making NOS DAC -- TDA1543 + CS8414 --


Japanese version here.

listmarkTIMEDOMAIN Theory

I got a opportunity to hear a lecture of Mr.Yosii who was progenitor of TIMEDOMAIN.
He said following things eagerly in his lecture.

"It is wrong that designing audio device to be flat in frequency domain and making it seasoned to have good sound."
"More than 70 years has passed since contemporary audio engineering was fixed, but we couldn't get any audio product which sounded source sounds correctly. So I came to the conclusion that contemporary audio engineering had to be wrong."

I thought he was talking about ridiculous things.
But when he turned "TIMEDOMAIN light" on, I got to believe in his theory at a moment.
Marvelous sound were flown out from a tiny plastic speaker.
I didn't feel about speaker position, it was as is it were really source.

In his lecture he tought following practical things.

about amplifier
It is not good to adopt negative feedback when you want true source sound.
If you use negative feedback, the induced voltage of speaker unit which willl be caused by acoustic pressure from outside will come in from line-out and amplifier will compensate including this voltage, so we will be unable to sound true sound.
It should be based not on frequency responce but on impulse responce.
Frequency response is a responce of sin wave. It cannot be a responce of "sound". If you want a true responce of sound, you should design based on impulse responce. This would be the core theory of "TIME DOMAIN".
about speaker
Bass reflex is wrong regarding reproducing the original sound.
The bass from the duct of bass reflex is deadlly delayed, so it cannot get enough punch.
It is far from the thought of reproducing the original sound in the first place.
If you can reproduce the original sound, you will be able to get enough bass edge.
You shud use speaker unit as light as possible.
Heavy unit can't immediately responce. Light unit as a feather can do.
Speaker unit should be grounded on completely unmoved point.
Enclosure will be vivrated by sound pressure of speaker unit. Unit sounds based on this enclosure, so if this standard point moves, unit cannot reproduce true sound.
We shoud prepare unmoved ground.

I made sound equipments which were against of TIMEDOMAIN theory.
I used two negative feedbacks in PCM2704 DAC and there may be negative feedback in TA2020.
Speaker is bass reflex type and its unit is 12cm.

I'm satisfied with my audio system, but I want to try TIMEDOMAIN theory.
So I decided to make DAC which would use no negative feedback.
I want sounds as true as possible.

listmarkTDA1543 and CS8414

What is oversampling?
From sampling theorem, we can completely restore original signal when we sample over 2f.
("f" is the maximum frequency included in original signal)
But we have to remove higher harmonic waves which are over fs(sampling frequency) in order to restore completely.
This removing tool is low pass filter.
This LPF bothers us. It is impossible to make ideal LPF, so we have to use dull one and high frequency component of original signal often becomes distorted by this dullness.
We usually adopt fc(cut-off frequency of LPF) as fs in order to remove aliasing noise.
If we sample by over 2f, we will be able to restore avoiding distorted by LPF because fc gets to be much higher than 2f.
In addition to this, we can suppress noise component because we can scatter it to the area which is from 0Hz to fs(over 2f).
(N is the sum of noise which is added from 0Hz to 44.1kHz. If we sample by n*2f, we will be able to suppress N into N/n.)

I remember as above.
My emplanation may be very doutful, so please refer following sites.
Oversampling in Wikipedia
An Introduction to Delta Sigma Converters

I think oversampling is very good method.
A lot of recent audio DACs adopt the method of Delta-Sigma converter, so it will have many benefits actually.
But I heard that it use negative feedback.
I have no idea that this negative feedback is also a violation of TIMEDOMAIN theory.

But TDA1541A and TDA1543 which are Non OverSampling DAC ICs made by PHILIP are famous of good DAC, so I tried to make DAC using TDA1543.


TDA1543 datasheet (pdf)
We can find TDA1543 is a old IC to see the year, 1991, of datasheet written.
"Dual 16-bit DAC (economy version) (I2S input format)" is largely written in its front cover.
"economy version"!!
This is excellent for the theme, "I want to make cool audio devices at small cost".

features of TDA1541 and TDA1543
1: resistor ladder 16bit DAC
2: current output
3: I2S format
4: include no digital filter

It is nice for new and fickle TIMEDOMAIN believer that TDA1541A and TDA1543 don't include digital filter.

They adopt current output, so we can use plurality of them in parallel.
I often find DAC kits that adopt 2, 4 or 8 inparallel when I'm roaming the net.
I want to try parallel use.

We can make NOS DAC using a DAI which supports I2S output format.

