my DAC design - pup1 DAC

[cobaltmute]

Over here, I showed off the fact that I made some nice looking coasters.

I'd rather not do that again. So I'd like to get some comments before I send out my boards again.

Considering that there is some many DIY DACS out there, why do another one? Well, quite simply, there was not one I could find that would meet my requirements:

1. USB only - for power and digital source
2. No capacitors at the output.

At a high level, what I have in my DAC which I'm calling the pup1 is:
USB -> PCM2707 -> I2S -> PCM1794/1798 -> IV stage

So in some detail:

The PCM2707 (U2) accepts the USB signal and converts it to I2S. The chip is running in self-power mode and is configured in a simliar manner to what is done in the gamma1 DAC. I had originally set it up as host powered, but the self-powered mode is a simpler configuration taking up less board space.

The PCM1794/98 (U4) is setup as to accept the I2S. The implementation is setup as per the PCM1798 datasheet. This is slightly simpler than the 1974 implementation (saves a bit of board space).

The IV stage is the same as the one used in the ezDAC. All kudos to Evan for his design. The only thing that I have done is change some resistor values to keep the ouput within available voltages (2V p-p). My choice of op-amps at the moment for this stage is either the LMH6643 or the THS4222.

Power on this board is a bit of work. USB power is filtered with a ferrite right off the cable. From their there is a 3.3v TPS (U1) regulator to provide the 3.3v sections to the PCM2707 and PCM1794 and a 4.75V TPS (U5) regulator to provide the power for the analog section of the PCM1794. In both cases the power is run through another ferrite post regulation.

To provide power to the I/V stage, a charge pump is used to create a negative rail. The charge pump I choose was the TPS60403 (U6) . This provides an unregulated -5V rail from the USB 5v source. Two regulators (U7, U8) are then used (one from the +5V and one from the -5V) to create +-2.5V. It is important to regulate the charge pumps output as the talk alot about the rail drooping under load. Post regulation there is again another ferrite as well as two rail capacitors to ensure that the output op-amp is feed properly.

Given all that to create clean power, I've added a couple of "test points" to the board. In the default configuration they can be used to check the voltages of the various stages. However, if desired they can also be used to:

a. run the output unregulated at +-5V
b. run the entire board on a power supply other than USB.


As a choice, I was originally going to go through hole like the gamma1. I then decided that certain components were too large, so I changed some of the larger components to 1206 SMD.

So my PCB looks like this:

Top:

Bottom (showing Top silk for reference):

I've got the schematic as a PDF file, but it is a little messy. It started as the schematic for a more complex design and then I took stuff off so it looks like a bit of a mess. I can attach is someone would like.

The two things that concern me right now:
1. The placement of the de-coupling capacitors (C19 and C17) around U4
2. The fact that I don't have any resistors on the I2S lines.

To more experienced eyes, am I in any trouble with what I am doing?

As a final note, I like to says thanks to:
1. Evan for his ezDAC I/V stage
2. AMB and everyone who commented on the gamma1 project. Learned a lot just from following the development thread at Headwize
3. MisterX for the charge pump idea. Took me a while to get to using it, but I got there eventually.

 

[MisterX]

Can't you already purchase a product based on a similar design?

HIPPOhifi - The Bloat

Quote:

Digital music and power is provided from your pc via USB to the Bloat. The digital signal is received by a BurrBrown PCM2707 USB Receiver IC. The digital signal continues into a BurrBrown DAC (Digital to Analogue Converter) PCM1798 (or PCM1794, see upgrade options link below). After the DAC the converted analogue signal goes into the first of the 2-stage amplifiers, also know as the I/V stage. It then enters the second stage amplifer and continues out via the 3.5mm stereo jack to your headphones.

 

[cobaltmute]

Quote:

Originally Posted by MisterX /img/forum/go_quote.gif Can't you already purchase a product based on a similar design? HIPPOhifi - The Bloat

I hadn't heard of that one until now.

But then again I wouldn't learn anything by just buying someone else's product. This is DIY after all. I'll reinvent the wheel if it means that I learn about how the wheel was built. This is about the process as much as the outcome.

