Which DACs are interesting for modest DIYs running 16bit/48khz?

[Hayk]

I have great hopes from PCM1728. This DAC looks to have been given great car for 16bit performance. If I subtract the noise from THD+N at -60db it results 0.4%. Now if you look the spectrum bellow, odd harmonics are absent that can result near 0.1% THD in differential mode, less than what TDA1541A ss can 0.25%. I will give particular attention to the output interface.

 

[Hayk]

Pioneer, uses this DAC renamed PE8001A in high end model PDS707.

The voltage outputs are only 3vp-p. I will drive the Sony LP filter directly from this plug-in, in differential without passing the OPA1632 differential.

 

[Hayk]

As I will bypass the OPA1632 I/V converter for voltage output DACs, I modified the SAA7350. The output opamps, of same phase, are paralleled with single feedback resistor of 15k instead of 30k, and the differential voltages will be presented directly to the Sony filter.
Pioneer pd07, which uses 1 SAA7350 per channel, advertises <0.0015 THD with 16bit, what the AD1862 achieves with 20bit.

 

[Hayk]

Problem with AD1853.
The digital filter disabled, the DSP left justified format is rather special.
First, the MSB needs to be delayed 1 clock.
Second, I don't understand how the LRCK should be.
I am even wondering if it is necessary to have the AD1853 along the AD1955.

 

[Hayk]

I am grouping again the DACs on PCBs, this time with the bit No order.
16 bits PCM56,(AD1856,AD1851,) DIP16&SMD TDA1543-5A. Can AD1860-1.
18 bits AD1860SMD, AD1865PLCC
20 bit. SAA7350, PCM1702
24 bit. PCM1728,WM8740, AD1955, CS4335.
The TDA1540 is seperate, not a plug-in.
I find the AD1860 unjustly neglected by DIYs. An example of 4xAD1860 is Sony cdp990.
Listen on youtube
2 SONY CDP990 by Marco C.

 

[Hayk]

When I listen to AD1860 and AD1865, the 60 is not engaging, more a nice sound to hear than to listen. On the other hand, the pcm69, although measurments are inferior, has very vivid, dynamic sound.

I will pair the AD1865 with PCM69, the 1860 can be used on the pcm56 board.

 

[Hayk]

The circuitry of 24bit DACs is nearly finished. The AD1955 only needs I/V converter, the others feed the Sony filter directly. The AD1955 needs another 74hc74 to function with bypassed digital filter, whereas the WM8740 needs to function the DF bypass to receive 24 bit data in Right justified parallel . All other DACs receive left justified series. A jumper will decide.
As you see, I still have some space left, I see after the wiring what I can add.

 

[Hayk]

Since the beginning, the PCM1702 is not convincing me for quality.
Over 25 years ago, @Bernhard on DIYaudio has tested in Denon dcd2700 in differential with deceptive results.
Post in thread 'New DAC measurement' https://www.diyaudio.com/community/threads/new-dac-measurement.49778/post-694492
I listened to several Denons, Harman kardon, both in 2x and 4x, I find metallic sounding.
The best I found is a serious Sony CDP D50 with 2x only.

I don't believe 4x makes any sense, asking people spend so much money for nothing substantial.

 

[Hayk]

I found, not easy, the schematic of the Sony above. Nothing special but I will do the same.

 

[Hayk]

The I/V converter, opa603 should play important role in the sound quality. This CFB single opamp is obsolete and sold over $5. Luckily the smd version available from Ali costs $5 for a pair with ship. To note that this opamp is not intended for audio with low noise low distortion, but what I can guess, the metallic sound can be the result of odd harmonics in phase. This type of harmonics act as dynamic expenders, as you can hear when Dolby is not activated where it should. The distortion of OPA603 might be acting inversely. Or else they would have used the NE5534.
I will not adventure in other opamps, so that, someone who will be interested, will know what exactly will get.

 

[Hayk]

The OPA603 has a unique input impedance in I/V mode. It has flat to 20khz 0.1ohms.
The graph bellow is 0db means 1ohms, -20db=0.1 etc.
To compare with VFA opamp, see next the LM4562.
The problem with 603, it cannot deliver better than 2vp with +/-5v supplies.
The opa683 has exactly the same Z and it comes with disable pin and doesn't have trouble to output 3v.
My I/V converter OPA1632 has a Z of 0.3 ohms but increases at high frequencies. I wonder if I need to replace it.
Another CFA OPA 694 available in double, opa2694 has 0.14ohms and also good with +/-5v.
Normally the complience of Philips outputs is +/-25mv. With 1-2ma it needs maximum 25-12ohms. Maybe this type of R2R DAC needs very low Z.

 

[Hayk]

The premium choice is the dual $10 (mouser) OPA2684, 22mohm.

 

[Hayk]

I replaced the opa1632 with opa2684. Lcsc has the SOT23/8 version about $10.
It needs the reference voltage extra pin to be 0v for symmetrical supply DACs, 3.6v for the TDA1543-5 and pcm69, 2.4v for AD1955 to be supplied on the plug-ins.
The SAA7350 has internal unity follower opamp to supply low impedance reference, someone experimented by an external lower impedance opamp with better result.
I will see if I do it too.

 

[Hayk]

The OPA2683, one may think is dual 683, no it isn't, the DS are seperate.
It is more popular, has 44mohm Z and costs $5 and in SOIC8.

 

[Hayk]

Printing error.
The Sony opa603 is wrong written, it is opa604 which is jfet opamp.

 

[Hayk]

Strange that it doesn't resemble to ordinary VFA as the LM4562.

 

[Hayk]

The opa2604 is a +/-24v opamp. It is obsolete but available from LCSC in SOIC8 for $7.5. I replaced on the motherboard the opa1632 with opa2683. The SAA7350, also received it and I added for the pcm1702 an opa2604 optional, the user to choose the I/V needed.

 

[Hayk]

The DS of PCM1702, preconises the OPA2604, see bellow.

 

[Hayk]

The AD1862, also 20bit DAC, the I/V preconised for high resolution is CFA opamp AD846, obsolete instead of SE5534.
A used one costs $10.
An AD8001 costs $6.5 from mouser with 120mohm.
To show that the dual OPA2683 is an excellent choice.

 

[Hayk]

This is the circuit of the motherboard.
It receives the SRC4190 board covering the shift registers with male-female pinheaders.
The SRC board needs some preparations.
The pin 1 track link to ground needs to be cut and 100k pull up resistor from 3.3v applied. This allows to select the interpolating filter delay.
The Vdd on the output side needs to be isolated from the bottom side, to receive instead PO8//PO9 to select Right or left justified, open is right.
The PO3 is shorted to be 96khz.
The output connector has 2 extra pins to select the bit dept following the MLCK pin. An extra hole is drilled for bit1 and the screw hole receives the bit2.
The PO7 has the 2 pins, the 16 and 17 parallel to select 16 or 24. The lead of pin 17 needs to be lifted up to seperate. It can be done using a scalpel and an extra blade to avoid the lead to bent sideways. Even if the lead gets broken from the root, is not a problem, you can scrub the plastic to reach the copper and solder a wire to link the connector with a 100k pull up resistor. The PO7 links the last connector pin.