making USB DAC -- VICS USB Audio Kit(PCM2704) --


Japanese version here.


I wanted something acoustic after making power amp, speaker and modifying soundcard.

I decided to make USB DAC(Digital Analog Converter).

I found a USB audio kit by VICS, so I tried to build it.

VICS USB audio kit (PCM2704) (written in Japanese)

PCM2704 datasheet (PDF)

The characteristics of this kit were following.

1: the IC of USB DAC is PCM2704 by TI/burr-brown
2: work in USB1.1
3: sampling rates are 32, 44.1, 48kHz
4: 16bit digital sigma DAC
5: THD+N = 0.006%(RL>10k ohm)
6: SNR = 98dB
7: Dynamic Range = 98dB
8: adopting Windows standard driver (USB Audio Class Device)
9: the size of board = 40 * 70mm

I wanted to make USB DAC better than modified SoundBlaster Digital Music LX.


I was relieved that PCM2704 was already on board because its legs were so small.
It would be very difficult for me to solder them.

The parts of the kit were mostly mediocre.
I would want to exchange them for audio-grade parts after completion, so I set high-class parts in advance.
The change are following.
code value detail use changed into value detail maker price note
R1,R13,R16 4.7kohm meetallic film resistor 1% 1/4W R1: adjasting the voltage for LED
R13,R16: the return resistor of OpAmp
---- 4.7kohm metallic film platy resistor 1% 1/2W NIKKO 42yen *1 R1 was removed
R2,R3 1.5kohm metallic film resistor 1% 1/4W R2: adjasting 3pin(DT) for LV TTL-level
R3: adjasting 9pin(D+) for TTL-level
---- 1.5kohm metallic film platy resistor 1% 1/2W NIKKO 42yen *2 _
R4,R5,R6,R7,R8,R11,R17 22ohm metallic film resistor 1% 1/4W R4,R5: damping resistor
R6 and C9 consists of LPF for OpAmp's V++
R7,R8 and C7,C8 may consist of LPF
R11,R17: damping resistor
---- 22ohm metallic film platy resistor 1% 1/2W NIKKO 42yen *5 R6,R11,R17 were removed
(But R11 and R17 should be set in order to remove reflection of high frequency signals)
R9 1Mohm metallic film resistor 1% 1/4W generating clock ---- _ _ _ _ removed
R10,R18 20kohm metallic film resistor 1% 1/4W with C14,C15 consisting of HPF for output ---- _ _ _ _ removed
R12,R15 10kohm metallic film resistor 1% 1/4W determining the gain of OpAmp ---- 10kohm metallic film platy resistor 1% 1/2W NIKKO 42yen *2 _
R14,R19 470kohm metallic film resistor 1% 1/4W determining input impedance of OpAmp ---- 470kohm metallic film platy resistor 1% 1/2W NIKKO 42yen *2 @
Y1 12MHz crystal unit supplying PCM2704 with 12MHz clock ---- _ _ _ _ changed into crystal oscillator
LED1 red LED power indicator ---- _ _ _ _ removed
U2 _ OpAmp OPA2340PA determining the impedance of output ---- _ _ _ _ changed into NJM2114D
C1,C10 22pF laminated ceramic capacitor 50V soaking up noises from crystal unit ---- _ _ _ _ removed
C7,C8 0.022uF laminated ceramic capacitor 50V R7,R8 and C7,C8 may consist of LPFH ---- 0.022uF metallic polyester film capacitor MKT1826 100V ERO 136yen *2 _
C3,C5 0.1uF laminated ceramic capacitor 50V soaking up high frequency noises from 10pin(Vbus) ---- _ _ _ _ I used them.
C2,C4,C9,C11,C13 1uF laminated ceramic capacitor 50V soaking up high frequency noises from IC ---- _ _ _ _ I used them.
C12 47uF ordinary capacitor decoupling capacitor which supplies Vccp/2 for single voltage supply ---- 100uF OS-CON SA 20V SANYO 231yen _
C14,C15 2.2uF ordinary capacitor coupling capacitor ---- _ _ _ _ removed
L1 100uH coil consists of LPF for USB bus power ---- _ _ _ _ removed
CN1 _ USB connecter _ ---- _ _ _ _ I used it.

