Buffalo II, Musiland 03US Cap Mods Results
The normal operation of the Buffalo DAC is to have the DPLL set at “BEST” and with this setting, there are no unlocks whether you are using I2S input or SPDIF input. Even during warm up, with the DPLL set at “BEST”, there are no unlocks.
The DPLL has several settings, and in theory, when set at “LOWEST”, the built-in ASRC rejects the most jitter. It has been widely reported that when the “LOWEST” setting is used with I2S input, the DAC will experience unlocks: a lot during warm-up and then occasionally afterwards. Unlocks will also happen with higher sample rate material if using the lower settings of the DPLL.
The mods I’ve been experimenting with and the resulting measurements have been aimed at understanding and correcting what causes these unlocks. Essentially, the DPLL in the DAC cannot lock into the incoming signal because of jitter. In this post, I report my latest results.
I have added input and output capacitors to the Musiland 03US [link] local regulators and also added output capacitors to the Bufflao II DAC local regulators [link]. These were fairly easy and straightforward mods.
As I said in the previous post, my experience (and my ears) tells me that the current state of the art has long passed my level of audible discernment. All of these devices sound really great with or without the modifications. So I have to use my “poor man’s jitter measurement tool” to see the effects of the mod.
DPLL set at lowest, 44.1KHz material.
This is a remarkable result: never in the past was I able to get unlock-free playing until after the 50 minute mark. In this test, note that there are no more unlocks after the 30 minute mark. Granted this is just a single test,
We are back to having unlocks until the 50 minute mark. So the first test was probably an exception. In fact, the behavior is mirrors the behavior of the average of 5 tests performed before the cap mod. Nevertheless, I was never before able to get zero unlocks after 30 minutes.
Now the full 24 hour test (click for larger image):
Zero unlocks at last!
Wow! This result is even more remarkable: FIRST TIME I GET ZERO UNLOCKS IN A 24 HOUR PERIOD. And this when during the afternoon and night hours I was using the computer that was playing music (browsing the web, writing this blog entry, etc) and the family was at home turning lights on/off, using appliances, computers, TV, etc.
Compare with my last long term test where I thought I had reached the best possible results. Those result were quite good but not unlock-free [link],
I think the biggest effect of the cap mods is to add further filtering of the power supply and thus reducing the effect of external disturbances that can cause unlocks after the initial warmup.
I modded the Musiland further by adding a ceramic capacitor to the LC filter of the first regulator. Now both regulators have increased input capacitance and also increased output capacitance in the output LC filter. The capacitance on both LC filters have been increased from 10 uF to 30 uF.
I ran the long term test again:
Again, solid performance after the warm-up. Note the lack of unlocks after the 30 minute mark.
Decided to re-install the capacitors in a different configuration and use larger values. Also while trying to tidy up the ceramic capacitor, the “end cap” detached from the capacitor body. The end cap of these ceramic capacitors seems pretty fragile. I decided to use a horizontal SMT electrolytic capacitor without the plastic encasing while I purchase new capacitors. Ceramic caps have lower ESR.
Here are a 4-hour test starting at around 6:30 PM which seems the electrically noisiest time at home
Again, this is what we would expect, clean (no unlocks) after warm up. I think the Musiland device can provide solid (unlock-free) performance when matched with the Buffalo DAC
Was able to salvage more Kemet Tantalum capacitors. These are 22 uF, 25v and they are not cheap, at about $1 each [Digikey]. According to spec, the ESR is not too bad, 0.8 ohm. Probably not as important here in the LC filter since it is not used as a bypass cap. Besides, it is already an improvement to the existing tantalum capacitor.
Here is the latest iteration of the mod:
After this last mod, I did another long term test for 20.5 hours. This time I got two unlocks (in the same 10 minute interval). I guess there is still room for improvements. For now, I think this is as good as it can get. I think further improvements will come from additional power and EMI isolation.
In summary, the cap mods were easy, economical and resulted in achieving zero unlocks after the warm-up period. It was very satisfying to see this particular mod cap (no pun intended) the long quest for zero-unlocks while having the DPLL set at “LOWEST”.
This is how it was before all the mods. The graph shows the number of unlocks per 10 minute interval just like the current charts, so about 5 hours worth of measurements.
These were the things I tried. I also implemented all of them except where noted:
Mods that definitely worked
- Cap mods on the local supplies for the Buffalo DAC and Musiland interface
- Shielding of the Sabre 32 DAC chip
- Shielding of the DAC board
Mods with potential value (there seem to be some improvements)
- IL715 Isolator (not used because in theory it adds jitter to the signals)
- Mechanical damping of the clocks
- Mechanical isolation of the case, etc
- Ferrite on the USB cable
Mods that very likely did not improve things
- Really short I2S wires (like cutting to 3″ from 12″ -after you cut the wires, no use lengthening them again :-))
- Ferrite bead on the I2S bit clock wire (not used, seemed to be worse)
- Night time use vs day time use (well, not a mod, but testing noise in the power grid)
In my setup I already had a power filter/conditioner, so it was already used for all the tests.