Building the ultimate NOS DAC using TDA1541A

[fff0]

Thanks again for sharing your insights.
Allow me to touch again on the topic of NOS. From your DAC output (E1~E4), I don’t see any filter or circuitry to reduce/eliminate unnecessary harmonic components unless you had left it out. I understand that a LPF(at least) is somehow needed to reduce any intermodulation distortion (let alone side band component).

From a patent shown here: Analog filter for digital audio system and audio amplifier for using the same - Patent 6721427 , it explains that:
The output signal contains a basic frequency component up to a frequency of 20khz as well as upper and lower side band components of the integral multiples of the sampling frequency Fs. …….. If these unnecessary components are reproduced by a speaker, cross modulation distortion may occur.

I believe due to the above, it’s an inherent problem of NOS thus leading some of the listener to hear some sort of distortions.
Do you believe in these type of filters are not needed at all in your experience? Is it possible that you share with me your experience or belief or knowledge on your decision even If you don’t believe some sort of filter is needed?

Meanwhile, correct me if I am wrong. I guess that your speakers and passive crossover might have already implemented some of that preventive/ elimination functionality in terms of sideband and cross modulation freq components. Is this how you solve the problem, if not did you do something after E1~E4 to resolve these?

PS: I want to know NOS is the way to go but how, using engineering terms.

[galeb]

Hi ECD, just one question;

I had the oportunity of observing 3 pcb from the AMR CD77. Let me explain, an owner has this incredible CD player, (I've heard it) but, there were a storm and the CD player and others things burned. Then, the importer of Spain, wich one I know, gave me 3 pcbs (uau¡¡):

1-servo+D/A
2-1st supply (general one)
3-one channel output

I've saw that there is one pre-regulation, (discrete), and, on the servo bd, there is more regulations, like TL431, and so on. My question is that if you believe, that even with a well designed supply, the second regulation isn't necessary.

Best regards,

[-ecdesigns-]

Hi fff0,

Allow me to touch again on the topic of NOS. From your DAC output (E1~E4), I don’t see any filter or circuitry to reduce/eliminate unnecessary harmonic components unless you had left it out. I understand that a LPF(at least) is somehow needed to reduce any intermodulation distortion (let alone side band component).....

My objective is to keep the signal path as simple and straight-forward as possible.


Let's follow the signal in the ISD player.

DAC current output passes a single JFET current buffer and runs through the 500R passive I/V resistor into +5V. By doing so, all bit currents flow into +5V.

The voltage across the 500R passive I/V resistor is buffered by a discrete JFET unity gain buffer. The output at the buffer has large bandwidth (oscillograms posted earlier).

Next a hybrid coupling cap with DC compensation circuit is used to put the coupling cap in a "sweet spot" where distortion is lowest. This signal then enters a frequency compensated 10K motorized ALPS pot (low noise conductive plastic film).

Signal is then fed into a balanced FET bridge power amp that remains stable up to at least 10 Mhz. I added a concept schematic of this power amp.

Interlinks are made using RF litz wire.

The bridge power amp is connected to the speakers using RF litz wire (8mm diameter). I added an oscillogram of the power amp output signal.

The sonic resonator passive filters are designed to handle large bandwidth input signals. There are 4 main filter parts, two multi-segmented honeycomb air chokes and two hybrid caps that have an extra PCB bypass cap. Two speakers, a bass-mid chassis and a tweeter are driven by this passive filter.

The signal on the woofer looks analogue when measuring it with the scope. This is caused by the passive xover filter. In other words, no visible steps, just a smooth signal.

The signal on the tweeter has step shape, but the tweeter starts to roll-off at approx. 25 KHz, removing RF energy by speaker mechanical limitations (RF converted to heat / losses).

Finally the human auditory system performs required brickwall filtering on a signal that's already band limited to approx. 30 KHz by the tweeter.


This method places the filtering at the end of the signal path instead of after the DAC chip. It uses given speaker properties that already provide band limiting starting at approx. 25 KHz.

