[Clutz]

Quote:

Originally Posted by dsavitsk Isn't the rule that you aren't supposed to review your own stuff? Nuts to that. Compared to myusb->spdif converter running into a diyparadise tda1543 dac, even not burned in it is quite a step up. That setup was, in turn, quite a step up from guzzler's 2702 based dac (sorry gus.) So, for the fact that the parts are under $50, it sounds great (though there is a bit of a cheat -- I am using about $90 in jupiter coupling caps that I picked up for another project rather than the $4 in blackgates that have yet to arrive, so that helps a bit.) Thanks again for your help. I don't think this would have gotten done without it. -d

Congrats!

What about it do you like better than the diyparadise or guzzler dac? (Other than it is your own design, of course!

)

 

[dsavitsk]

Quote:

Originally Posted by Clutz Congrats! What about it do you like better than the diyparadise or guzzler dac?

I can essentially make up whatever I want here...

As compared to the diyparadise version, I think the level of refinement is much higher. In general I find that NOS DAC's sound is difficult to describe as they don't sound like anything, they just sound like music. This one just happens to get you closer and sound more delicate than the other one. I do think that some tinkering with the I/V is in order, however. I used all 1K resistors (PRP for I/V and Vishay dale for bias) but many feel that increasing these values and using carbon comps improves things, so that is the next experiment.

As to the 2702 based dac, I think it is a pretty nice sounding chip, but in the end the built in opamps limit what it can do. For a USB powered option, though, it's great. (It is worth noting that the usb specs say that usb will deliver 500mA, which would certainly be enough to power a pcm2707 and a tda1543. I don't believe it, though. Also, my experience is that reserve power is a good thing for sound quality.)

Quote:

(Other than it is your own design, of course!)

That's a big part of it too. I have gotten to the point where I spend tons more time thinking and designing than soldering. In fact, as I was putting the first one of these together I realised that I had not soldered anything in weeks. This, I think, is a good thing as I don't find stuffing circuit boards to be particularly fun. Instead, I like figuring out how stuff works, and while I am still very much a beginner, starting, designing, and completing a project is really rewarding.

As a future project, the TDA1541a looks like it would work in this circuit without too many changes (i think it just takes i2s as well). It is a much larger chip though, so space will be tight moving stuff around, but I am thinking it might be worth doing. The only real issue is that tda1541's cost a ton these days (double crowns are over $100 on ebay.)

Oh, another issue came up which is that there is a second (small) bug in my board which is that RCA jacks case mounted next to the usb jack extend over the board and bump the output cap if you use a blackgate (if you use a film cap it is off the board anyway, so no big deal -- but the leads must be isolated.) This means that a case really needs to be an extra inch wide, or tall, or you have to use short rca jacks. Not a huge deal, but worth tinkering with.

 

[Francis_Vaughan]

I was thinking abou the longer term value of this board as a basis for further experimentation. Trouble is, as you observe, it is a little cramped. But the basics are there.

A TDA1541A could be cute, although I have a general worry about the whole NOS deal. The reasons usually trotted out for why NOS is better are mostly misunderstandings of the theory. Not that there may not be some good reasons why NOS may work well, but I tend to feel that they are second order effects.

Things that I feel would help another design greatly would be stuff like reclocking the I2S signals. There is of course a huge following of the 1541 that is playing with clocking the DEM pins too. There are actually good reasons why this could be beneficial. (And nothing to do with NOS.) So a high stability clock would be good. Even replacing the 12 MHz crystal with a feed from a high quality clock (like a Kwak) would be a very interesting experiment. The 270x series will take such a feed.

Different I/V converters too. I have a hankering to try Malcom Hawksford's current steering I/V. Of course personally I would use a PCM1704 DAC. You could even run these NOS

 

[dsavitsk]

Quote:

Originally Posted by Francis_Vaughan I was thinking abou the longer term value of this board as a basis for further experimentation. Trouble is, as you observe, it is a little cramped. But the basics are there.

AsI look at it more closely, the TDA1541A (which I think is DIP-28) probably won't fit without some drastic alterations. Though, if the transformer were to be moved off board, there would be tons of space.

Quote:

Originally Posted by Francis_Vaughan A TDA1541A could be cute, although I have a general worry about the whole NOS deal. The reasons usually trotted out for why NOS is better are mostly misunderstandings of the theory. Not that there may not be some good reasons why NOS may work well, but I tend to feel that they are second order effects.

I'm not sure how much I care about (or understand for that matter) the ups and downs of the various theories. I feel like I have a handle on what the bits are doing in a NOS design, but not otherwise, so I feel like I can put one of these together. I'm also not convinced that a digital filter is inherently better than an analog one. To my ear, the NOS designs I've heard sound better than similarly priced oversampling designs, but so long as I understand it, I'm willing to try anything.

