Buffalo II DAC PS Output Cap Mod
The Buffalo II DAC still is one of the best engineered and the best “bang-for-buck” ESS DAC out there. The power section was designed with solid engineering practice (4-layer PCB, close-to-the-chip bypass capacitors, etc), using low noise series LDOs in the digital section and shunt regulation in the analog section. I think the robustness and reliability of the IC regulators in the digital section is actually an advantage over using shunt type regulators such as the Trident regulators. The Digital section of the DAC does not require quick response and close tracking to changing current demand as in the analog section, but rather just provide a solid, steady and noise-free power delivery. This is perfectly accomplished by series LDO regulators.
One possible place to mod is the output capacitor for the on-board power supplies for digital section of the chip. The regulators are based on the LT1763 low noise LDOs with a specified noise figure of 20 uV RMS. According to some reported measurements the noise figure can be further reduced by increasing the output capacitor. The results have been reported in diyaudio:
[link] Connecting a TL1763 LDO, however, I managed to get noise down to 8-10 uV, with reduced HF components… Throw away your 7805 and LT1086s, and try something better. The LT1763 has 20 uV specified noise, but a 150 uf OSCON at the output did the business.
[link] I have now finally had the patience to rig up the 100 mA LT1761-5. Powering a 45.1584MHz XO, I am in the 10nV/sqrtHz noise again, with a 150 uF OSCon at the output . Like the LT1763, the 1761 needs a 3×2 cm ground plane connected to -ve but under the module to reject a slight hum pickup.
The current configuration of the on-board LT1763 consist of an input capacitor of 100 uF (an OSCON SVP type), a bypass capacitor of 10 nF and a ceramic output capacitor of 10 uF. If you look at the photo above for the bottom supply, C18 is the input capacitor, c21 is the bypass capacitor and c22 is the output capacitor [link]. It is configured as per spec except for the input capacitor which is 100 uF. In addition, there is further filtering provided by a series ferrite bead which, together with the bypass capacitor near the ESS chip, forms an LC/RC filter. This is not just a solid implementation of the LT1763 regulator, but goes further in the noise department. Adding the output capacitors will further improve this already excellent implementation.
The bypass capacitor is already at the maximum required (10 nF or 10000 pF) to reduce the output noise to its minimum specified. Thus no need to mod this part.
We have to make sure that the added capacitor, especially because it is a low ESR type will not cause stability issues in the regulator. The stability graph tells us we are safe even with ESR=0.
INSTALLING THE CAPACITORS
The specification does not measure noise with varying output capacitor size, so we are basing this mod on experiments by other audiophiles as reported above. This mod is very easy. Just piggy back the capacitors on the existing ceramic output caps as shown in the following photos:
Installed the capacitors horizontally for two reasons:
- Keep the same size profile as the original board
- It was very hard to get the solder tip to the soldering position with the capacitors in upright position, so there was no easy way to install them upright
I decided to use 100 uF for the digital supplies and a larger 820 uF 4V capacitor for the clock power supply. The clock supply is more critical and needs to be more stable in order to keep the clock stable.
The capacitors already had the leads cut for board stuffing, making them very short. I had to extend the leads with cat5 wire.
Soldering onto the ceramic output capacitors was very easy. These ceramic capacitors are fairly big in comparison to what is used in the Musiland boards.
Even reworking was easy. Here I first installed a 100 uF capacitor, but later decided to install a 820 uF capacitor. Removing the first cap and installing the new one was a breeze. Use leaded solder and high flux content…
A NOTE ON CAPACITOR RATED VOLTAGE
The capacitor I used for the clock supply is rated at 4V (5.2V surge). Is this safe enough? The output voltage of the regulator is 3.3v, thus the margin is about 20%. I think this is safe enough.
In case of failure, we can expect the following:
According to the OSCON-GENERAL CATALOG the failure mode of an OSCON capacitor is a short circuit:
The main causes of failure are thermal stresses cause by the soldering or thermal use environment, along with heat stresses, electrical stresses or mechanical stresses. The most common failure mode is a short circuit
Fortunately, the LT1763 has current limit protection. In the event of a short circuit in the capacitor, the LT1763 will protect itself by limiting the current. The output voltage will be zero and the clock will just stop operating. The Placid supply which is feeding the BII DAC is also limited to about 300-400 mA and when it cannot supply the current shunt by the short circuit, the entire DAC will just gracefully shutdown. I think the only damage will be the capacitor itself.
Thus there are already several safety measures built-in in the event of the OSCON capacitor failing in a short.
At this level of noise and resolution, I can’t tell by listening whether there has been a positive change or not. 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 great with or without the modifications. And in many cases, the gratification comes in being able to add the mods and tweaks, sort of like saying “the journey is the destination” :-).
I am also doing some unlock measurements, will report later…
A NOTE ON OSCON CAPACITORS
The capacitors used in the mod are the older series, now discontinued, which uses organic semiconductor dielectric. The new series (lke the OSCON-SVP types used in the BII DAC), uses conductive polymer as dielectric. They are identical in all respects except the new series exhibits very minor variations in temperature characteristics vs minor variation in temperature characteristics in the older series. You can read all about it in the OSCON-CONSTRUCTION specification.
In fact the SP series are specifically “Optimum for audio”:
Additional info here: [link]