Now it may be difficult to get such a DAI because I2S may be old format.
CS8412 and CS8414 can output I2S.
PCM2706/7 which is USB audio reciever can also do, but I wanted one which support S/P DIF input because it is useful for many occasions.

I finally found CS8414 in web.
Further more, it was already converted into DIP package.
I appreciated it!


CS8414 datasheet (pdf)

This is 96kHz Digital Audio Reciever.
It outputs I2S recieving S/P DIF.

I tried to make NOS DAC using TDA1543 and CS8414.

listmarkdesigning circuit

I/V conversion was needed because TDA1543 is current output.
There are various methods of I/V conversion.
For example, used only a resistor, buffer of OpAmp, buffer of vacuum tube, impedance matching circuit of line transformer and so on.
I/V conversion holds the key of sounding.

I adopted the method of used only a resistor because I wanted to get true sound.

We usually need LPF which its cut-off frequency is sampling frequency in output of DAC.
The output feature of DAC is LPF, but I couldn't find out whether TDA1543 had LPF in its output in datasheet.
In my case, power amplifier had its own LPF.
(TA2020 is D-class amplifier and it uses PWM, so it would not good to include the noise which frequency is as high as switching one. I don't know whether TA2020 has LPF in its input before pulse width modulation, but I guess usual D-class amplifiers will have it, maybe I hope...)
So I didn't set LPF in DAC output.
If fatal problem occures, I'll set LPF.

In order to suppress jitter noise, I adopted re-clock circuit.
I tried asynchronous reclock method which re-ranged SDATA, FSYNC and SCK by higher and high-accurate clock compared with MCK(master clock of reciever).

(For larger image, please click.)

I planned to set chip capacitors which had about 0.1uF capacity between the lengths of regulator and each IC.
I also designed to set 22pF chip capacitors into crystal oscillator.

Digital inputs were 2 ways (coaxial and optical).
Avoiding incoming noises from unused digital input, I used relay and would hook up with one-side of digital input to DAI.

I used three TDA1543 (3para).

I used three regulators and they would separately supply voltage to CS8414, OSC + 74HC74 and TDA1543.

listmarkcollecting parts

_ parts quantity detailed information price a piece
_ TDA1543 3 _ 1,050 yen
_ CS8414-CS 1 _ 2,300 yen
_ 74HC74 2 _ 50 yen
_ 74AC74 2 _ 70 yen
_ optical signal reciever 1 TORX179 230 yen
_ regulator(5 in a pack) 1 NJM7805 250 yen
U1 crystal oscillator 1 67.108864MHz 300 yen
_ relay 1 Y14H-1C-5DS 80 yen
Q1 transistor 1 2SC1815 21 yen
_ 22pF 1 ceramic capacitor 10 yen
C2,C3 0.01uF 2 APS NISSEI 21 yen
C1 0.068uF 1 MTFF050J683 21 yen
D4 0.1uF(25 in a pack) 1 chip capacitor 100 yen
C5,C6 22uF 2 BG-NX low ESR and noise 189 yen
R2 470uF 3 OS-CON SP 250 yen
L1 47uH 1 TSL0709-470KR94 105 yen
R9-R11 22ohm 3 NIKKO 42 yen
R19 51Ohm 1 metal oxide film resistor 21 yen
R1 75ohm 1 RIKEN RMA 78 yen
R8 120ohm 1 NIKKO 42 yen
R2 470ohm 1 RIKEN RMA 78 yen
R3,R5 750ohm 2 RIKEN RMG 160 yen
R18 10kohm 1 metal oxide film resistor 21 yen
R4,R6 100kohm 2 NIKKO 42 yen
VR rotary potentiometer 2kohm 1 3296W 80 yen
_ case 1 UC9-5-12DD 2,050 yen
_ switching AC adaptor 1 9V 2.5A 600 yen
_ RCA jack red, black and white 3 MR-565 136 yen
_ blue LED(5 in a pack) 1 _ 100 yen
_ DC plug 1 MJ-14 63 yen
_ switch 1 T-881 84 yen
_ switch 1 2MS6-T1-B3-M1-QE-1 70 yen
_ universal board 2 ICB-93SG 341 yen
_ socket 3 8P 25 yen
_ socket 2 14P 30 yen
_ socket 2 line 20P 70 yen
_ socket 2 line 9P 40 yen
_ heat sink 3 TO-220 52 yen

I used OS-CON for decoupling, RIKEN RMG for the resistors of I/V converter and valuable parts for other important parts.

crystal oscillator OSC-004BTNF 67.10886MHz

It looked so nice, so I judged that it had to have high accuracy.
I was not particular about the frequency of 2^28Hz.
I believed that it was a nice one.