Quote:

Originally Posted by Francis_Vaughan Things that I feel would help another design greatly would be stuff like reclocking the I2S signals. There is of course a huge following of the 1541 that is playing with clocking the DEM pins too. There are actually good reasons why this could be beneficial. (And nothing to do with NOS.) So a high stability clock would be good.

This would probably require a daughter board.

Quote:

Originally Posted by Francis_Vaughan Even replacing the 12 MHz crystal with a feed from a high quality clock (like a Kwak) would be a very interesting experiment. The 270x series will take such a feed.

The datasheet is pretty specific that, at least in the case of the 2707, this is a bad idea. They say you cannot turn off the internal clock, and thus an external clock will just aggregate with it rather than replacing it which will be worse. As I type this, I am struck that I don't really know if I used the wrong type of crystal here? I think the data sheet says to use a resonator. So much to learn.

Quote:

Originally Posted by Francis_Vaughan Different I/V converters too. I have a hankering to try Malcom Hawksford's current steering I/V.

I'm not familiar with this. Any links?

Quote:

Originally Posted by Francis_Vaughan Of course personally I would use a PCM1704 DAC. You could even run these NOS

I think a pair of these would filt better than the TDA1541A though other associated circuitry could start to crowd things.

 

[steinchen]

Quote:

Originally Posted by dsavitsk Quote: Originally Posted by Francis_Vaughan Different I/V converters too. I have a hankering to try Malcom Hawksford's current steering I/V. I'm not familiar with this. Any links?

Malcom Hawksford's current steering I/V (pdf)

 

[dsavitsk]

A couple of quick pics. I ordered too few ferrites the first time around, so I am waiting on DigiKey to ship them so I can assemble a second one of these to start I/V tinkering.

Stuff worth mentioning here: the I/V resistors are 1K PRPs, the bias is a 1K Dale. The resistors from out to ground are 47K Dralorics, and the output caps are BG N series 4.7uF. Also, I put insulating pads under the USB jack and the crystal since they are metal and there is no solder mask to prevent shorts.

 

[guzzler]

lovely stuff, those cases you do are excellent.

You can ground the X-tal body as well, no worries there.

 

[dsavitsk]

Quote:

Originally Posted by guzzler lovely stuff, those cases you do are excellent.

Thanks. I've got a block of padauk that the next one of these is going into. Here it is sanded, but not oiled (sittng on a cherry table for comparison.)

It is a little hard to see thanks in part to a terrible camera, but the wood is bright orange when you cut it (I turned a friend's shop totally orange the other day) and deep red once oiled -- very pretty. Here's a link with better pics: http://www.exotic-wood.com/african_padauk.htm.
I think the plan is to fit the DAC and a small head amp (based on an AD8397) for a sort of portable thing -- the block is slightly larger than an RA-1, an should have just enough room. Whether I do cmoy style for the amp, or wait for a pint is up in the air. I am both impatient, but also way to busy to be fooling with this stuff for the next few weeks. Plus I have about 8 other projects at various stages of disassembly that need to get done.

Quote:

Originally Posted by guzzler You can ground the X-tal body as well, no worries there.

The problem is that there is a power trace and a connection to ground under it that I don't want to bridge.

 

[dsavitsk]

I built up another one of these with a few changes. I used AB carbon comp resistors for I/V and bias (2.8K/1.4K), polyprop film caps instead of polyester in the PS, and an AN8008 voltage reg for the TDA1543. Not sure which change is responsible, but this one sounds cleaner and more dynamic. The caps have the same lead spacing but are fatter and thus had to be mounted "Pinky style" (i.e., on the bottom of the board) so I need to move some stuff around in the layout to fit them up on top.

 

[dsavitsk]

Board files released, parts list added, new pictures, etc.

http://www.ecp.cc/NOS-USB-DAC.html

 

[steinchen]

I'm interested to give a non-OS DAC a listen but at $51 for one miniboard and $32 for shipping to Germany this is a no go for me. The little 3.8"x2.5" $51 board doesn't even have a silkscreen or legend.

Expensive boardhouses in Germany take about 50 EUR (~$60) for a single prototype eurocard-sized (160x100mm) board with silkscreen and position plot.

ExpressPCB seems way too expensive for me, or did I miss something ?

 

[skyskraper]

the software is free, you gotta end up paying for it somehow!

 

[steinchen]

just found out that you get 3 boards for $51, that's reasonable, but $32 shipping is a joke. If someone in the States is going to order a set of miniboards I'm interested in buying one.

 

[halman]

Steinchen >>

try:

http://www.olimex.com/pcb/index.html

 

[cetoole]

Quote:

Originally Posted by halman Steinchen >> try: http://www.olimex.com/pcb/index.html

The problem is that the board was designed in ExpressPCB, a proprietary piece of software that doesnt allow users to export to gerbers or any other format, practically forcing one to purchase boards from them.