Photos of D1V3 built by DIYers around the world. 
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
those look great finneybear...
no plans to do a vcxo board. the d1 secondary pll is reasonably dependent on the cs8412/4 so limits it's usefulness to d1v3 builders who choose to run newer chips. I'm also more interested in building a d1 that is closer to the Colburn/Pass design - the D1V3 is really very nice but leaves me pondering how the dac would sound with mosfet I/V, 78xx regs, and passlabs approach to supply filtering.
no plans to do a vcxo board. the d1 secondary pll is reasonably dependent on the cs8412/4 so limits it's usefulness to d1v3 builders who choose to run newer chips. I'm also more interested in building a d1 that is closer to the Colburn/Pass design - the D1V3 is really very nice but leaves me pondering how the dac would sound with mosfet I/V, 78xx regs, and passlabs approach to supply filtering.
Reading up trying to learn something about pcb layout and grounding techniques and came across some interesting references:
1) http://www.hottconsultants.com/pdf_files/june2001pcd_mixedsignal.pdf
2) http://www.hottconsultants.com/techtips/tips-slots.html
3) http://www.tentlabs.com/InfoSupport/Technology/page35/files/Supply_decoupling.pdf
4) http://www.sigcon.com/Pubs/edn/multipleadc.htm
These articles drew my attention to what the PCM63 datasheet says in relation to power supplies:
I'm possibly missing something here, but the D1V3 layout seems to break the "rule" that agnd and dgnd are to be connected to the same ground plane as close to the chip as possible.
If I'm reading the layout correctly the dac agnd and dgnd only communicate via L21. Because the two sections aren't connected internally, doesn't this mean the ground path for the digital and analog sections of the 4 PCM63 chips describe large loops that intersect at L21? Won't this degrade performance?
Please correct me if I'm misunderstanding something.
cheers
Paul
1) http://www.hottconsultants.com/pdf_files/june2001pcd_mixedsignal.pdf
2) http://www.hottconsultants.com/techtips/tips-slots.html
3) http://www.tentlabs.com/InfoSupport/Technology/page35/files/Supply_decoupling.pdf
4) http://www.sigcon.com/Pubs/edn/multipleadc.htm
These articles drew my attention to what the PCM63 datasheet says in relation to power supplies:
I'm possibly missing something here, but the D1V3 layout seems to break the "rule" that agnd and dgnd are to be connected to the same ground plane as close to the chip as possible.
If I'm reading the layout correctly the dac agnd and dgnd only communicate via L21. Because the two sections aren't connected internally, doesn't this mean the ground path for the digital and analog sections of the 4 PCM63 chips describe large loops that intersect at L21? Won't this degrade performance?
Please correct me if I'm misunderstanding something.
cheers
Paul
spzzzzkt said:those look great finneybear...
no plans to do a vcxo board. the d1 secondary pll is reasonably dependent on the cs8412/4 so limits it's usefulness to d1v3 builders who choose to run newer chips. I'm also more interested in building a d1 that is closer to the Colburn/Pass design - the D1V3 is really very nice but leaves me pondering how the dac would sound with mosfet I/V, 78xx regs, and passlabs approach to supply filtering.
1. The merit of simple reclocking circuit is questionable. Most of the time it will simply give you a different sound. One of the better solutions is to build something with FIFO, VCXO, PLL, etc, something similar to the Lavry solution. I know one guy in China is working on this through a FPGA. Or you can take a different route, switch to I2S. Another buy in China is selling an I2S interface board for CD transport. You will have to provide the recipient side yourself. Sure, everything has to start with a good CD transport. This then will be a major upgrade to D1V3.
2. MOSFET probably will give you a warmer, more forward sound, at the cost of some high end ultimate resolution. Yet you will have to know that the signal paths on D1V3 have been highly optimized. They are in the shortest possible paths.
3. LM317 is not necessary better than 78xx. The main difference is the noise level. Those LT devices are definitely good. Another thing is the response time. Fast response time does not necessarily mean better sound. So it's always worth of a try yet my take is that the output cap is more important than the active device itself. This is especially for digital power supply. We chose LM/LT devices on D1V3 simply because of the lower noises. As for the analog output power supply, the one we put in is fairly good already. You can try other solutions to see whether they sound better to you. The options of tuning in definitely plenty.
