OT, but they are on a rolling production schedule. I'm set to receive mine near the 28th.
Would you mind sending me a PM when you've figured it out?
OK, there's an error there (mine), the 10R resistors are 1210. I have amended the BOM to show this.
The 10R resistors aren't in the signal path, they're in the PSU CRCRC filters, thick film are probably adequate there. I used a Panasonic part from Farnell, described as an anti-pulse or anti-surge resistor, rated 0.5W 5%. There are pads on the board for 2 in parallel so you could use 2 * 22R, The actual value is dependent on the wallwart you use (how much voltage you can afford to drop) and the dissipation (how hot they get). I've got 1 * 10R in each position, the amp should draw ~50mA quiescent which is 0.5V dropped across each, you can feel that they're above ambient but they're not hot.
R17, 18, 23 - 28, the voltage setting resistors round the regulators again don't need to be anything special, 1% is nice but I just selected some matching values from 5% thick film. I have a selection kit, 50 each of numerous values, only a few dollars from ebay. The values shown, 150R and 1k5, result in +/- 13.75V rails. You could parallel a 10u 1206 cap with the voltage-setting resistor to reduce ripple by piggy-backing it, but I found the hum inaudible once I cut the board. Similarly, the voltage dividers for the protection circuit and the switch-on delay can be thick film. This just leaves the gain-setting resistors in the amplifier proper and the integrator. I used Panasonic ERA series metal film 0.1% for this first build. These are rated low current noise, excellent non-linearity and are 27 pence here, not too bad.
All the ICs in the amp are rated +/-18V, so you can run it this hot if you feel the need and you have a suitable wallwart. I have some 16V ones. These provide ~22V after the diode drop. Lose 2V in the CRCRC filter, the LM317/337s have a dropout voltage of 2V, that would allow 100mA before the supply would begin to sag, a good compromise keeping the dissipation in the regulators from getting too high. You can see that you could get 200mA by reducing the 10Rs to 5R at the cost of some supply ripple, or go for an 18V wallwart at the cost of having the regulators run a bit hotter. You might want to bolt them to the enclosure and bring wandering leads to the board in that case.
Very helpful, thank you.
pity they took i2s off the header =(
Here's the resspin, so far. I've improved the legend legibility, modified the ground planes, added decoupling, ripple rejection caps on the regs. None of it's critical apart from the ground plane split, I'm just taking advantage of the available real estate.
OK, looking at the top channel in the diagram;
R14 is the feedback resistor from the output to the input, it works in conjunction with R13 from the inverting input to ground to set the gain. Despite the fact that there are 2 chip amps in the loop, this can be simply treated as an inverting opamp circuit with a gain of (1+(R14/R13), which works out to a gain of 2.
Looking at the circuit again, the 2 resistors R6 and R4 in series from the integrator output control the servo gain. The R4 and R6 are both 4k7 for a series resistance of about 10k. Since the feedback resistor R14 is 1k the servo authority is R14/(R4 + R6) or approximately 10%. That means that to correct an offset of 1V the servo opamp must put out 10V. This is about as much output swing as the servo opamp can manage with +/- 13V rails, which is about what you get with the resistance values shown (R17, 18 etc.) round the regulators. (13.75V actually)
The output of the servo opamp is a virtual ground for AC purposes, so R6+R4 are effectively in parallel with R13. This makes the gain (1+(R14/(R13 || (R6+R4)) or 1 + (1k/(1k || 10k).
1k || 10k = 1/(1k+1/10k) = 1/(11k/10k) = 10k/11k = 0.909k
...so the gain of the circuit becomes 1+ (1k/0.909k) = 2.1
In addition to this R12 can be paralleled with R13 by the addition of a jumper. This makes the gain 1 + (R14/(R13 || (R6+R4 || R12) = ~ 4.2. The gain, with the values shown, is approximately 2 without the jumper installed and approximately 4 with the jumper in place.
Obviously these values are not set in stone. Few modern sources exceed 2V output so the gain can be increased without the risk of clipping and there is room to increase the rail voltages too. If it was desired to approximately double the gain (to x4 and x8), I would increase R14 to 2k2, R4 & 6 to 10K and omit one of C28/C30. Again these measures would need to be taken with both channels.
Increasing R18 etc. to 1k8 will increase the rail voltage to 16.25V. This may be useful depending on the voltage of your (AC) wallwart, as the less voltage dropped across the regulators (down to 2V, which is the dropout voltage) the cooler they will run. A12V (AC) wallwart is good for 13.75V rails, a 15V wallwart for 16.25V rails, but check the voltage at the regulator input is 2V greater than the required output. The values in the CRCRC PSU filter (R33, 34, 35, 36) can be tweaked to trim these values.
Edit: Previously said 'R4 & 6 to 1K' corrected to 'R4 & 6 to 10K'
Lovely project, and just what I have been looking for since I have got some LME49990 and LME49600 samples from TI recently. Would not have done much differently myself.
I guess it's too late for a prototype board and too soon for a re-spun board? Ah, well, at least I'm not like a year late as with The Wire amp...
I'm doing up a Digi-Key based BOM for this build.
Is R20 supposed to be 47K or 1m?
It's 47k in the schematic and 1m in the BOM.
Also R25, R17, R27, R23, R26, R18, R28 and R24 are missing from the BOM.
Is this the correct relay?