So here I go again with the ever-more-difficult-and-more-expensive projects. This time I wanted to try out a DAC that doesn't get much exposure around here since you really need a microcontroller to set it up properly - the AD1955. On paper, it's very competitive with PCM1794 and I'd like to give an active I/V a try - everything I have is either voltage out or passive I/V. In my continuing quest for ever-smaller boards, I've managed a layout in only 1.5"x2.5", which wasn't easy. That will make the build fun though, an average of 26 components per square inch in that space on both sides of the board.

Before I go ahead and send the design off for fabrication, I'm asking you folks to proofread my schematic and make any comments. It's pretty much lifted straight from the AD1955 datasheet with some simple changes. Anyone interested?

Here's the schematic. Please check out the PDF since it's too large to be readable at forum-size.

I'm interested. Will you be making more than one board? I'd also want to use a THS4131 for I/V as recommended by Twisted Pair.

What are you going to use to get the I2S? Are you just going to leave that open or are you going to put some sort of SPDIF or USB input? If you leave it open, I might be interested in getting a board too.

.

Quote:



That's actually something I was wondering about. A smattering of posts around the web seems to indicate that people are using that datasheet schematic with different opamps like OPA627 or OP275 without ill effects, but I'm having a hard time finding much detail and the datasheet is rather lacking as well. My layout uses dual opamps; there's no way it would fit with singles though, so I can't use AD797. I was planning on using LME49720 or LM4562 since I've liked them both a lot, they aren't too expensive and are fairly easy to use. Can you give any advice on tuning the component values?

And no, I'm not sure double-sided will be sufficient. The ground planes are a bit of a mess - but I was able to get everything routed and the main analog and feedback paths have no vias and good separation. I need to take a closer look at the major current loops and see if there's anything really bad, if so I may need to go to 4 layer. I might try it anyway with a 2-layer board since it's pretty cheap to try it out and if it fails I can spend the money on a 4-layer redo. 4-layer prototypes aren't cheap. When the layout is finalized, I'll post more details on what I end up with. I might relax the size requirement a bit to clean up the ground.

Quote:



I'm laying out a USB board right now, which I'm planning to fab at the same time, and I've got my Opus SPDIF module (though I'll probably do my own just for kicks, and to use the input mux). It seems like I'm just replicating the Opus line, but it's more for fun and to try some obscure stuff than for the pursuit of the right sound. I've also got PGA2320 volume control and a bipolar power supply board in the mail to go along with these.

Quote:



I'll probably make a couple, but not more than that to start. I'm short on funds at the moment and since I've got extra time, I'm starting a lot of projects. If you like the final layout, we can probably arrange something before I go to fab if you really want in, but keep in mind you'll need an AVR programmer as well to flash the microcontroller. Unfortunately THS4131 won't work with this design, which uses vanilla opamps. Most dual opamps should work though.

Go crazy, built SRPP output! Lampizator CD player upgrade tubed output stage
It's a CD player modification actually but you can follow the ideas - passive I/V conversion on a resistor and a tube buffer after it.

I've been looking at the same part too, for a certain iPod project (heh). I've got the AD1955 in hand, but not started the board yet.

I believe AD uses the same schematic for their eval board, which is said to sound fantastic. And I've seen the same circuit in a analog design article somewhere. Unfortunately I didn't bookmark it, but I think it was a Walt Jung article and I was searching for active I/V converters and low pass filters.

Those 100 Ohm resistors in the op-amp feedback loops are recommended for the AD797 from it's datasheet. I don't know if you'd need them for an alternative op-amp.

Overall, the circuit is also a fourth-order low pass filter, and the only thing I can remember from the article is that C29,C35,C39 and C40 need to be tweaked to obtain the desired response. But since AD ended up with those values, they should be OK.

I take it you're using 0.1 uF decoupling per op-amp per rail? The reference circuit also adds a 4.7 uF tant in parallel with each one. I've been looking for alternatives, but they all have much lower ESR than tantalums, which may be a problem.

How did you derive your values for R5 and R6?

Quote:



That was my thinking as well: it's in a publication from a major manufacturer, the values should be well-chosen. I need to look into it a bit more though. I wouldn't be surprised if Walt Jung designed this circuit himself, since he has worked for AD for many years. I'll see if I can source the article you mention, as it would be good reading.

I am using 0.1uF per op-amp per rail, plus a bank of 4x100uF polymer electrolytics shared between them. I believe this should be sufficient, but since I can probably fit them in, I might add those tants as well. There is also 0.1uF and 47uF after an LP2985 for each AD1955 supply.

Well, "I believe" may have been an understatement... I came across a pdf that appears to contain the schematic and BoM for the AD eval board.

The bad part is the tantalums they listed were Panasonic EF series parts that have been discontinued, I can't find data on their ESR. As an alternative I was looking at Panasonic FK series in B size (1.35 Ohm impedence and more than 4.7 uF) or their SP-CAP and possibly a resistor in series with it. (Both for an all surface mount board design) That said, I don't know for sure if the BoM is real, or if it was actually used.