4. One major power supply upgrade I do recommend is to add one more power filtering stage. Separate transformers for each power rail. About 20,000-40,000uf power filtering for each rail. Use Rifa, BHC, or Epcos caps. Use big fat soft recovery diodes. I even dont mind if you use chokes. Pack all of these in a separate box. You may want to parallel the caps with smaller ones to speed them up. Keep the power caps on the DAC board as they are.
5. Try BlackGate N for the output cap. Big film caps have microphonic problems. Film caps with oil filled are the best yet you can hardly find affordable 10uf oil caps.
6. Some have replaced resistors on the analog output paths with SMD parts. Be sure to get the tantalum film ones. Definitely better characteristics than through hole parts.
7. After everything is done, seal the analog output parts with epoxy mixed with thermal paste. Cover the section with copper box, if you can.
8. At last, shield the board with EMI absorbing sheets. No need to get the super expensive multi-layer ones.
I can keep going and going....
spzzzzkt said:
Well, rules are just, rules. Not even mention those rules are not universal.
PCM63's datasheet says the benefit for separate analog and digital power supplies is minimal. This is hardly a surprise for most chip designers. Well, whatever power supply it will be, the ground planes are all connected together in most chips. In other words, when you do the layout floor planning for a PCM63 DAC, you will have to keep in mind that the analog and digital grounds are already connected inside the chip.
Any ground plane has current flowing. I.e. different points at the ground plane may have different voltage. One goal of layout is to eliminate the voltage difference and to provide paths to dump those currents as soon as possible.
What we did on D1V3 are:
1. Put ground planes on the top of PCB, instead of bottom. Along with the chassis, this will provide some extent of shielding for signals running on the other side of the PCB.
2. Divide the analog and digital ground roughly in two big planes. The boundry runs across the center of PCM63 chips.
3. We know that the grounds are connected inside PCM63s. Along the plane boundary, we also provide a few connectors through beads. Now we know any current between analog and digital grounds will flow through those pipes.
4. Since you know where those pipes are, you can provide adequate shielding around the pipes so the ground current will not interfere with other circuits. The beads will absord some HF noise current, too.
5. Now here's the trick. You may find tiny holes around the very corner of each ground plane, or you can scratch off the ink to expose the copper. Solder a thin silver wire to this hole, then run the other end to the middle of the ground plane. Repeat this to the rest three corners. Now solder all four wires together in the middle of the plane. You should be able to find a screw hole or whatever around the center. Solder the wires to this contact point. Now you have done the star ground for one plane. Repeat this to another plane. At the end you will have 2 star grounds. Either you can connect them to chassis separately or to a single point at the chassis. This star ground will be the short current dumping path.
6. Sure, you can make more star grounds. Even better, adding by-pass caps or inductors to absorb ground current.
7. We have done other tricks on the layout so most people cant see the need to go to this length. This is why we didnt mention it... yet since you are asking...
Oh, BTW, have you noticed right beside every IC there's a small hole? You stick a copper foil to the top of the chip, solder a wire to it then connect it to this hole. Nice?
spzzzzkt said:
You can not just move the board level concept into the silicon.
Basically there are certain safeguards built in around the pins. Also for different voltage planes, usually they are connected though diode bridges. Chip designs have to deal various conditions. For instance, if pin 7 and 12 are connected together externally, and the analog 5V comes in first, this will create a voltage jump in the digital domain and fry the circuit.
Some DAC chips are particularly fragile. CS43122 is one famous example. If you have separate power supplies for digital and analog, the chip will get fryed easily.
-finney
finneybear said:
You can not just move the board level concept into the silicon.
Basically there are certain safeguards built in around the pins. Also for different voltage planes, usually they are connected though diode bridges. Chip designs have to deal various conditions. For instance, if pin 7 and 12 are connected together externally, and the analog 5V comes in first, this will create a voltage jump in the digital domain and fry the circuit.
Some DAC chips are particularly fragile. CS43122 is one famous example. If you have separate power supplies for digital and analog, the chip will get fryed easily.
-finney
But that begs the question: how can you say the gnd's are internally connected or not if you cannot directly test?
spzzzzkt said:
I am afraid that you missed my point. They are connected inside through certain circuit, for instance, protection bridges hence you can not just measure the resistance between the two pins to decide whether they are connected.
Or take it this way. inside PCM63, the digital domain has lots of switches keep switching quickly. These will create lots of bounces on the ground plane. The analog domain has lots of bounces, too, when the R2R network is working. Definitely it will be nice to have separate power supply loops for digital and analog so you can reduce those circuit noises more effectively. Yet why BB claimed that you would not get any benefit with PCM63? Can you think of any reason?
