = STD on Gryphon
Measure = MIN on Gryphon

I do not suggest testing blind, but do not expect "more expensive = better" or "newer = better", or you will hear exactly that:

https://tu-dresden.de/mn/psychologie/ifap/kknw/die-professur/news/we-hear-what-we-expect-to-hear



I rather doubt that.

Why? The SPDIF input, at least on xDSD uses the same circuitry as is used in the SPDIF iPurifier. In other words it is essentially absolutely impervious to huge levels of jitter. And the internal USB Audio system is build around the XMOS as U2 uses and the clock is EXACTLY the same as drives the SPDIF side.

So connecting USB directly or via an external USB Audio device and then via SPDIF will not show any differences in jitter etc. for the xDSD (which I designed). Unless someone made a major mistake implementing the USB subsystem in the Gryphon, I'd expect the same results as the clock / SPDIF subsystem I designed into the xDSD is clearly still present in the Gryphon.

And if it works for SPDIF, how can it NOT work for USB (USB is asynchronous and the clock driving the DAC & reclockers inside the gryphon do not change from SPDIF to USB).

So unless something else changes the signal (e.g. the digital filter processing inside the "gryphon" on USB which is absent on SPDIF AFAIK) there can physically be no difference in signal jitter at the DAC and thus no jitter related changes in sound.

Thor

Thanks for the feedback. I'll defer to you re technical aspects of their design - interesting to get your unique insights on the design history here, really appreciate it. I do stand by my impressions re the Douk via toslink vs USB input into Gryphon - I'm not talking subtle or marginal differences here.

Given the difference I heard I assumed it must have been due to the Gryphon using its internal clock to reclock the asynchronous USB input versus using the clock signal embedded in the synchronous s/pdif stream, but not buffering and reclocking it.

I should also mention I have two versions of the Douk - one with a better oscillator, the Crystek CCHD957 which sounded better still, and whose sound I was describing in my post above. Using a Holo Titanis USB reclocker ahead of the Douk improved the sound further. I posted about my observations re this chain some time ago upthread (I'll see if I can find the post). The audible improvements - especially with the Crystek Douk were of similar character to those I've experienced with external clocks and other DDCs such as those in my sig, so are to my mind very likely explicable by the greater temporal precision (lower jitter) of the s/pdif signal fed to the Gryphon. And likely also to some extent the 100% galvanic isolation of toslink vs USB borne noise, though this is only relevant to the delta between s/pdif and USB, not the marked difference between the two Douks (and with /without the upstream Holo Titanis) which both fed the Gryphon via toslink.

I have a couple of s/pdif iPurifiers (the OG and the 2) and whilst superb and indispensible I wouldn't say they are quite impervious or perfect (what is!?), as they can definitely benefit from an even cleaner power supply (the iFi Power X + DC iPurifier 2 is a fab combo with the V2 s/pdif iPurifier feeding my R26 DAC) and a cleaner upstream signal source. I made a series of improvements to my streamer which routinely flow audibly through the iPurifier to my DAC.

Edit - I found my listening impressions post - it was over in the Douk thread:

https://www.head-fi.org/threads/dou...ble-clean-digital-signal.958462/post-17488646

 

[Thorsten Loesch]

I do stand by my impressions re the Douk via toslink vs USB input into Gryphon - I'm not talking subtle or marginal differences here.

I do not doubt the least that you are hearing differences. HOWEVER, I note that jitter SHOULD not be the cause. It may be worth chasing down the real cause of the difference.

Given the difference I heard I assumed it must have been due to the Gryphon using its internal clock to reclock the asynchronous USB input versus using the clock signal embedded in the synchronous s/pdif stream, but not buffering and reclocking it.

That is not the case, the clock is always the same and the recovered clock still runs on the same programmable local clock. The system is designed to completely reject any external clock jitter or phase noise etc. This system is actually common to any iFi product with an SPDIF input. SPDIF input signals are buffered and re-clocked internally, the external clock is completely discarded and cannot have any effect.

I have a couple of s/pdif iPurifiers (the OG and the 2) and whilst superb and indispensible I wouldn't say they are quite impervious or perfect (what is!?), as they can definitely benefit from an even cleaner power supply (the iFi Power X + DC iPurifier 2 is a fab combo with the V2 s/pdif iPurifier feeding my R26 DAC) and a cleaner upstream signal source.

Power supply noise can very much have an effect on the OUTUT jitter with the SPDIF iPurifier.

We tested the SPDIF iPurifier with HUGE levels of jitter, to a point that a very high end DAC could not even lock onto the signal, nothing as present in the output.

https://audiophilestyle.com/forums/...pdif-ipurifier/?do=findComment&comment=613811

Is it perfect? No. It's pretty cheap.

The local reference clock could be better (Ovenised OCXO - for ten times the selling price of the iPurifer perhaps?), power supply could be better. Don't forget it used to be under 150 Bux...

