Building the ultimate NOS DAC using TDA1541A

[poobah]

The end result is linear interpolation... your filtering must still be designed accordingly.

[rfbrw]

Someone did something similar a while ago in order to create a fast enough data rate to allow the use of the external digital filter interface of the PCM1792, bypassing the internal digital filter and with some success it would seem.
Must say from a purely technical viewpoint without the extra samples created an FIR oversampling filter or something similar, there is likely to be a degree of error in the HF area due to a lack of samples though there now appears to be at least two cases that suggest this may not be a problem.
As for the title Ultimate, I'd have to take issue with that, not least because it uses the TDA1541A, but mainly because the Japanese audio crazies build these things with 64 dacs per channel, with or without a preceeding OS filter.

[-ecdesigns-]

Hi,

I try to post the Twindac diagram but couldn't get the file small enough. I will give it another try (This DAC uses the 8th order active filter and differential stage).
About the Direct Interpolation questions, there is one major difference with existing methods. Brickwall filtering is necessary to filter out the noise introduced during oversampling / digital filtering. The Direct Interpolation DAC uses no oversampling, no decimation and no digital filtering, so there is no need for a brickwall filter. Yes the output signal of the Direct Interpolation DAC still has steps, but these are 4 times the ANALOG resolution (18 bits). Because the virtual sample rate is now 176.4KHz a standard 2nd order EXTERNAL filter is sufficient, just like in the standard designs that do use oversampling and digital interpolation filters. And remember, the prototype is working perfectly without interference, the output signal is very clean and has no steps after the 2nd order Bessel filter, so this is not just a theory.

[analog_sa]

Originally posted by -ecdesigns-
Hi analog sa,

The white connector you referred to is actually the analog output. I use a custom made fully differential SPDIF interface like in professional sound studio equipment. So no 75 OHm coax cable, certainly no TOSlink, actually I am using standard SVHS cable for this purpose. So I have a normal and an inverted SPDIF signal to cancel out interference that causes jitter. Obviously I had to modify my CD player

I have no doubt the 8412 is much happier with your interface. Still, one may question how well the 110 ohm terminates the mixture of cables and connectors.

Your mixed signal interpolation idea appears interesting and elegant to me but the chain of 5 opamps seems absolutely frightful and against the NOS minimalism.

[-ecdesigns-]

Hi all,


Here some information about the Direct Interpolation DAC.

Twin Direct Interpolation by parallelling DAC outputs:

It uses the WS signal to split up the 64 BCK pulses long sample into two 32 BCK long samples DAC1 receives the standard signals and outputs L and R. DAC2 receives an inverted WS (wordselect) causing a 32 BCK long delay (WS determines when the sample appears at the DAC output). This all creates 2 samples of 32 BCK pulses each: 1-32, 32-64. so the virtual sample rate becomes 44.1 * 2 = 88.2 KHz. This can be verified by observing the output signal with an oscilloscope. Because all analog outputs are ADDED, the following intermediate steps are created:
1-32 (DAC1)
32-64(DAC1+DAC2)

These addittional steps increase the output resolution to (2*65536=131072) equivalent to 17 bit resolution as DAC output voltages are added stepwise.

Here is an overview of how to connect each of the 2 DAC's

DATA DAC1 = DATA
BCK DAC1 = BCK
WS DAC1 = WS
L OUT = L OUT
R OUT = R OUT

DATA DAC2 = DATA
BCK DAC2 = BCK
WS DAC2 = NWS (inverted)
L OUT = R OUT (swapped)
R OUT = L OUT (swapped)

So basically if you already have a twin DAC and you want to try this, just invert WS on the second DAC and swap its L and R outputs. You will get 88.2KHz virtual sample rate and 17 bits. If you like a differential output stage, invert DATA of DAC2: NDATA


Quad Direct Interpolation using differential output stage

It uses the WS signal to split up the 64 BCK pulses long sample into two 32 BCK long samples then a 16 BCK delay (two sets of 2 * SN74HCT166 cascaded, one for DATA the other for WS) is applied to both DATA and WS of DAC 3. DAC4 gets the same treatment as DAC2 plus an addittional 16 BCK pulses so 48 BCK pulses delay. This all creates 4 samples of 16 BCK pulses each: 1-16, 16-32, 32-48, 48-64. so the virtual sample rate becomes 44.1 * 4 = 176.4 KHz. This can be verified by observing the output signal with an oscilloscope.

Because all analog outputs are ADDED, the following
intermediate steps are created:
1-16 (DAC1)
16-32(DAC1+DAC2)
32-48(DAC1+DAC2+DAC3)
48-64(DAC1+DAC2+DAC3+DAC4)

These addittional steps increase the output resolution to (4*65536=262144) equivalent to 18 bit resolution as DAC output voltages are added stepwise.