Perhaps I did, but you are also evading mine
I'm not sure I know the motivation - I'm tempted to say because the the analog has the more demanding requirements than digital, so a supply which is adequate for the analog section will more than meet the needs for digital section.
Still seems like you are swimming against the tide regarding the dac grounding however:
http://www.analog.com/analog_root/static/raq/raq_groundingADCs.html
I'm not sure I know the motivation - I'm tempted to say because the the analog has the more demanding requirements than digital, so a supply which is adequate for the analog section will more than meet the needs for digital section.
Still seems like you are swimming against the tide regarding the dac grounding however:
http://www.analog.com/analog_root/static/raq/raq_groundingADCs.html
spzzzzkt said:Perhaps I did, but you are also evading mine
I'm not sure I know the motivation - I'm tempted to say because the the analog has the more demanding requirements than digital, so a supply which is adequate for the analog section will more than meet the needs for digital section.
Still seems like you are swimming against the tide regarding the dac grounding however:
http://www.analog.com/analog_root/static/raq/raq_groundingADCs.html
Tidewave? They all look more like Calculus 101 to me!
I think probably you know not much about chip designs hence have no idea about what I was talking about.
In most mixed signal chips, the digital ground actually has far more serious current bounce problem than the analog. Take the PCM63 as an example. The voltage swing on the digital ground is so server such that the ground is practically floating, a big no no for a high performance DAC. It's crucial to maintain a common reference point for the analog domain and certain circuit in the digital domain.
If you connect the grounds through external pins. First of all, you have to deal with the impedance of the I/O pads. The thin metal wires inside the chip have voltage drop as well. Eventually this reference loop is still way too long. A solution is to build a voltage reference inside the chip then refer both grounds to it. You need to build some sort of bridge for the digital ground such that the device will maintain the ground reference yet stop the digital noise/voltage swing running into the analog ground. Looks like BB engineers were so confident with their solution such that they felt separate power supplies are not required.
Those bridging devices are built in silicon. They are not active until you apply power to PCM63 so just picking up a PCM63 chip then measure the pins will not tell you anything.
Sure, it is still important to connect both ground pins together to reduce the capacitive load among the planes to improve the current dumping
If you look at D1V3's layout. Since we know both digital and analog planes have the common reference inside the chip, instead of being paranoid about connecting AGND and DGND pins immediately, we have to see how to provide the dumping paths for those ground currents. Since SMD beads are out of questions here, we could not join the digital and analog grounds under the chip through beads. Beads are critical here since we need to prevent the HF digital noises running into the analog plane. Direct connection between both grounds will not get you much benefit, instead, will just bring you more troubles. Our solution is to put the joins right beside the chip. And the connections are not just one, instead, we use two beads in parallel. We feel this is the best compromise.
We still use separate analog and digital power supplies mainly to get a better rejection rate.
As you can see, connecting both ground pins immediately together is just a basic concept. It's far more important to see what kind of ground currents you are dealing with and find out a good solution to address the challenge.
Analog power supply is very different from digital power supply. You cant say one is more demanding than the other, BTW. So what's your next guess why the datasheet would say separate power supplies cant gain you much?
Oh, probably you do not know D1V3 is a joined effort between Spencer and me?
dw8083 said:
Recommendation is recommendation, right?
Sure, if SMD parts are an option, it's a good idea to connect both ground right under the chip. However, you will have to keep in mind that SMD parts have very low current rating. AD1955 can surely generate lots of nasty ground current.. eventually you may have to have analog/digital joins somewhere else... then you are just creating another ground loop. Not a very pretty thing.
Anyway, it's just an eval board, isnt it? Dont expect it solves all the problems for you.
AD1955 is an excellent chip. The output current source is simply the best in the industry, very easy to live with. Hell, ADI chips all have better current outputs than BBs.
finneybear said:
Oh, probably you do not know D1V3 is a joined effort between Spencer and me?
I read diyhifi.org and have seen your comments regarding your joint project there, so I gathered that Spencer had called in the reinforcements in a couple of threads.
thanks for the explanation. I'd spent some time reading up on things on the web, in particular a walt kester tutorial that covered much the same ground:
http://www.analog.com/en/content/0,2886,761%5F%5F97529%5F65,00.html
Obviously you disagree with most of the references I've found on grounding strategies, but I guess that it the privilege of those who know enough to break the rules
spzzzzkt said:
Yes, yes, the ADI application note will put you more on the right direction, or I'd say, the real world scenario. The handling of ground planes is a very nasty nasty problem.
