Upgraded Single Board PCM1794 NOS DDDAC

With the help of many contributors to the http://www.diyaudio.com/forums/digi...dac-pcm1794-waveio-usb-input.html#post3252373 thread I implemented a series of upgrades to Doede’s DDDAC which resulted in my single board DDDAC-Buffer-Cinemags sounding far better than I thought ever possible by any DAC. Doede’s brilliant NOS PCM1794 design and the optimized operating parameters are the foundation on which these upgrades are applied.

I reported my upgrades on the main forum thread but they are so scattered that I decided to put them all together here with even more explanation and pictures. Everything except the buffer applies to multiple DDDAC stacks as well.

The following is a summary list of the upgrades in order of sq improvement/cost:

1. Digital Sources. I have a WaveIO with a Belleson Vreg and simple reclocker which I am very happy with. I plan to try a BBB with an Acko reclocker because of the many posts that the rpi and BBB sound even better. Clean digital source stream is a must.
2. ½ clock delay circuit. If you have a blue Main Board then you already have this. If you have a red Main Board then you MUST add a 2 chip circuit which provides this I2S Data delay. Four out of 4 persons who have done this report a major improvement in sq. Music and voices sound more coherent and real. See group buy for 2 chip PCB at end of thread.
3. Analog 8VDC power supply. I use a Salas Shunt power supply. Others have had great results from Tent Shunt supplies. A Belleson series Vreg can also be used. This increases the musical energy and dynamics as well as more detail. This power supply may partially offset the need for a buffer.
4. CCS. The standard circuit for pin 20 is a 6K resistor to ground. Changing the resistor to a premium type provides very little improvement in sq. Single JFET CCS circuits using the 2SK246Y or the 2SK208 “O” or “R” sound more “musically exciting” and increase detail. When changing back from a CCS to the 6K resistor on pin 20 the sound became a bit compressed with less detail by comparison. Why does the CCS circuit sound so much better than a resistor? I've concluded that it actually makes the PCM1794 DAC chips sound better. Both JFETS also have a low temperature coefficient. Toshiba officially lists the 2SK208 as the replacement for the discontinued 2SK246. They sound the same to me. See Assembly instructions for CCS circuit board. http://www.diyaudio.com/forums/digi...e-board-pcm1794-nos-dddac-37.html#post4229210
5. Buffer. A single board DDDAC is inadequate to drive the DAC output through a transformer without sounding a bit thin and weak in the bass. The music sounded pleasant and polite but I was left wanting a more real musical experience. Four or more DAC boards can properly drive the signal through a transformer. The forum consensus suggests that a partially upgraded single board DAC sounds better in some ways than a standard 4 DAC stack. A fully modded 4+ DAC stack sounds incredible but is expensive and tall and requires a huge power supply. A fully modded single board DAC with an output buffer is the alternative. The buffer increases the musical authority, bass and detail through the Cinemags. The Cinemags no longer sound weak or thin. The downside of the buffer is that it requires +12V and -12V power supplies, 4 matched pairs of BF862 SMD JFETS, and a PCB. This post has more explanation of the Buffer: http://www.diyaudio.com/forums/digi...e-board-pcm1794-nos-dddac-11.html#post4163372 Direct feed NOT recommended as discussed in this post: http://www.diyaudio.com/forums/digi...e-board-pcm1794-nos-dddac-18.html#post4173717 See GB for CCS circuit board.
6. Digital Vreg. I use Belleson SPM 3.3V regulators on the PCM1794 digital circuits. Others report using the Tent shunt regulator with success. Even if the standard LF33 Vregs are not replaced, the digital caps should still be OSCONs. I also use Belleson SPM 5V regulators with additional caps on the main board. The sound improvement from the Belleson’s and OSCONs was modest, but the music sounded more “settled down” and the details seemed to have much fewer unnatural artifacts.
7. Rload Resistors. A premium resistor like the tantalums, Rhopoints, TX2575, or Susumu SMD sounds clearer than the standard Dale or the PRP. I use tantalums and Susumu’s on my two DDDACs. I also tried 120 ohm tantalum resistors which didn’t sound quite as good as the 135 (2x270) ohm Tantalums. Doede tested many different resistor values and 133/number of DAC boards is the sweet spot for sq. I tried 135 (2X270) ohm Susumu RG's with success. See http://www.diyaudio.com/forums/digi...e-board-pcm1794-nos-dddac-32.html#post4202948
8. Cinemag Load. Having a load of 8K-15K on the outputs of the Cinemags, including the preamp input resistor, sounds best. Having a 220k ohm preamp load sounded unstable in the higher frequencies to me. Adding extra resistance in the output requires high quality resistors. I use two LDRs in series set at 11Kohms. Dwjames and I tested Wima FKP 3 film caps on the Cinemag outputs and found that 150pf brings the musical detail into proper focus. Higher capacitor values start to collapse the sound stage.