Here is an overview of how to connect each of the 4 DAC's

DATA DAC1 = DATA
BCK DAC1 = BCK
WS DAC1 = WS
L OUT = L OUT
R OUT = R OUT

DATA DAC2 = NDATA (inverted) causes 32 BCK pulses delay but swaps outputs
BCK DAC2 = BCK
WS DAC2 = NWS (inverted), causes 32 BCK pulses delay
L OUT = NR OUT (swapped, inverted)
R OUT = NL OUT (swapped, inverted)

DATA DAC3 = DLY16CP DATA (delayed 16 BCK pulses using 74HCT166 shift register)
BCK DAC3 = BCK
WS DAC3 = DLY16CP WS (delayed 16 BCK pulses using 74HCT166 shift register)
L OUT = L OUT (parallel to DAC1)
R OUT = R OUT (parallel to DAC1)

DATA DAC4 = NDATA DLY48 (inverted)causes 48 BCK pulses delay but swaps outputs
BCK DAC4 = BCK
WS DAC4 = NWS (inverted)
L OUT = NR OUT (swapped, inverted, parallel to DAC2)
R OUT = NL OUT (swapped, inverted, parallel to DAC2)

[-ecdesigns-]

Hi, analog sa

You are mixing up two things. The posted diagram is the Differential DAC with 8th order active filter.

The new Direct Interpolation DAC basically needs no filter.

High quality OP-amps (OPA627) are used in the "old" DAC and they sound very clear. It is true the OP-amp quality makes or breaks a design, and they have a bad reputation, but the OPA627 is really an exellent OP-amp. The distortion is extremely low. And more important, you can hear every last detail crystal clear. It sounds very realistic. Remember interference from the standard unfiltered NOS DAC output covers up details that are fully revealed when using the 8th order filter. The filter has been maticulously fine tuned to preserve every last bit of audio information, constantly comparing it with the unfiltered output. It's the actual sound quality that counts, not the "feeling" about a circuit.

[poynton]

Originally posted by -ecdesigns-
Hi, analog sa

You are mixing up two things. The posted diagram is the Differential DAC with 8th order active filter.

Hi.

He's not the only one a little confused !

So, the circuit diagram in post 23 is your old design shown in post 5?

The breadboard layout of your new design is shown in post 14.


What is the pcb in post 9?


Andy

[rfbrw]

Originally posted by -ecdesigns-
Hi all,


Here some information about the Direct Interpolation DAC.

Twin Direct Interpolation by parallelling DAC outputs:

It uses the WS signal to split up the 64 BCK pulses long sample into two 32 BCK long samples DAC1 receives the standard signals and outputs L and R. DAC2 receives an inverted WS (wordselect) causing a 32 BCK long delay (WS determines when the sample appears at the DAC output). This all creates 2 samples of 32 BCK pulses each: 1-32, 32-64. so the virtual sample rate becomes 44.1 * 2 = 88.2 KHz. This can be verified by observing the output signal with an oscilloscope. Because all analog outputs are ADDED, the following intermediate steps are created:
1-32 (DAC1)
32-64(DAC1+DAC2)

These addittional steps increase the output resolution to (2*65536=131072) equivalent to 17 bit resolution as DAC output voltages are added stepwise.

Here is an overview of how to connect each of the 2 DAC's

DATA DAC1 = DATA
BCK DAC1 = BCK
WS DAC1 = WS
L OUT = L OUT
R OUT = R OUT

DATA DAC2 = DATA
BCK DAC2 = BCK
WS DAC2 = NWS (inverted)
L OUT = R OUT (swapped)
R OUT = L OUT (swapped)

So basically if you already have a twin DAC and you want to try this, just invert WS on the second DAC and swap its L and R outputs. You will get 88.2KHz virtual sample rate and 17 bits. If you like a differential output stage, invert DATA of DAC2: NDATA

With no delay to the data to DAC2, there will be an error. DAC2 will convert the wrong data


.

[-ecdesigns-]

Hi rfbrw,

You are absolutely right of course, but I already fixed this problem this morning. I used the correct delay for DAC2 and DAC4 now 32 and 48 BCK delay and the outputs are no longer swapped, however it takes 12! 8 bit shift registers, but it is definitly worth it, I am quite pleased with the result now. I will post schematics as soon as possible

[tubee]

I am still busy, with more interuptions than procedings, with a four paralleled NOS 1541A dac to put it in a decent DCP. It will get its data by I2S, shift register divide the signals to the four dacs. I/V will be with tubes or nice opamps, not sure about that.
Is your design better then the this one, Ecdesigns?