other parts
case UC9-5-12DD TAKACHI 1610yen
RCA terminal gold plated RCA jack (black and red) 399yen
AC adapter 9V 2.5A 24W switching AC adapter 600yen
DC plug MJ-10 30yen
bracket LED CTL701R SANSEI 168yen
switch DQ-11 3pin 105yen
resistor oxidated metallic film resistor 62ohm (adjasting voltage for LED) 21yen

listmarkimproving clock

The origial crystal unit of VICS kit was written "12AKSS5DT".
I didn't get any information for this because no site was hit with the keywords "12AKSS5DT".
I guessed it would be ordinary crystal unit, so I decided to exchange more high-class one.

Clock is the heart for audio IC.
Sampling rates and other frequencies are generated by this clock.
If we set very accurate crystal unit, we will get clear outline sound.

I searched 12MHz one.
Then I decided a crystal oscillator which was looked nice.

TCXO is guranteed for temperature change.
We can see the words "KSS", so it will be made by KYOCERA.
But I couldn't find any datasheet of this, so I didn't know about this.
I believed that it would be high-grade crystal oscillator.

This picture is back side of it.

I determined its circuit as following.

1pin : NC
7pin : GND
8pin : output clock
14pin : 5V supply

It is recommended to insert 22pF capacitor between 14pin and 7pin as closer as possible.
This is for removing high frequency noises that oscillator will emit.
Laminated ceramic capacitor will be good for this role because of its high quality of high frequency.

330 ohm resistor will work to remove reflection of high frequency waves that will be caused by the gap of impedance between OSC and signal line.
It is often called as "damping resistor".

I also inserted 220uF OS-CON to 5V supply in order to remove ripple.
This would stabilize the 5V power supply.

crystal oscillator 12MHz TCXO KYOCERA 525yen
330ohm resistor metallic film resistor 1/2W 1% NIKKO 42yen
22pF capacitor laminated ceramic capacitor accompaniment to kit _
220uF capacitor OS-CON SA 10V SANYO 231yen

I cut universal board into properly size and I fastened it to print board of kit with legs of resistor and lines of GND.

listmarkdriving PCM2704 self-power mode

We can find in the datasheet that self-power mode is much better than USB bus power.
Self-power mode is lower noise-level and better dynamic range.

But there are two problems in practice of self-power mode.

1: we have to solder PCM2704's tiny legs
2: it is necessary to modify print board(cutting pattern)

I guessed that it would be ultimately difficult to solder PCM2704.
In addition to this, If we cut pattern, we would be unable to restore it.

But I tried it.

I flattened a lead line which was the leftower of resistor's and capacitor's legs using file.

Then I soldered this into 4pin(PSEL) and 21pin(HOST).

I soldered carefully not to unite to next leg.

Now I cut pattern of the land of C2 and 4pin(PSEL).
I used cutter knife for this.
It is recommended to check isolation by tester.
If isolation is inadequate, it will be troublesome condition.

Then I supplied 3.3V to PCM2704.

destinations of 3.3V
connection pin the place of pin
7pin(VDD) the leg of C2(the square land)
13pin(VCCL) the leg of C4(the circle land)
16pin(VCCR) the leg of C13(the circle land)
20pin(VCCP) the leg of C11(the square land)

In addition to this, it is necessary following things.
connectig 4pin(PSEL) to GND
(connecting 21pin(HOST) to USB 5V and inserting 1Mohm resistor between USB 5V and GND)

The circuit for self-power mode.

Now I wired actually.

Now it would be the certain improvement for sound quality.

listmarkdriving OpAmp both power supply

OPA2340 which is pertained to VICS kit is a rare OpAmp.
It works single power supply.
datasheet of OPA2340

For ordinary OpAmp, it is necessary to prepare both plus minus power supply or set bias voltage of V+/2.
OpAmp is important part which is correspondents with vacuum tube and determines sound quality.
So OpAmp enables us to change sound taste.

I liked powerful deep sound, so I chose NJM2114D.
I tried to drive it both power supply.

