[00940]

Ok, here's for comments a small (8cm/5cm) I/V stage. It's optimized for the PCM1794/8. It's a simplified/cheaper version of what I'm currently using.
 
I/V opamps are opa1632. A buffered trimmer set the input voltage at 0V. No caps in // because the opa1632 is apparently sensitive to capacitive loading.
Balanced to unbalanced and analog filtering is pretty much straight from the PCM1798 datasheet.
The power supply is a pair of simple shunt regulators. CCs transistors are bc337-327 or any 600mw to92, shunt transistors are bd139-140. Input should be +/- 10V, output +/-7V. +/-7V shouldn't make the opa1632 too hot.
 
The board has been drawn with wide tracks, large pads and easy tolerances. V+ and ground are on top, V- and signal tracks on bottom. It should be possible to do it at home. Current input at the right, output in the middle, power input at the left.
 

 

 

[00940]

Update.
 
The board was a bit too tight, so I added 2cm in length. The transistors for the ccs were a bit weak, so I changed them to bd139/140 too. And finally, going for a symmetrical layout eased the ground layout.
 

 

[cobaltmute]

Maybe you could rotate the CCS transistors 90 degrees to allow heatsinks over the board?

 

[00940]

Hmm... what I could do is to push them to the edge, so that they can more easily be heatsinked or bolted to the case ?
 
I'm not quite sure about it though, as they will only dissipate about 5V * 80ma (15ma per opa1632, 5ma per opa134, 20ma for the shunt, 10ma for variations) : 400mW. I'd think that the bd139 could reliably sustain it without heatsinking.  At first, I had more a 3V drop in mind so that would be rather 240mW, definitely fine without heatsinks and "ok" (but not great) for to92 with small heatsinks. Still, if I wanted to power this thing from +/-15V, heatsinking could be nice...
 
The transistors in the shunt parts would have like 30ma@7V, 210mW so they're quite ok.
 
 

 

[cobaltmute]

CCS heatsinking would just be a nice thing.  Don't think you need to go the full case route, but having the ability to add a heatsink there would be beneficial.
 
It would also allow you to run more current through the shunt transistors.  I can't remember where I heard it, but I did read somewhere that you want 2-3x the load current in the shunt for best performance.

 

[00940]

That seems a lot (but with audiophiles, nothing surprises me anymore 

). I'd think having 3 times the worst current variation in the shunt would be enough.
 
Btw, I made a small mistake in the maths above, it's 20ma per shunt I want. So, I've got (15ma+ 2* 5ma+20ma+5ma)*2= 100ma rather than 80 to dissipate.
 
Now, I've to find something to complete my panel at olimex. A tpa6120 based preamp maybe ?
 
 

 
edit: the mouting holes are weirdly placed, it's to be compatible with an USB dac board I've left.

 

[cobaltmute]

IIRC, the logic was that making the variable current draw a "small" portion of the draw on the regulator helped performance.  I've seen reference to doing to same thing to a three term regulator by loading the output with a resistor.
 
A TPA6120 preamp would be cool.  I've had thoughts about a LME49713 headamp.  Not as much current output, but better offset numbers.

 

[00940]

What I'm thinking about is a tpa6120 in the a vfb opamp's feedback loop, with a ds1802 in front of it. I had started to draw a board a long time ago, time to ping jcx  

 

[00940]

A reunion got cancelled in last minute tonight so I had the evening to consider this preamp thing. The schematic is a complete mess (jumpers in eagle aren't well implemented) so I'll describe it.
 
- Starting left you have the ds1802. Under it, a pair of +/-2.5V zener regulators. PS quality isn't very critical here.
- Middle top is a opa2134 (or pretty much any dual opamp). It's set up with the tpa6120 (middle bottom IC) in its feedback loop.
- The tpa6120 is soldered to both top and bottom groundplane, linked by vias.
- The two opamps are powered through jumpers running under the board.
- At the right, a typical dual lm317/337 regulator.
 
The dimension is the same as the I/V board (10/5cm). They could be stacked and the I/V powered from this one. The I/V will be set to give 2 or 3Vpp into the ds1802.
 
There are still some little things to fix (pads for output gnd for example

) but I think my Olimex's panel is slowly getting full.
 
Btw, starting groupbuy and the like isn't my thing but if anyone wants the eagle files for his own use (and at his own risk), just send me a pm with your email.
 

 

 

[cobaltmute]

Take a look on Pg 7 of the LT1033 datasheet.  There is way to make your LM317/337 supply dual tracking there using one pot and two resistors instead of two pots.
 
Everywhere I've read lately comments on the fact that you shouldn't use the caps after the zener/follower supply.  Obviously though this could be a choice to not install the part.
 
The jumpers are going right into the middle of the TPA?  Hate to have to install them.
 
Wouldn't it be better to have the control pins for the DS1802 on a header block?  I'd have them further away from the part to have the option at least.

 

[holland]

curious if it would be advantageous to keep the signal differential and have the TPA6120 perform the conversion after the pot.  I don't know if the DS1802 can be chained in that way, one for each channel.
 
one other curiosity, as I've been meaning to play with this myself, is similar to what audio-gd does.  that is use the pot for I/V, following a current amplifier.  DAC->current amplifier->I/V pot->buffer.