74HC74AP and 74AC74

These are flip-flop IC.
74AC74 has higher speed than 74HC74, but it needs more electric power.
I thought its difference would be very small.
So I chose 74AC74.

nichicon FineGold 3,300uF 16V

I used this for smoothing capacitor.
This was in my pocket luckily.
This catch phrase is "realizing ritch bass and well spread mediant and treble".
I liked this because it added beautifulness to the sound.


I separated boards for ditial and analog.

part of ditial

The orange color of polypropylene film capacitor looks nice.

This is a photo wchich I've set CS8414, TDA1543, OSC and 74AC74.

I tried it connecting to a I/V converter as a test not putting in a case.
Luckily it easily sounded, but its sound was noisy and distorted.

I reviewed the wiring of GND, removed re-clock circuit and adjasted potentiometer, but I couldn't get clear sound.

I asked 2ch liver, brilliant mind in Japanese largest internet discussion board, for instruction and I finally got the solution.
The resistance of I/V convertion seemed to be too high value.
I replaced 750ohm resistor with 680ohm one and adjasted potentiometer, then I got very clear sound.

I wanted to confirm the potency of para-using TDA1543, so I removed two tda1543, in other words it got a single TDA1543 DAC.
Then it got very clear and beautiful.
Single TDA1543 is much more marvelous than 3para TDA1543.

Para-using of TDA1543 was famous method to improve sounding.
Why it didn't suit in my case?
Was its reason time lag caused by the differences of the path-length?

I modified the I/V convertion circuit for single TDA1543.

part of analog

(image will expand if you click)

I changed the parameters of I/V convertion for single TDA1543.

I chose RIKEN RMG for R1,R3.

SIZUKI VE10Y225J 2.2uF 100V laminated metallized film capacitor

It costed 210 yen.
I forefelt it would work well.

The blank space was to add smoothing capacitors.
I guessed that I would probably want to convert something after completion.

I wired using gold-plated lead lines of RIKEN RMG.
I felt a bit of luxury.
And I added MKT1826 0.022uF capacitors for coupling which were luckily in my pocket.

processing case

I used UC9-5-12DD made by TAKACHI once again.

I'm happy it looks so nice regarding flush.

I regret I've not use a flash.

First floor is the part of digital and second is the one of analog.
Braid black and red cables reminds me of poison spider.

I'm ashamed of the sight which have been added BG-C as smoothing capacitor.

I used blue LED as a indication lump.
I drilled a small hole.


Because blue LED was too bright, I tried to set a creamy white acrylic plate.
It was very easy to process acrylic material.

This is a photo which DAC has beeing sounded with optical input.

impletion of this nos DAC
It was very clear and beautiful.
Its sound seemed to be like the one of USB DAC with PCM2704 which I remove about two veils of it.
NOS DAC was a bit inferior to USB DAC with PCM2704 which uses two OpAmps in bass sounding, but wasn't defeated in punch.
NOS DAS's bass had a stable center core, so it pressed me strongly.

I was happy to have closed in true sounding.

I felt coaxial input from CM102-A+ was better than optical input from Mod SoundBlaster Digital Music LX because it was thicker.
It was strange that they had to be the same digital signals.


I'd come to the conclusion that removing high-frequency noises was the assured way for Hi-Fi because DAC used high-frequency.
So I took a action to remove.

mesure 1:
Setting a lot of capacitors which had good high frequency characteristic into the power supply legs of IC, OSC and regulators.

NISSEI APS polypropylene film capacitor 220pF 15yen

It looks like vitamin vitamin drops, we may be able to get carotene.

measure 2:
In order to suppress switchinig noise of regulator, setting a PI Low-Pass-Filter which is built by capacitors and toroidal core coil.

toroidal coil TC-300M-2A-5026 30uH 2A 52yen

This is like licking candy.

These were very efficient.
It got much more clear like urban escape.

I felt 2.2uF was even low for cappling capacitance by simulating with LTSpice.
Its bass might be cut off, so I added 2.2uF furthermore.

NISSEI MTF laminated metallized film capacitor 2.2uF 147yen

It served its purpose, bass got more powerful.
It may be good to adopt para-using film capacitors in couppling.

I was about to eat them by mistake.

It gets beautiful in night.
When I watch a movie without lamp, it produces a flavor of romantic mood.

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.