I prepared a 8pin-socket in order to replace OpAmp easily.

I connected 4pin and minus power supply, 8pin and plus power supply.
I bended socket's 4pin and 8pin, and I soldered them to cables which were linked to power supply.

In the upper picture, the black cable is the one which connects 4pin and -5V.

If we fix the gain for high value, it will be necessary to remove DC by high pass filter before OpAmp.
That is because the output voltage of PCM2704 is biased from +0.3 to +3V and if we set high gain, the output will be clipped.

The circuit of these is following.(the single channel)

This is non-reversing amplifier.

The gain is A = 1 + R2/R1.
R3 determines input impedance, 470kohm resistor will suit for this.
C1 and R' consists of HPF(High Pass Filter).
the cut-off frequency is 1/(2*PI*C1R')[Hz] .

It won't necessarily need output coupling capacitor if input voltage is correctly 0V.
This is the advantage of both power supplying.

But PCM2704 needs coupling capacitor because its output is biased.

listmarkgenerating minus power (charge pump)

How to make minus voltage is a disstress.
I guessed following three methods.

1: preparing two independent power supplies
2: using DC-DC converter
3: adopting charge pump method

OpAmp doesn't need large current. So I chose charge pump method.
It would be easy to make and cost low.
It is the way charging a capacitor and then switching it to the output for generating negative voltage.

There are ICs which work this role.
7660 models - including boost mode (ICL7660/7662, JRC7660 etc)
1044 models - non-boost mode (LTC1044/1044A/1144, MAX1044 etc)
These are famous.

Boost mode enables to heighten the switching frequency over 20kHz that is beyond the bandwidth to which man can listen.
It is recommended to choose boost mode for audio device.

Because I used 9V power supply, I searched charge pump IC which could endure over 10V.
I found LTC1144 which had boost mode and could endure 15V.

We can use external clock for LTC1144, so I bought a multi-frequency crystal oscillator, EX03, and I tried it. But it didn't work correctly. I guessed it wouldn't work with over 100kHz, so I used LTC1144's internal clock.

This is the circuit for charge pump. (boost mode = on)
The capacitor for charging should be over 10uF and low ESR.

It is also recommended to insert capacitors which are good at high frequency and about 0.1uF between 3pin and 8pin as close as possible.
It will remove high frequency noises emitted from IC.

If we use this output, we will be able to get negative voltage for OpAmp.

charge pump IC LTC1144CN8 15V ---- plus minus 15V 300yen
100uF capacitor SANYO OS-CON SA 20V 231yen *2

(There are EX03 and its output resistor in this picture, but they are not connected into LTC1144 because they have not worked.)

listmarkgeneratin 3.3V, 5V, -5V (switching regulator)

Three kinds of power supply were necessary for me.
1: 3.3V which was for self-power mode
2: +5V which was for crystal oscillator and OpAmp
3: -5V which was for OpAmp

Low noise power supply was recommended.

Switching regulator enables to generate clean power which is low ripple.

I prepared 3.3V, 5V and -5V regulators.



3.3V switching regulator
TA48033 84yen

I sticked it into heat sink using double-faced tage.
5V switching regulator
7805 63yen

The voltage of package is 0V.
-5V switching regulator
7905 63yen

Be careful that 7905 is different from 7805 in layout of GND and IN legs.
The voltage of package is input voltage. So mold packaged regulators will be useful because they are isolated.

The circuit for 78xx is following.

I chose following parts for C1 and C2, but they would be enough about 100uF.
C1: 3300uF (MUSE FW 25V)
C2: 6800uF (ELNA RJJ 10V) for 3.3V
: 3300uF (MUSE FW 25V) for 5V

The circuit for 79xx is following.

Be careful that 7905 is different from 7805 in layout of GND and IN legs.

C1: 100uF (OS-CON)
C2: 3300uF (MUSE FW 25V)

Heat sink will be unnecessary for regulators because PCM2704 and other IC don't need large current.
But it is desirable to place heat sink thinking of summer.
TO220 will do this role.

listmarkprocessing case

I used the same case as the one of TA2020-20 D-class AMP.

I drilled using hand-spinning drill, Chassis reamer and metallic file.