And it's listed at $229US. That's about $287CDN right now. Excluding my screw up from my costs,its about $55CDN for the parts + $10CDN for each PCB board = $65CDN (or $54US) for the whole thing minus case.

 

[error401]

Don't know if you saw dsavitsk's HPDAC2 and 3, they're kinda similar to what you're trying to do. Might take a peek.

What was the issue with C17,19? They seem fine to me, though I'd move them to the side of U4 to shorten the loop a little. Might take a bit of shuffling for C17, but it shouldn't be too hard. I don't see any glaring issues, but I haven't seen the schematic so...

Why not go to mostly SMD components if you're going to go as far as you did?

 

[MisterX]

Quote:

Originally Posted by cobaltmute /img/forum/go_quote.gif But then again I wouldn't learn anything by just buying someone else's product.

What are you hoping to learn here?
This is like an exact repeat of your last design check thread.
What did you learn there?

Quote:

This is DIY after all.

So you won't mind posting the gerbers then?
(we can only do so much by looking at your images, especially since there is no schematic and it appears you are using something besides Eagle)

Quote:

Excluding my screw up from my costs,its about $55CDN for the parts + $10CDN for each PCB board = $65CDN (or $54US) for the whole thing minus case.

Do you get free electric to run the computer you "designed" this on?

 

[cobaltmute]

Quote:

Originally Posted by error401 /img/forum/go_quote.gif Don't know if you saw dsavitsk's HPDAC2 and 3, they're kinda similar to what you're trying to do. Might take a peek.

I've seen them. You are correct in that they are very similar. The HPDAC2, which is the closest to what I'm interested in still used the capacitors at the output.

Quote:

What was the issue with C17,19? They seem fine to me, though I'd move them to the side of U4 to shorten the loop a little. Might take a bit of shuffling for C17, but it shouldn't be too hard. I don't see any glaring issues, but I haven't seen the schematic so... Why not go to mostly SMD components if you're going to go as far as you did?

Having read a couple of articles about the de-coupling, I was curious how much difference it would make if they were not placed right beside the DAC.

As for full SMD, it really is a bit of a trade off. As I see it so far, if you have a ground plane on the bottom of the board, and you need to do a de-coupling capacitor to ground, you need to get to that plane somehow. It will either be through a via or a a through hole for a part. If you are limiting yourself to 1206 at the smallest size, for small value capacitors with small lead spacing, it would seem to be a toss-up on which is the better part to choose through-hole or SMD

 

[cobaltmute]

Quote:

Originally Posted by MisterX /img/forum/go_quote.gif What are you hoping to learn here? This is like an exact repeat of your last design check thread. What did you learn there?

That either a. my approach was wrong or b. it was the wrong place to solicit an opinion as I did not get a single response.

So I learned a lot in the reading all the datasheets, learning what everything did on the parts, the choices that needed to be made and how to layout the board. What unfortunately I didn't get was any feedback to know if what I thought was correct actually was. And as I said, that may have been due to the way I approached the community.

And it is not an exact repeat of last time - at least this time you said something.

Quote:

So you won't mind posting the gerbers then? (we can only do so much by looking at your images, especially since there is no schematic and it appears you are using something besides Eagle)

You are correct in that it is not Eagle. Using Eagle lasted about the length of trying one board.

Gerbers attached. I'll clean up the schematics and post them sometime later today. They'll still be the same multi-page design as the first time.

Quote:

Do you get free electric to run the computer you "designed" this on?

Work laptop, which usually makes it from day to day charges at work, so pretty much yes. In fact spending a number of days staying late to work on the 24" monitor as opposed to 12" laptop display.

And yes even I would put designed in quotes. Integrated building blocks would be more like it. I still have much to learn. But then again, everyone needs to start somewhere.


EDIT: I have removed the gerbers as they were incorrect.

 

[error401]

Quote:

Originally Posted by cobaltmute /img/forum/go_quote.gif Having read a couple of articles about the de-coupling, I was curious how much difference it would make if they were not placed right beside the DAC.

Probably not much, but may as well do the best you can if you've got room. You want the cap as close as possible to the power pin you're decoupling, but also the loop that the path back from the corresponding ground pin to the other lead of the cap makes should be as small as possible. Right now it's gotta go around some traces and stuff.