 

[00940]

Rushing to sleep (working tomorrow morning), so I'll be short.
 
Quote:

Take a look on Pg 7 of the LT1033 datasheet.  There is way to make your LM317/337 supply dual tracking there using one pot and two resistors instead of two pots.
 
Everywhere I've read lately comments on the fact that you shouldn't use the caps after the zener/follower supply.  Obviously though this could be a choice to not install the part.
 
The jumpers are going right into the middle of the TPA?  Hate to have to install them.
 
Wouldn't it be better to have the control pins for the DS1802 on a header block?  I'd have them further away from the part to have the option at least.

- Interesting trick, I might use it... The main interest in my view would be to save a pot

Dual tracking is nice for discrete designs with sensitive operation points, not really necessary for opamps imo.
 
- Any reason given for that ? Such stabilizers (one cannot really speak of regulators) are as old as npn transistors and zeners and I've always seen them with caps on the output.
 
- Ideally, they would be thick, twisted wires going flat under the board, soldered bottom side and carefully trimmed components side. I tend to design pcbs with "extra layers" in the air.
 
- Very good point and easy enough, I'll push them away. I'm a penny pincher, I solder everywhere I can rather than using connectors. It's not a good habit, I'm wasting time because of that.

 
Quote:

curious if it would be advantageous to keep the signal differential and have the TPA6120 perform the conversion after the pot.  I don't know if the DS1802 can be chained in that way, one for each channel.
 
one other curiosity, as I've been meaning to play with this myself, is similar to what audio-gd does.  that is use the pot for I/V, following a current amplifier.  DAC->current amplifier->I/V pot->buffer.

- This is the line of reasoning I've followed in my current DAC (well, the one that's broken until I get a new USB-I2S converter). However, there is one catch: you must usually drive the balanced receiver (the tpa6120 in this case) from a low impedance. Thus, the chain looks like this: DAC - I/V opamp - volume control - buffer - balanced receiver. An alternative is to use the pot in between the two phases (as a rheostat) but the ds1802 cannot be used that way. Another possibility is to use a balance receiver presenting a high input impedance, which is what I'm using based on a suggestion of PRR. See this thread (bottom of the page): http://gilmore2.chem.northwestern.edu/ubb/showpage.php?fnum=3&tid=7616&fpage=1 and http://www.head-fi.org/forum/thread/85561/post-pics-of-your-builds/7530 . The PCB in this thread are safer, more conventional options.
 
- The good old current conveyor... The practical problem with that approach is to present a signal without DC offset to the pot, or they get scratchy.
 
 
PS: I'm currently listening to the small dac I designed with Guzzler and my cheapie amp ( http://gilmore2.chem.northwestern.edu/ubb/showpage.php?fnum=3&tid=7773 )... Honnestly, considering how good this combo is, everything discussed here is well into the diminishing returns' realm. But it's fun to skin the cat in a dozen ways 

 

[cobaltmute]

Quote:

Rushing to sleep (working tomorrow morning), so I'll be short.
 

- Interesting trick, I might use it... The main interest in my view would be to save a pot

Dual tracking is nice for discrete designs with sensitive operation points, not really necessary for opamps imo.
 
- Any reason given for that ? Such stabilizers (one cannot really speak of regulators) are as old as npn transistors and zeners and I've always seen them with caps on the output.
 
- Ideally, they would be thick, twisted wires going flat under the board, soldered bottom side and carefully trimmed components side. I tend to design pcbs with "extra layers" in the air.
 
- Very good point and easy enough, I'll push them away. I'm a penny pincher, I solder everywhere I can rather than using connectors. It's not a good habit, I'm wasting time because of that.
 

 
I've lazy, so I like terminals and easy to install stuff.
 
Of course, I can't find many references now for no cap after the zener/emitter follower.  One I could find:
http://www.diyhifi.org/forums/viewtopic.php?p=42344#p42344
 
From all the postings I've seen at that same place, the comment is that it screws up the performance, but I don't have much to back that up right now.
 
 

 

[Avro_Arrow]

I built that and the tracking wans't so good.
Even with 1% resistors, the voltage was off
by a good half a volt at lower voltages (+-5 volts).
You will need to do some tight resistor matching
to get that to track well.
 
 
Quote:

Take a look on Pg 7 of the LT1033 datasheet.  There is way to make your LM317/337 supply dual tracking there using one pot and two resistors instead of two pots.
 
Everywhere I've read lately comments on the fact that you shouldn't use the caps after the zener/follower supply.  Obviously though this could be a choice to not install the part.
 
The jumpers are going right into the middle of the TPA?  Hate to have to install them.
 
Wouldn't it be better to have the control pins for the DS1802 on a header block?  I'd have them further away from the part to have the option at least.

 

 

[cobaltmute]

Quote:

I built that and the tracking wans't so good.
Even with 1% resistors, the voltage was off
by a good half a volt at lower voltages (+-5 volts).
You will need to do some tight resistor matching
to get that to track well.

 
0.5V on 10volts is only 5% error.  Not much worse that you can get out of using fixed regulators.  A 7805 can vary from 4.75-5.25V.