Little by little I was progressing with processing aluminium.


It looks similar to TA2020-20 digital amp does because they adopts the same cases.
I like this case.

Wiring is very simple.

But inside is messy.
Smoothing capacitors and heat sinks takes so much space.
If we adopt USB bus power, it will become much smaller.


PCM2704 works by Windows standard driver, USB Audio Class Device.

I switched on.
LED shone and it sounded.
I was touched.

I compared with Modified SoundBlaster Digital Music LX.
It costed about $100, this was about twice the cost of SoundBlaster, and took about twice the volume of SoundBlaster.

I would be unable to be satisfied if it couldn't sound twice SoundBlaster did.

treble It is comfortable. It isn't harsh sound. The sounds such as when glass breaks are real as if they sticked my ears.
midium SoundBlaster is a bit cloud, but it is clear. I feel it has peeled SoundBlaster's film.
bass I feel it has a large capacity for bass. It sounds softly and powerfully.
generally It has larger capacities than SoundBlaster. It is confortable.

I'm very glad that it sounds much better than SoundBlaster.
Making audio device is very pleasant.

listmarkadding bass boost circuit

I had adjasted tone using software equalizer, but it had been heavy load to CPU.
So I tried to make a equalizing circuit.

I liked deep bass sound, so I decided to make a bass boost circuit.
Its circuit is following, it will work as LPF(Low Pass Filter) using OpAmp.

Figure: bass boost circuit using OpAmp

This is just a non-reverse amplifier added C1.
If there isn't C1, the gain will be A = 1 + R2/R1.
But it becomes a bit complex when C1 is added.
The gain of this is determined by following equation.

A' is dependent on frequency.
In other words, we can change A' by frequency.
Adjasting the values of R1, R2 and C1, we can get original bass boost circuit.

I noted my favorite equalizer setting using software equalizer in advance.
Then I determined them which would work like it.
I used Excel for calculating the characteristic in frequency.
Bass.xls (Excel file)

"Bass.xls" calculates the behavior of bass boost circuit just changing the values of R1, R2 and C1.

I set the values as following.
R1 = 12k ohm
R2 = 15k ohm
C1 = 0.022uF

I bought these parts.

OpAmp NJM2114D JRC 168yen
12k ohm Riken MA 1/2W 5% 78yen *2
15k ohm Riken RMG golden legs 1/2W 2% 160yen *2
470k ohm NIKKO 1/2W 1% 42yen *2
0.022uF MKT1826 100V metallic polyester filmy capacitor (ERO) 136yen *2

Riken resistor is over production, so I chose it.
It was a bit expensive, but it might have a premium later.
(I could enjoy thinking like that, so It would be worth the price)

RMG looks luxury because of its golden legs.
It is carbonic resistor for audio.
I heard it sounds vivid and nice.

RA is poor at its accuracy, so I didn't want to use as a resistor determining the gain because the ones of Left and Right channels would be different values.
But I bought some RAs and chose two of them which had certain accuracy using tester.
So they would be all right.

I made on a small piece of universal board actually.

The electrolytic capacitors are BlackGate NX 22uF 168yen *2 for coupling.
I inserted coupling capacitor because PCM2704's output had DC voltage as large as 2V.
(After taking this picture, I noticed that I didn't insert R3 between BlackGate NX and OpAmp.)

I connected this bass boost circuit to VICS kit.
As a result, I used two NJM2114D, in output buffer and this circuit.


It sounded like software equalizer did.
The design seemed to be correct.
In addition to this, it sounded more deep than before.
This was like pennies from heaven.
I don't have enough idea about this. Is this a benefit of Riken, BlackGate NX or two NJM2114D? Is analog circuit better than software bass boost?
Anyway I'm very glad.

After this, I modified in order to control bass volume.

It looks like TA2020 Amp does.
I set these join each other.

The back shape.
I regret that I've allocated terminals thoughtlessly.

listmark If you have a question or comment, please use BBS.
If BBS doen't satisfy your needs, give me a mail. (replace # by @)

Notice!! I don't take any responsibility for the trouble that can occur for this homepage. Please do at your own risk.