Got my mouser parts in today and oh man
Didn't realize that these opamps are SOP sized.
Hi - which op-amps, the AD8656? Mouser has both the larger SOIC-8 (the ARZ ending on the part number) and the smaller MSOP (the ARMZ ending).
I just sent you a PM about something.
For fun, I've looked at the effect of the two rail splitter versions in simulation. I used a generic UniversalOpamp model with a 200 mA Ilimit, and a 9 V block was modeled crudely with 45 ohms of source impedance. Both channels are in parallel, the input is a 10 Hz, 1600 mVpp square wave.
The only difference between the two versions is the point at which C4 connects.
Now look at this:
The correct "single-supply" version (right) exhibits some highpass-related sag, but generally does a decent job. (My square wave unfortunately does not start at zero volts, hence the initial offset that diminishes over time.)
The stock cMoy version (left) is a mess, however, with the opamp being decidedly unhappy. You can tell by V(Vminus) how wobbly the virtual ground is. It gets better if you make the input and/or output coupling cap much smaller, but still not perfect - the peaks remain oddly rounded.
(Yes, "ground" is virtual ground here, and Vminus is my "real" ground that input and output are referred to. It looked nicer that way. As it's been explained before, "ground" is an arbitrary reference. If you were to use Vminus as your reference as it's often done, virtual ground would be at -V(Vminus).)
A split supply version with DC-coupled output is not shown, but it's as well-behaved as you'd expect.
The amplifier with the stock splitter immediately starts behaving properly when cap-coupling the feedback network to Vminus instead of going to virtual ground.
Attachments
Last edited:
A-ha! (Not the band.) That does look much better, in fact. Resulting bass rolloff cannot compete with the right-hand circuit though, it seems to be equivalent to having a 150-ish µF output coupling cap (or 75-ish µF per channel). Changing the voltage divider makes things even worse, while large buffer caps improve matters.
A pretty clever circuit, that cMoy. It can operate both split and single supply with decent results, and gets by with a minimum of electrolytics. I would still prefer the RHS circuit in a single-supply application though - it gives about the same performance with 100 µF / 25 V + 100 µF / 16 V (buffer) + 2x 470 µF / 16 V (coupling) as a cMoy with 2x 1000 µF / 16 V (buffer). (Both would cost about the same, but fitting the smaller caps ought to be easier when space is tight.)
A pretty clever circuit, that cMoy. It can operate both split and single supply with decent results, and gets by with a minimum of electrolytics. I would still prefer the RHS circuit in a single-supply application though - it gives about the same performance with 100 µF / 25 V + 100 µF / 16 V (buffer) + 2x 470 µF / 16 V (coupling) as a cMoy with 2x 1000 µF / 16 V (buffer). (Both would cost about the same, but fitting the smaller caps ought to be easier when space is tight.)
Interesting! Yeah it looks like you guys are right. With C4 gone the result is probably more of a AC/noise bypassed voltage reference.
With both C4 & C5 together the result really is a virtual ground and bypasses the return signal to both rails. In a perfect world, with the leakage currents of C4 & C5 exactly equal, R3 & R4 wouldn't be needed at all of course. The VG would be just C4 and C5. But as mentioned previously those then become a voltage divider for rail noise. May be useful to have both. The C4 + C5 VG to return the output signal to, but a separate divider with just the bottom resistor bypassed to act as a low(er) noise midpoint voltage reference for the non-inverting input, in the case of the inverting amp, if trading off the output coupling caps for the C4 & C5 VG.
With both C4 & C5 together the result really is a virtual ground and bypasses the return signal to both rails. In a perfect world, with the leakage currents of C4 & C5 exactly equal, R3 & R4 wouldn't be needed at all of course. The VG would be just C4 and C5. But as mentioned previously those then become a voltage divider for rail noise. May be useful to have both. The C4 + C5 VG to return the output signal to, but a separate divider with just the bottom resistor bypassed to act as a low(er) noise midpoint voltage reference for the non-inverting input, in the case of the inverting amp, if trading off the output coupling caps for the C4 & C5 VG.
Last edited:
The other great advantage of a single supply configuration is that there are less risks to interface it with other pieces of equipment, especially when earthed. With a virtual ground, one must always keep track of "real ground", in order to avoid a short.
You can probably even reduce the C5 cap in the single supply version by using higher values resistors for the voltage divider and a jfet opamp.
Inverting CMOY for 5V USB with attenuator - works!
Well the boards for the "5V USB inverting CMOY" amp in the posts above arrived and actually works, minus the typical V1.0 layout screw-ups. I had the polarity of the USB connector power pins reversed, the polarity of the relay coil reversed, and the rotation direction of the pot reversed - goes high to low. Luckily I noticed just before sending the board out that I had a "A" USB socket in there instead of "B" socket and fixed that.
BUT... it works! A single chip (an AD8656 rated for up to 220mA out on each channel) powered off of the 5V USB. No computer noise at all either in the headphones or looking at the output on the scope, so the common mode choke and in-line ferrite bead on the USB power input seems to be doing the job. I'm measuring 4.56Vdc out to the board under load.
