Improving on the Musiland 03US?
After using the Musiland 03US as a USB-I2S interface for the Buffalo II DAC for some time now and lately comparing it with the Amanero board with respect to my “poor man’s jitter test”, I am coming to the conclusion that this device is comparable in performance and low jitter with the best USB-I2S currently available in the market and it is a good solution for the long haul. It is to Musiland’s credit to produce a good product in “spite” of using such “non-audiophile” approaches such as DC-DC switching power regulators and derived clock frequencies. In addition, the company continues to update the software at a regular pace and the device driver implements a unique approach to transfer data, namely the use of Bulk transfer mode.
Having said that, I think it is time to investigate possible mods (if for no other reason than to satisfy the diyaudo modder inside me :-)).
The clocks in the Musiland device are generated by the 4 Digital Clock Managers in the FPGA. The incoming master frequency to the FPGA is 48 MHz and is provided by the Cypress USB receiver. This 48 MHz frequency is referenced to the on-board 24 MHz oscillator (more info on how the clocks are derived here:[link]).
Because the final I2S frequencies are generated by the reference clock passing through complex circuitry in two chips, I doubt changing the 24 MHz clock would result in any improvement (e.g. with lower jitter clocks). The inherent added jitter as the clock signal passes through the USB receiver and the FPGA clock derivation circuitry will predominate over any lowering of jitter by replacing the existing oscillator.
The following photo shows the digital power section for the Musiland 03US. These regulators provide the 3.3v and the 1.8v for the USB receiver chip and the clocking circuitry implemented in the FPGA.
The “C5BC” regulator seems a DC-DC switching regulator because of the presence of a large LC network filter at the output. There is a 4.7 µH output inductor and a 10 µF tantalum capacitor for each of the two regulators. Though I wasn’t able to identify the “C5BC” regulator, the topology and components are similar to the circuit for the newer Musiland devices shown below.
The following photo shows the digital power section for the Musiland 03USD and Musiland 03 US Dragon.
The PHKI is an adjustable switching regulator from TI: The TSP62200 (or one of its variants. -look at page 2 of the datasheet under “Symbol”). As per datasheet, there is a 10 µH output inductor and ceramic output capacitors (likely 10 µF) on each of the two regulators. Since the choice of component values seems so close to the components shown in the previous circuit, I think we can use the recommendations from the datasheet of this regulator and apply them to the “C5BC” regulator in the Musland 03US Regular Edition.
Shown below is the recommended configuration for TPS6220 in the Musiland 03 US Dragon Edition.
According to the specification (p 13), the choice of input capacitance is as follows:
Because the buck converter has a pulsating input current, a low ESR input capacitor is required. This results in the best input voltage filtering and minimizing the interference with other circuits caused by high input voltage spikes. Also the input capacitor must be sufficiently large to stabilize the input voltage during heavy load transients. For good input voltage filtering, usually a 4.7 µF input capacitor is sufficient. It can be increased without any limit for better input-voltage filtering. If ceramic output capacitors are used, the capacitor RMS ripple current rating always meets the application requirements.
The datasheet also specifies the choice for output capacitor (p 13):
The advanced fast response voltage mode control scheme of the TPS6220x allows the use of tiny ceramic capacitors with a value of 10 µF without having large output voltage under and overshoots during heavy load transients. Ceramic capacitors with low ESR values have the lowest output voltage ripple and are therefore recommended. If required, tantalum capacitors may be used as well.
TWO EASY MODS
Thus an easy mod is to add input capacitance to the regulators. The output capacitor, at 10 µF seems adequate enough, but can be increased 2X without exceeding recommendations.
Here is the input capacitor mod (before and after) I used 100 uF OSCON SH capacitors:
This mod is exactly the same mod I did to the Amanero board. However, probably not as beneficial as to the Amanero board because there is already a large capacitor (560 uF) on the USB power line.
However these OSCONs are low ESR and are positioned close to the regulators, improving input filtering.
I tapped to the the power line by scraping some of the solder mask to expose the metal in order to solder the + legs of the capacitors. The OSCON caps I have are made for stuffing into boards and thus they have short leads. I had to make an extension for one of the legs. The negative legs were soldered to the nearest ground connection, which is actually the ground connection of the original input caps.
To facilitate soldering,I first put some solder on the exposed line, put some solder on the leg of the cap and then just reheat the leg of the capacitor when it is put in place until you see the solder flow. It the best way I know to minimize heating time without having to worry about cold solder joints.
Had to solder the capacitors at an angle (not as pretty :-)) because there is not much space to stick a soldering iron without potentially burning the surrounding components.
As it turned out, the capacitors had to be at an angle to fit into the case. Lucky me! 🙂
Here is the output capacitor mod.
I added 22 uF only to the second regulator first. It was hard to fit an additional output capacitor to the first regulator, so I left it as implemented.
Tantalum capacitors are easy to identify. This one was a salvaged and measured around 20 uF. I soldered first a (cat5) wire around the capacitor. After installation I cut the wire to separate the + and – terminals. Again, adding some solder to wires and later reheating the wire as you put it in place, makes soldering much easier. Having a single piece of wire to start with helps with keep the wire in place when soldering it to the existing capacitor.
During the test for the Amareno board, damping the DAC case seemed to help reduce the unlocks. I added some damping foam to the oscillator.
ALL THE MODS
I tested the number of unlocks twice and compared the results to two previous tests without the mods. This is what I found:
The start up behavior is pretty much the same before and after mods. I also did a long term test and it looks “squeaky clean” .
FURTHER MODS TO THE MODS
(You know how it is; can’t leave it alone :-))
Boosted the output capacitor of the other regulator by adding a ceramic capacitor and replaced the input capacitor with a larger value one
Changed the other input capacitor to a larger value. Also while trying to tidy up the ceramic capacitor, the “end cap” detached from the capacitor body. So reworked the cap and used a horizontal SMT electrolytic capacitor without the plastic encasing.
Replaced the mini electrolytic with a tantalum
For the “ultimate” PS mod, diy audiophiles would replace the switching regulators with linear regulators, but in this case it is not easy to do. One may cut the power to the existing regulator and add external modules providing the necessary voltages. But this is a bulky solution and it would not fit inside the case.
But there is another way:
The MUSILAND 01 us 2012 Edition
I completely missed the introduction of the 2012 edition of the Musiland 01 US. This is the “little” brother of the Musiland o2 Dragon, Musiland 03 US and o3 US Dragon, but it has all the digital goodies (including I2S output lines) of its more expensive siblings. It is also the latest and newest introduction of the entire Musiland USB lineup. In fact it inherits the linear regulators from the 02 US Dragon board and 01 USD boards.
How does this board relates to modding the Musiland 03US?
If you look at the photos show, you can see that there are NO DC-DC switching power regulators. All the regulators in this board are linear LDOs. In fact, of all the current boards with a DAC produced by Musiland,I think this is the only one that is free from having any DC-DC switching regulators on board.
This board has “implemented” the mods an audiophile would do on the Musiland 03US: replace the switching regulators with linear regulators. In addition, input and output capacitor mods can be implemented for further improvements. If anyone can figure out what “5EBI” or “5EB” is, please let me know.
The Musiland 01US 2012 Edition can be purchased for about US$80. Thus for USB-I2S duty (after tapping the lines), the Musiland 01US looks like a worthy upgrade to the Musiland 03US and for half the price!