Our idea is pretty much the same as Fig.10. Using beads has a lot of drawbacks. The capacitive load, that is, the voltage Vn created between the noises in diginal and ground planes, is another headache. Both will hurt the LSB resolution of the DAC yet we figured, PCM63 was not famed for resolution anyway. Nowhere it can compete with modern delta sigma gens such as PCM1794, CS43122, etc, so we took the compromise. We just wanted to put out a design which was fun and flexible. It would give out the best possible performance for DIYers. Beyond that, probably it will be a board nobody can handle manually.
The soul of D1V3 is the I/V stage by Mr.Pass. This I/V and PCM63 simply have the syngery. The impedance curve, the current rating, the bandwith, simply have the best match. You cant help but just call Pass a genius! Anyway, this simple I/V with no NFB can simply bring out the best of PCM63. The strength is in the image, the sound stage, the advantage of a good R2R DAC. It also has a very smooth and relaxed sound.
We tried to mate this I/V with PCM1798... well, just didn't work. The bottom is a bit thin. Now I see why Charlie had to use transistors in his Ayre machines.
Saw this in the article you posted the link:
"Most ADC, DAC, and other mixed-signal device data sheets discuss grounding relative to a single PCB, usually the manufacturer's own evaluation board. This has been a source of confusion when trying to apply these principles to multicard or multi-ADC/DAC systems. The recommendation is usually to split the PCB ground plane into an analog plane and a digital plane. It is then further recommended that the AGND and DGND pins of a converter be tied together and that the analog ground plane and digital ground planes be connected at that same point as shown in Figure 8. This essentially creates the system "star" ground at the mixed-signal device.
All noisy digital currents flow through the digital power supply to the digital ground plane and back to the digital supply; they are isolated from the sensitive analog portion of the board. The system star ground occurs where the analog and digital ground planes are joined together at the mixed signal device. While this approach will generally work in a simple system with a single PCB and single ADC/DAC, it is not usually optimum for multicard mixed-signal systems. In systems having several ADCs or DACs on different PCBs (or on the same PCB, for that matter), the analog and digital ground planes become connected at several points, creating the possibility of ground loops and making a single-point "star" ground system impossible. For these reasons, this grounding approach is not recommended for multicard systems, and the approach previously discussed should be used for mixed signal ICs with low digital currents. "
This is exactly the scenario we encountered on D1V3 which had 4 PCM63s on the same board.
"Most ADC, DAC, and other mixed-signal device data sheets discuss grounding relative to a single PCB, usually the manufacturer's own evaluation board. This has been a source of confusion when trying to apply these principles to multicard or multi-ADC/DAC systems. The recommendation is usually to split the PCB ground plane into an analog plane and a digital plane. It is then further recommended that the AGND and DGND pins of a converter be tied together and that the analog ground plane and digital ground planes be connected at that same point as shown in Figure 8. This essentially creates the system "star" ground at the mixed-signal device.
All noisy digital currents flow through the digital power supply to the digital ground plane and back to the digital supply; they are isolated from the sensitive analog portion of the board. The system star ground occurs where the analog and digital ground planes are joined together at the mixed signal device. While this approach will generally work in a simple system with a single PCB and single ADC/DAC, it is not usually optimum for multicard mixed-signal systems. In systems having several ADCs or DACs on different PCBs (or on the same PCB, for that matter), the analog and digital ground planes become connected at several points, creating the possibility of ground loops and making a single-point "star" ground system impossible. For these reasons, this grounding approach is not recommended for multicard systems, and the approach previously discussed should be used for mixed signal ICs with low digital currents. "
This is exactly the scenario we encountered on D1V3 which had 4 PCM63s on the same board.
So it seems that Kester is suggesting the DAC(s) should placed over the analog ground plane and powered from the analog supply. I suppose that would be a fairly clean way to partition the board and powering the DAC VD regs from the same 15V supply used for the DAC VA regs doesn't seem like a real issue. The resistors on DATA, LE, CLK could the bridge the split in the ground planes in a similar manner to Figure 6.
spzzzzkt said:
From the board level, this is a clean arrangement yet the DAC chips may not like it. PCM63, based what BB engineers stated, actually loves this setup!
Still, the decoupling cap on the digital power rail is critical. It's hard to find an ideal cap for this purpose.
The bottom line is since we decided to go for separate analog and digital power, this was not an option.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.