Quote:

As for full SMD, it really is a bit of a trade off. As I see it so far, if you have a ground plane on the bottom of the board, and you need to do a de-coupling capacitor to ground, you need to get to that plane somehow. It will either be through a via or a a through hole for a part. If you are limiting yourself to 1206 at the smallest size, for small value capacitors with small lead spacing, it would seem to be a toss-up on which is the better part to choose through-hole or SMD

Size wise you don't always save, but you can manage a more contiguous ground plane and it's less work to stuff the board. You're obviously not scared of SMD from the rest of the board design, so it seems odd to go with half through hole since it offers no real advantages other than no skill required soldering. Nothing wrong with it, it just seems weird.

 

[cobaltmute]

Quote:

Originally Posted by error401 /img/forum/go_quote.gif Size wise you don't always save, but you can manage a more contiguous ground plane and it's less work to stuff the board. You're obviously not scared of SMD from the rest of the board design, so it seems odd to go with half through hole since it offers no real advantages other than no skill required soldering. Nothing wrong with it, it just seems weird.

The partial SMD and through-hole is just an evolution thing. It started as the cub1 design at Headwize. As I decided to do the charge pump, I decided to remove the ASRC. Then I got to doing the PCB layout, I decided that to save space, the ferrites should be SMD. The 1uF and the 2.2uF as well became choices for SMD as they better ones seemed to only available int 0.2" lead spacing.

Now seeing the boards from the first run, it seems tighter in real-life than on the screen. After I build one, we'll see if I was foolish or not.

 

[cobaltmute]

As promised, schematics:

USB Section


DAC Section:


I/V Section:


And for the record, here is my Digikey BOM. All the parts are listed with their part numbers so you can see what goes where.

Edit: Just remembered that R19 on the BOM has the wrong value - that is the resistor for the LED and 10 Ohm is quite wrong there.

 

[digger945]

I don't know if this will help, but it might spark some ideas (several different products to choose from, my first idea was a vertical adapter I saw on ebay).
I don't know what size and/or pin count you have.

 

[cobaltmute]

Well, my boards came in yesterday and I assembled one board. Having a problem with the 3V reg and the charge pump, but I think I may know where those are.

Will post pictures later.

 

[millwood]

Quote:

Having read a couple of articles about the de-coupling, I was curious how much difference it would make if they were not placed right beside the DAC.

I used to have tremendous amount of trouble to power on a 8dip PIC device: one out of 10 times it would power on but others it just sit there.

it turned out it was a lack of decoupling. I had thought, incorrectly, that it draws minimal current so decoupling wasn't needed.

i couldn't be more wrong on that. apparently the chip needs consistent Vdd during power on or it hangs.

so I would say decouple your dac too.

 

[cobaltmute]

Quote:

Originally Posted by millwood /img/forum/go_quote.gif so I would say decouple your dac too.

I would suggest you review the schematic and the datasheets. Everything is decoupled as TI suggests.

 

[cobaltmute]

I found the issue on the 3.3v regulator - I had a bridge on the PCM2707 that was sitting behind the pin so it was not very visible. Once I got that straightened out, the board would power up and register as a USB device and then died with Windows reporting a device malfunction. No amount of reflowing the PCM2707 seemed to fix it so I've resigned myself to replacing it.

Now removing it was a challenge. I tried heating up all sides to get it loose, but that didn't seem to work. I tried wicking off all the solder and then doing the same, with no result. My final attempt has met with reasonable success and I only removed one pad off the board - I simple used a set of wire cutters to cut all the pins and then used my soldering iron to clean off the pads. There is one pad that is in question and it does worry me as it goes the the crystal.

I need to re-clean the board (as I have no compressed air right now to blow the crud off after I use my flux cleaner) but here is what it looks like:



My only other issue that I know about right now is that the charge pump (U6) is only putting out -1v as opposed to the -5V that it should be doing. Isolating it from the op-amp stage didn't make any difference for its output nor did re-flowing all the joints around it, so I'm a little puzzled by it right now.