Are we in agreement, that DACs are sounding better when a good quality low phase noise clock is being used? Yes? keep reading… There is a lot of information and good articles on this topic, so I will keep it brief and not go in all possible backgrounds of phase noise, clocks and their power supplies. Lately (the past year or so) Andrea Mori started a thread on the DIY Forum with also a group buy for a really top of the line development of a clock to be used in audio equipment. For example his future FIFO and DAC or like my DDDAC1794and Ian Canada FiFoPi. I will not go into detail what all this is. Check the threads in the DIY Forum:
Link to Andrea Mori clock thread Please check post #1 for further links to group buy etc
Link to Ian Canada FiFoPi thread This is a VERY long thread, so my be start a check at #1
Clocks and their frequencies
Just to line it up – what frequency do you need? The choice seems wide. The principle is that the clock frequency is the MASTER clock or MCK. From this the Bit clock (BCK or SCK) for the I2S signal (being used in many and my DDDAC for example) is being extracted by dividing it by 4 (256FS) or by 2 (128FS) – I know this already sounds complex. You always see these power of two relationships. Some DAC chips use 256FS (divide by 4) and for example the Ian Canada FiFoPi uses 128FS (divides by 2) – Finally this signal is being divided by 64 to arrive at the Word or Left/Right clock (WCK or LRCK). The last signal contains basically the left and right channel sound samples in 32 bit format per channel (this is most used – there are others, but not now….)
Why is this and what is the relation to my sound track’s FS (and… what choice to make?)
The sample frequency FS of a sound tracks being reordered or rendered is given in kHz. Like the good old Redbook CD = 44,1 kHz. Ripped SACD tracks you find for example as 88,2 or 176,4 kHz. High end studio tracks often as 96 or 192 kHz… In the table at the right you will find the relationships for the FiFoPi. You will notice they can be summarized into two families where easy doubling can occur: 44.1 (CD and SACD) and 48 (typical high res stuff and DVD ) families.
In the MCK column you will recognize the typical values of XO clocks being offered by suppliers and vendors. The FiFoPi for example is automatically detecting the clock frequency and selects the right one for a family of 44.1 or 48 tracks. So you could reason just take the 45 and 49 MHz clock and you covered for most available tracks? YES correct. As I did indeed for my Accusilicon XO clocks in my DDDAC1794, but….. according to Andrea’s measurements, the phase noise is lower at lower speed clocks, so you take what you need as maximum for your tracks available or what you want to play. To make things even more complex: Andrea claims that you can preserve quality by not taking a higher frequency Crystal, but doubling the clock frequency with a special circuit……
Lots of PCBs
In that case you can get a pretty full set up. The image at the right shows a set up where you could go up to 188,4 and 192 kHz tracks, maintaining optimum phase noise quality. Well, that is a big jump from just two DIL14 XO metal cans like the Accusilicon Clocks I use too in my DDDAC1794 setup (Link to Audiophonics website selling these clocks)
This will get you 8 boxes with PCBs for two clock signals. Up to the DIY’r of course – it is hobby after all right? Of course I was intrigued and wanted to try Andrea’s clock. But I was not ready and willing to go this long “many box” way and decide to just test the base frequency of 5,6448 MHz to test CD redbook tracks (and convince myself if it is would be worthwhile)
My plan for testing Andrea’s clock
OK, so I bought the one base clock 5,6448 MHz. This would give me the best possible phase noise and I wanted to know what THAT would do with sound quality and also decided to NOT go immediately “over the top” in the test setup. The reasoning behind was that even with relatively simple gear, the low phase noise clock should make a difference compared to other clocks. You can always take it from there. Experience shows, that with better power supplies for example things always improve somehow similarly.
Below is the “reduced” setup I am using for the test. In regard to the power supply, I also decide to NOT use extreme low noise regulators or complex battery supplies. Just a simple LM317 supply. I hear you think, “you will never be able to judge the quality of the clock if not using maximum high end PSU” Probably true, but I wanted to start simple. Tweaking is always possible and there is always room for improvement afterwards. (may be later part two of this post 😉 )
Enough contemplations, back to my nice new work shop and get things ready. For testing purposes I made a DDDAC1794 setup with a one deck and two OTPs (the old Audio Creative ones – which works with one deck) and the standard DDDAC Power Supply without big transformers or CLC filtering. As I2S Source a Raspberry PI3 and the Ian Canada FiFoPi Q2. This is already a major leap DAC versus many standard stuff DACs by the way… But anyway, see some details below. The clock measures a bit below with 5,6446 MHz in stead of 5,6448. For Audio this absolute value is totally negligible. The power consumptions is steady with ~ 70mA as you can see. The clock worked best between 16 and 18 Volt by the way (Sine wave signal form)
Listening with Music
My rack is pretty full, so with tests I have to put in in front, LOL. Here you also see the small PCB with the 16,5 Volt LM317 power supply. Coming back to the power supply. The Sine-to-Square converter (the small sine to square stick up PCB at the DIL14 position on the FiFoPi) is using the ultra low noise regulator already on the FiFoPi. So it is probably not that bad after all when the square wave clock signal is being applied to the FiFoPi….