I will add posts to this thread which provide more details on each of the above items. In order to keep this thread from getting “scattered” please confine comments to the buffer only. All other topics should be posted on the http://www.diyaudio.com/forums/digi...dac-pcm1794-waveio-usb-input.html#post3252373 thread because they are of more general interest. You can also send me a PM if you wish.
 
Last edited:
CCS on Pin 20

Many different constant current devices and circuits have been tested on pin 20 for sq improvements and temperature stability. Two JFET types meet these requirements. The 2SK246Y is a TO-92 style and not in production but is readily available and affordable. Make sure to get the "Y" version. The 2SK208 SOT-23 JFET is in current production with the “O” and “R” versions sounding great. In fact Toshiba officially lists the 2SK208 as the replacement for the discontinued 2SK246Y which may explain why they sound so similar. The 2SK246Y has a small negative temperature coefficient. The 2SK208 “O” has a slight positive temperature coefficient no greater than 1% from 25C to well above 40C.

See circuit below. The JFET source resistance values vary considerable depending on the JFET device used. A 1000 ohm fixed resistor and a 2000 ohm pot cover the entire range of 2SK246Y JFETS with 1800-2500 ohms covering most of them. A fixed resistor is a more temperature stable device, so it is desirable to have as much resistance from the fixed resistor and the least from a Bourns pot. A Susumu RG SMD resistor is preferable to a Dale RNxx. A bread board can be used to determine what resistance value gives the desired 0.40 ma so that the optimum components can be purchased. This circuit can be easily made using a perf board.

The 2SK208 JFETS require much less source resistance. I used a 470ohm fixed resistor and 500 ohm pot for the “O” types I tested. A forum friend reported that 308 total ohms provided the desired 0.40 ma for his “R” types that measured an Idss of 0.60-0.63 ma.

Also below is a mini-circuit board that was made when cascode circuits were being tested and only through-the-hole CCS devices were being considered. The JFET and Bourns pot can be installed from either side or any orientation depending on the space requirements of your surrounding DAC components. If the ground termination end of the 6K resistor is too difficult to reach then run a wire from any ground "star" pad hole to the Rload common or any other analog ground point. The next revision would have alternate pads for a SOT-23 JFET and probably eliminate the cascode circuitry. Please send me a PM if you have any interest in these CCS boards or in a future revision.
 

Attachments

  • CCS Circuit.jpg
    CCS Circuit.jpg
    58.5 KB · Views: 5,951
  • 2SK246 CCS.jpg
    2SK246 CCS.jpg
    362.7 KB · Views: 5,793
Member
Joined 2010
Paid Member
Thanks, subscribed.

I'd be interested in a little board which will allow for the one SOT-23 package + resistor/potmeter. Any suggestions anyone?

From the post from carlsor the small board already exist as per his initial circuit with cascode and led. Should not be so difficult to redesign a new small board with the SOT 23 with or without cascode.
 
Thanks, subscribed.

I'd be interested in a little board which will allow for the one SOT-23 package + resistor/potmeter. Any suggestions anyone?

I am thinking of using the resistor and transistor in smd package, it will not be difficult soldering these to the legs of a bourns multiturns pot. This will keep the footprint small for easy mounting at the resistor position on the dac board.