LP2985 is something I'll have to look at. I was considering REG101s for the 5V supplies and 2.8V non-inverting references. I picked up some of the 2.8V ones, but he 5V were out of stock.

Oh, and just to avoid any misunderstanding, I saw the article on some website to do with analog design, not Analog Devices' website.

Quote:



Me too: http://home.tiscali.cz:8080/saudio/AD1955EB.pdf and it seems legit. I have no reason to believe it's not accurate, at least to the point of being usable.

DigiKey part numbers are listed and they seem to still have the EF series datasheet on their site if you want it: http://www.panasonic.com/industrial/...ABF0000CE5.pdf . ESR doesn't seem to be specified however, but I wouldn't be that concerned about it. They are decoupling capacitors and should be 'the lower the better'. Thus, I think my use of UCC PSA series (25mR ESR) should be fine.

Quote:



REG101 seem to be getting hard to source so I've stopped using them. LP2985 specs almost as well and is much easier to source. It's also about half as much per device. If you're planning on using an active device for the reference voltage (I'm using the resistive divider with decoupling), you might look at a voltage reference instead of a regulator. You should be able to find one with much lower noise and better temperature stability than either REG101 or LP2985. Personally I think the divider is sufficient since the bias voltage only drives opamp non-inverting inputs, it should be a very static, very small load. An active device may actually be worse since it's probably more temperature dependent.

And you previous asked about R5 and R6. These are calculated based on a formula in the datasheet: R = (V+ - VBIAS) / (IBIAS + (VBIAS / Riv)). In my case, this works out to R = (15 - 2.80) / (3.24mA + (2.80 / 2000)), R=~2.63K. As far as I can tell, Riv is the feedback resistor and Ibias is given in the datasheet as -3.24mA.

It's a nice circuit. Very well designed.

The USB board is pretty much done. Check out the PDF. Just a basic PCM2707 implementation, nothing special at all, but the board size and layout matches the plan for the DAC so far...

Right, found the article that has a very brief mention of the circuit:

Using Op Amps with Data Converters - Part 5

last one on the page, and the values are different.

I don't think DigiKey has any stock left of the EF capacitors in the schematic, hence my searching for an alternative. And unforutnately, this is one case where lower isn't necessarily better...

If you look at the impedence frequency plot of a cap, it's v-shaped. The left hand side decreases due to capacitance, but the rising arm on the right is due to ESL. The minimum point is determined (among other things) by the value of the cap, a larger valued cap will have a minimum at a lower frequency.

Now when you put a large valued cap in parallel with a small one, eg the classic 4.7 uF and 0.1 uF, the ESL of the larger cap and the capacitance of the smaller one form a resonant tank circuit, leading to a unexpectedly high impedence peak between the two valleys.

The way to get around this is to add some series resistance into the circuit, this has the unwanted affect of raising (rounding off) the valley of the cap you put the resistor in series with, but the plus is with the right value, you can wipe out the unwanted peak. (Or in RF terms, you add a Q-spoiling resistor.) And tantalum caps just happen to have about the right amount of resistance to do that...

I actually built this board up almost 3 months ago, and I haven't posted here because I wasn't able to get it to work! Both 5V power supplies were being dragged down to ~1V and were in current-limit. I couldn't find any obvious bridges and reflowing didn't help.

The major error I found today after a few hours of probing was that I had ordered 2.21 and 2.80 OHM resistors instead of 2.21K and 2.80K for the VBIAS divider! 5 ohms to ground on AVDD! I didn't have any proper values on hand, but I was able to make some substitutes that get VBIAS at about 2.9V, which is pretty close. After making that substitution, I still had the same basic problem, and finally decided to replace the AD1955 itself, figuring it must have died at some point. It was a tough decision since they're almost $15 each and replacing a 28-SSOP isn't exactly trivial.

Long story short: it works!. And to my ear, it sounds pretty fantastic. Definitely more crisp, with more bass and cleaner mids than the AlienDAC I had hooked up just prior to trialing this one. My worries about the active I/V were unfounded, as it sounds great and has minimal DC offset (~3mV). Definitely this is going to be my new 'benchmark' DAC, burned terminal blocks, lifted pads and all. I think I like the sound better than my WM8740-based Opus, but that could easily be placebo . It's definitely on the same level though. I haven't done any microcontroller code though, so it's just running with default values right now. I wonder if it's even worth bothering, it sounds that good!

Now my entire signal chain except for the amplifier is self-designed and self-built. I've got a MiniPow for power, MiniUSB -> MiniDAC -> MiniVol for signal, and the CKKIII as my amp right now. Sublime.

I don't think I'll be producing more of these, the build is pretty difficult, it's far too dense to be practical. Here's the specifics for anyone interested though:

MiniDAC schematic and layout (pdf) (yes, the bottom slikscreen is a disaster)
MiniUSB schematic and layout (pdf)

If anyone wants more specifics (e.g. BOM, gerbers, whatever) feel free to ask and ye shall receive.

*goes to listen to some Pink Floyd*

We should do a group PCB buy for this, not many AD1955 DACs out there.

I was thinking the same thing.

Can you do an RMAA for comparative purposes?