However anything in front of it is eliminated in terms of jitter and the decoding of SPDIF is extremely robust. Even Amir found that, though he was caught out again by the limited CMRR of the AP he uses and the presence of a powersupply with common mode noise.

So based on all I know, I still suspect the actual audible difference, which I do not dispute is there, is not related to clock jitter. And being curious, I'd like to find out what causes it. I could imagine you might feel the same?

Thor

 

[Centauri]

XBass is an analogue bass boost. It's effect / curve has been documented.

The big advantage of doing boost in the analogue domain is that a digital EQ cannot actually boost levels, as a 0dBFS signal will clip heavily. So to boost a given frequency range by (say) 12 dB a digital EQ must lower the whole level by 12dB to avoid clipping, or 2 Bits worth of resolution. So ideally digital EQ is left to "cut" use.

Analogue EQ can boost however, without such consequences at sensible volume settings. If you subtract the Bass boost from X-Bass from the desired EQ you will have less or no boost in the digital EQ which is a good thing.

XSpace is a crossfeed, based not on the traditional (Bauer/Linkwitz/Meyer) coefficients which are wrong (hence usually most people do not really like this crossfeed) but instead based on a generalised HRTF derived using dummy heads and multiple live subjects (sound engineers), so it is not like the common (and not correct) crossfeed options. It can be used with digital eq etc. - no problems.



Take with a bit of salt. And remember to use a target that suits your taste (e.g. Harman, diffuse field etc.) and to subtract the XBass boost.

Thor

Thank you for the breakdown.

 

[Jake2]

I do not doubt the least that you are hearing differences. HOWEVER, I note that jitter SHOULD not be the cause. It may be worth chasing down the real cause of the difference.



That is not the case, the clock is always the same and the recovered clock still runs on the same programmable local clock. The system is designed to completely reject any external clock jitter or phase noise etc. This system is actually common to any iFi product with an SPDIF input. SPDIF input signals are buffered and re-clocked internally, the external clock is completely discarded and cannot have any effect.



Power supply noise can very much have an effect on the OUTUT jitter with the SPDIF iPurifier.

We tested the SPDIF iPurifier with HUGE levels of jitter, to a point that a very high end DAC could not even lock onto the signal, nothing as present in the output.

https://audiophilestyle.com/forums/...pdif-ipurifier/?do=findComment&comment=613811

Is it perfect? No. It's pretty cheap.

The local reference clock could be better (Ovenised OCXO - for ten times the selling price of the iPurifer perhaps?), power supply could be better. Don't forget it used to be under 150 Bux...

However anything in front of it is eliminated in terms of jitter and the decoding of SPDIF is extremely robust. Even Amir found that, though he was caught out again by the limited CMRR of the AP he uses and the presence of a powersupply with common mode noise.

So based on all I know, I still suspect the actual audible difference, which I do not dispute is there, is not related to clock jitter. And being curious, I'd like to find out what causes it. I could imagine you might feel the same?

Thor

Cheers again for the design insights.

I should clarify I'm in no way critiquing the iFi s/pdif iPurifiers, especially the V2, they're highly effective and great bang for buck. As is the DC iPurifier which I'm seriously considering getting another of. Ifi do some great stuff, for which I'm an enthusiastic advocate.

Back to the Gryphon, so just to check my impressions as it's been a while since I did any ABs - I don't normally have the Gryphon hooked up like that - in case they'd been a bit exaggerated and on a later listen were more subtle as can be the case, human nature being what it is - I had a quick listen last night.

I alternated between USB direct to Gryphon from iPad vs Toslink from the Douk with/without the Holo Titanis in the USB chain from the iPad. Sure enough the difference is as I described it - particularly with the Titanis which must provide a much cleaner USB feed to the Douk making its job that much easier. With the Titanis/Douk/Gryphon chain the sound was significantly more expansive and dynamic, with more depth and separation between instruments, bass notes were fuller and more impactful - more juicy - whilst female vocals were more palpable and textured. With the double bass on "Gavrilo's Prinzip" on the album Bass Room by Nenad Vasilic, its timbre was richer, with a more extended and nuanced reverb and decay.

The scale of the difference is greater than when say I hook up the U18 DDC or a master clock to my R26 DAC. I wish I could share it in person with you and other fellow hifi enthusiasts. Indeed the delta is so pronounced - the quality so high - I am tempted to show my non-hifi mates who can't not hear it and appreciate it, especially with that Double Bass track. In fact I may try to take a vid to capture the delta via speakers though this is tricky.

As to the technical cause, always interested to explore this of course. I can add it to the list of other very positive improvements with debatable mechanisms of action I've experienced, repeatably, in recent months.. incl signal grounding and vibration damping, both of which I tried on the Gryphon and Douk with satisfying effect.

I'm off on leave for a few weeks so won't be in a position to experiment much in person but will continue to follow this and other threads with interest.

Cheers
Jake

 

[dpeterfy32]

hello all. I would like to ask how much the sound is delayed when using the gryphon as a bluetooth receiver? noticeable? while watching a youtube video or movie?