Regards,
 
1/2 Clock Delay Circuit

This upgrade takes most of an afternoon to build and install, but is well worth it.

These are my initial sq impressions from this change: I hear MORE music!
- Cymbals sound 100% right for the first time.
- Voices are more articulate
- Sound is more crisp and vibrant. No smoothing here.
- Same amount of detail, but it sounds more coherent and "right"
- I got drawn into musical nuances of familiar CDs that I never noticed before.
- More focused placement of instruments in the soundstage. I could almost hear the size and shape of instruments.

The board is about $10 usd of parts, but I also bought a 5V Belleson SPM78 regulator ($36 usd) to comply with the Blue Main Board voltage. Yes, a 5V regulator makes the DAC with this upgrade sound more accurate and smooth compared with my original 3.3V Belleson. According to the data sheets the 74VHCxxxx chips switch state many times more quickly when operating at 5.0 volts. Adaptor boards can be bought on eBay but could take a month to receive. 3pcs SMD Converter Adapter PCB SOIC14 LGA14 to DIP14 | eBay

The I2S resistors should all be 100 ohms – the 1000 ohm resistor for DATA must be replaced with 100 ohms! The bottom of the main board had to be modified by cutting BCK foil traces and adding jumper wires. See pictures below. You can see the delayed signal wire from the 2-Chip Board going through one of the clock foil trace holes to the five 74VHC164 chips.

I initially built and installed a 1/2 clock delay circuit above the main board. Later I built a second 1/2 clock delay circuit board but installed it on the bottom of the red main board which puts it out of sight. I think some of the wires lengths may be a little shorter, but this location is no less time consuming to implement. I used 24ga CAT 5e solid wires. For multiple connections I loop a single wire through the points as much as possible.

For the bottom installation I drilled an M3 hole through the adaptor board pair and main board with a plastic nut spacer and screw to hold it firmly in place.

The Belleson and some other Vregs need more load than from the seven 74VHCxxxx chips to provide optimum regulation and sound their best. The connection to the WaveIO isolator provides enough extra load, but if you are using an rpi or reclocker with its own isolator you may need to add a 330 ohm resistor across the Vreg load.

My engineering mind cannot figure out why new components, circuits, PS, etc in the digital domain of 1's and 0's have to go through a burnin period, but my ears tell me this is necessary.

Below is the 74VHCxxxx chips circuit diagram pulled from page 3 of the DDDAC 1794S NOS manual that should be used as a guide for wiring and installing this circuit board. Also included is a BOM and installation pictures.
 

Attachments

  • dddac1794 delay half-cycle.jpg
    dddac1794 delay half-cycle.jpg
    483.8 KB · Views: 2,251
  • 2_Chip Back.JPG
    2_Chip Back.JPG
    346.6 KB · Views: 2,135
  • 2_Chip Installed.JPG
    2_Chip Installed.JPG
    551.5 KB · Views: 2,070
  • Main Bot Mod.JPG
    Main Bot Mod.JPG
    289.3 KB · Views: 1,722
  • Bot half clock delay board.jpg
    Bot half clock delay board.jpg
    352.1 KB · Views: 1,712
  • top half.jpg
    top half.jpg
    291 KB · Views: 717
  • BOM half clock delay.jpg
    BOM half clock delay.jpg
    254.1 KB · Views: 687

I'm also thinking to first build the CCS’s and calibrate them, then replace the pots with fix resistors, to reduce the tempco difference the pots are prone to. I think the tempco behavior is crusial when they are stuffed in between a dac stack.

I’ve ordered some 2SK246Y from Ebay. That way I can just work with a bit of perfboard if I have to.

My workbench still needs setting up. Work on my kitchen is taking priority at the moment..
 
1/2 delay clock for red motherboard dddac 1794

Carlsor advises to insert 1/2 delay for the red motherboards for better SQ.

I am wondering if someone has created or will create a PCB for this, having a Group Buy. I am interested to order, but have not the knowledge myself to produce.