I arbitrarily picked three clocks for the listening test:
- The stand cheap 49Mhz XO clock coming with the FiFoPi ( a kind of “two dollar thingy” I guess)
- A 11.286 MHz Tentlabs Clock. Many years this was the de facto standard to use as clock upgrade in CD players and DACs
- The 5,6 MHz DRIXO Andrea clock
I did not wanted to unscrew my DAC for the Accusilicon clock. I will do this in part two update I guess…. and I was hoping and expecting a nice gliding scale from 1-3 of course. Also there are many users out their with TENT XO clocks 11,286 MHz. so they might be able to relate?
As you also see below, no super power cables, no solid core silver connections. Also pretty standard stuff – so ready to to start listening
Changing the Clocks
I needed a procedure to make the comparison meaningful; so this was my plan of action:
- Pick a few 44.1 Redbook tracks I know well and where there is a lot “to hear” (instruments, voices, spatial stage etc.)
- Play these tracks with Roon (it has up sampling in the DSP section which works quit OK)
- Prepare for up sampling from 44,1 to 192 kHz – so FiFoPi will pick the cheap 49,15MHz XO clock
- with Roon enable / disable the DSP button with one click, Roon now plays the 44,1 of course and the FiFoPi now selects directly the Andrea clock. So there is straight A-B comparison possible
- After listening in and getting the differences, I just manually unplug and re-plug the Tentlabs clock and Andrea clock. This can be done “life” no problem, so it goes pretty quickly without any need to power down and up again
Conclusion and Summary
These are my impressions:
- First I just listened to the 44.1 tracks with the cheap 44.1 family clock. Up sampling from 44.1 to 192 kHz and listening to the 48 family clock makes things a little bit more fussy and flat, but it is small and the character of the cheap clock is now a known in my head for comparison with the Tentlabs next (Just click DSP on/off and enjoy direct A-B)
- moving to the Tentlabs clock I hear what I already knew from moving to the Accusilicon clocks: Space is getting more precise, voices are better defined an texture increases, transients are more isolated and helps getting a good spatial sound stage. If this would be all you ever could lay hands on or being available on the market. I would be happy with this and consider this a good upgrade to the FiFoPi and my sound set up
- Doing the same comparison between 48 family cheap clock and Andrea. OK, this is a significant step now. Click the DSP off and on for A-B listening and you immediately hear things opening up and does all the Tent does but improved more
- Finally comparing red book 44.1 tracks with unplugging and re-plugging the Tent vs. Andrea clocks. OK, this is not a shock anymore and I needed to go back and forward a few times to get the picture. But the picture IS very clear. The sound stage is better defined, voices are having a more detailed texture and instruments have this tad of sounding more natural. Not “HUGE”, “Playing against the wall”, NOT “the other was COLD and uninspiring”…. I read this too much, it is not like that at all. BUT, you can hear it and it is very nice – So you want to keep this. Any YES, I will need to find away later on to integrate this in my system and does more tweaks as there is surely room for improvement.
It is soooooo hard to express these things and get them in relation. I typically are a bit holding back where others are euphonious on every little step. This makes you wonder if they were listening 3 steps back to a shower radio or something? Well, everyone should do their thing, I do mine.
On a scale from 1-10 for this particular test with three clocks I would say 1-7-10, for what it is worth…
To give you some other idea, I thought the step from from TENT to Andrea to be kind of similar to comparing my LINN Krystal cartridge versus my LINN Kandid… Both great but clearly the Kandid is the more refined one. AND……. you pay an extra 1500 Euros for this. That will put it both ways into perspective I guess…. Make up your own mind on what I mean with this….. Oh, the cheap clock would compare with a standard Ortofon 200-400 Euro moving coil cartridge or similar I would think.
- Do the same comparison against the Accusilicon clocks
- Apply ultra low noise power to the Andrea Clock and compare
- Apply Andrea clock directly in my own DDDAC1794 set up (with CLC supply, ultra low noise power supplies, Sowter TVC, silver cables etc etc) and compare against Accusilicon (which ar currently in the my DDDAC) and compare again
- If I am so enthusiastic I want it all to change, I need to contact Andrea and see if he still sells the clock-doublers and other clock frequencies to try these (for Step THREE 😉 )
… Good I semi-retired and have time for this, LOL…