The LT Spice simulation predicted a 1.6Vpeak swing with the AD8656 chip into 32 ohms. In real life the scope shows a 1.5Vpeak swing into a 32R load, but part of that is probably due to the missing 0.5Vdc on the power rail (4.5Vdc vs. 5.0Vdc in the sim). That is 1.5V peak max on the positive half, the negative does go down to -1.6Vpeak before clipping.
Fits just fine in the tiny Hammond 1455C801 case. One nice surprise here. I was planning on the bottom slot but I wound up mounting the relay underneath to take care of my coil polarity layout screw-up. I found that the board will also fit just fine in the second slot up from the bottom. The relay will clear the bottom just fine and all the parts still clear the top.
Sounds great on a couple of headphones that don't need more than the 1.5Vpeak swing (1.06Vrms). I'm not going to publish the gerbers yet since there are layout screw-ups. I'll likely have another test board made first.
The inverting attenuator circuit works exactly as expected! Takes the signal from zero through unity gain and on up to 4x voltage gain (the 50K pot in parallel with the 200K resistor and the 10K inverting input resistor). Pretty slick.
Since sending the board out to fab I've discovered they actually make a 1500uF 6.3V solid organic aluminum electro, in stock at mouser, for only slightly more than the 470uF 6.3V's and slightly larger at 10mm diameter vs. the 8mm. I might go with those in the next board turn. The frequency de-rating (heating) for these 470uFs is huge, 0.05x at 120Hz. Luckily the ripple current rating is high to start with, 5700mA.
Some photos...
* The first two are the test setup. The USB power is coming from a laptop so should be "dirty power" (noise) coming in. The USB "B" socket fit just fine, the wires are to deal with my layout polarity reversal screwup.
* Next is the DMM readings with some music an sensitive headphones. The DC output offset from the AD8656 runs around 500uV (=0.5mV) with no signal in.
* The next 3 show the size of the board and the Hammond case. The second shows how that relay rather suprisingly fits just fine under the board.
* The final 4 are scope shots with a 1Khz sine just before clipping at 1.45Vpeak, then 1 Khz at clipping, then a 10KHz and a 100Hz. The scope vertical is at 500mV/div in all of these.
Well the boards for the "5V USB inverting CMOY" amp in the posts above arrived and actually works, minus the typical V1.0 layout screw-ups.
I had the polarity of the USB connector power pins reversed, the polarity of the relay coil reversed, and the rotation direction of the pot reversed - goes high to low. Luckily I noticed just before sending the board out that I had a "A" USB socket in there instead of "B" socket and fixed that.BUT... it works! A single chip (an AD8656 rated for up to 220mA out on each channel) powered off of the 5V USB. No computer noise at all either in the headphones or looking at the output on the scope, so the common mode choke and in-line ferrite bead on the USB power input seems to be doing the job. I'm measuring 4.56Vdc out to the board under load.
The LT Spice simulation predicted a 1.6Vpeak swing with the AD8656 chip into 32 ohms. In real life the scope shows a 1.5Vpeak swing into a 32R load, but part of that is probably due to the missing 0.5Vdc on the power rail (4.5Vdc vs. 5.0Vdc in the sim). That is 1.5V peak max on the positive half, the negative does go down to -1.6Vpeak before clipping.
Fits just fine in the tiny Hammond 1455C801 case. One nice surprise here. I was planning on the bottom slot but I wound up mounting the relay underneath to take care of my coil polarity layout screw-up. I found that the board will also fit just fine in the second slot up from the bottom. The relay will clear the bottom just fine and all the parts still clear the top.
Sounds great on a couple of headphones that don't need more than the 1.5Vpeak swing (1.06Vrms). I'm not going to publish the gerbers yet since there are layout screw-ups. I'll likely have another test board made first.
The inverting attenuator circuit works exactly as expected! Takes the signal from zero through unity gain and on up to 4x voltage gain (the 50K pot in parallel with the 200K resistor and the 10K inverting input resistor). Pretty slick.
Since sending the board out to fab I've discovered they actually make a 1500uF 6.3V solid organic aluminum electro, in stock at mouser, for only slightly more than the 470uF 6.3V's and slightly larger at 10mm diameter vs. the 8mm. I might go with those in the next board turn. The frequency de-rating (heating) for these 470uFs is huge, 0.05x at 120Hz. Luckily the ripple current rating is high to start with, 5700mA.
Some photos...
* The first two are the test setup. The USB power is coming from a laptop so should be "dirty power" (noise) coming in. The USB "B" socket fit just fine, the wires are to deal with my layout polarity reversal screwup.
* Next is the DMM readings with some music an sensitive headphones. The DC output offset from the AD8656 runs around 500uV (=0.5mV) with no signal in.
* The next 3 show the size of the board and the Hammond case. The second shows how that relay rather suprisingly fits just fine under the board.
* The final 4 are scope shots with a 1Khz sine just before clipping at 1.45Vpeak, then 1 Khz at clipping, then a 10KHz and a 100Hz. The scope vertical is at 500mV/div in all of these.
Attachments
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.