[Bernhard]

Originally Posted by -ecdesigns-

I also added measurement of music reproduced by the SD-player. The signal transients between samples are almost invisible as jitter levels at the DAC output are extremely low (<40ps rms) and bandwidth is larger than 30 MHz. This in turn is required for accurately reproducing sample pulse energy and achieving high dynamic resolution.

The resolution in time domain in your oszillogram is very small, if it was higher, one could easily see the settling time of the DAC. Where do you expect to see jitter in such an ultra low resolution measurement, how many samples on the scope screen, I did not count them ?
But that is not the point, just a notice.

[Bernhard]

Originally Posted by -ecdesigns-

The bridge power amp is connected to the speakers using RF litz wire (8mm diameter). I added an oscillogram of the power amp output signal.

Still the same very low resolution in the oscillogram, but now one can easily notice settling time.
It seems you have unintentionally created a situation where the power amp acts as substitute of what is usually found in DACs: the active I/V stage which badly affects settling time of the DAC, be there NFB or not.
Look at your pictures, obviously the power amp can not follow the short rise time of the DAC chip.

Another notice:
The tweeter provides hf rolloff, but far from the required steepness.
Use google to find out that despite ultrasound can not be heard directly, it causes audible intermodulation distortion ( difference tones ) inside the ear.

[fff0]

Hi EC designer,

Have you ever tried a sin(x)/x filter ? It seems according to this user that a sin(x)/x filter is important, how to design sinx/x filter for TDA1541A? .

If not, may I know whats your rationale?

[oshifis]

I made a test CD that has one track 0 dB 22.05 kHz square wave (alternating 0000 and FFFF samples). I got exactly 0 dB 22.05 kHz squre wave during playback on a NOS DAC. So where is the sin(x)/x rolloff? Perhaps I have to try with pink noise recording and an FFT analyzer, which I haven't done yet.

[Bernhard]

So where is the sin(x)/x rolloff?


Try 1kHz and 20 kHz sine wave.

[-ecdesigns-]

Hi fff0

Have you ever tried a sin(x)/x filter ? It seems according to this user that a sin(x)/x filter is important

I tried passive circuits that provide certain amount of trebles boost. I removed them shortly after as I didn't like the sound at all, it didn't sound natural and caused listening fatigue.

When applying "mild" analogue filtering after the DAC, higher frequencies are sine wave shaped The filters also introduce phase distortion.

Without any filtering after the DAC these higher frequencies have square wave shape. If I am correct square wave shape has form factor of 1 and sine wave shape has form factor of 0.707.

Measurements of AC magnitude : BASIC AC THEORY

This means that without "mild" filtering I end up with more power for trebles at the tweeter, and thus don't necessarily need a trebles booster circuit.

[Bernhard]

Originally Posted by -ecdesigns-

I tried passive circuits that provide certain amount of trebles boost. I removed them shortly after as I didn't like the sound at all, it didn't sound natural and caused listening fatigue.

I have quite the opposite impression.
Treble rolloff results in a sound that is lacking transparency.
If a linear freq. resp. causes listening fatigue, then the problem is elsewhere in the audio chain.

Originally Posted by -ecdesigns-

When applying "mild" analogue filtering after the DAC, higher frequencies are sine wave shaped The filters also introduce phase distortion.

Unfiltered doesn't sound homogenous.

Originally Posted by -ecdesigns-

If I am correct square wave shape has form factor of 1 and sine wave shape has form factor of 0.707.
This means that without "mild" filtering I end up with more power for trebles at the tweeter, and thus don't necessarily need a trebles booster circuit.

but:

Originally Posted by -ecdesigns-

The signal on the tweeter has step shape, but the tweeter starts to roll-off at approx. 25 KHz, removing RF energy by speaker mechanical limitations (RF converted to heat / losses).
Finally the human auditory system performs required brickwall filtering on a signal that's already band limited to approx. 30 KHz by the tweeter.

Very interesting:

Where is your extra power for trebles gone ?
Transformed into thermal energy heating up your ear and your speakers.