Hi,
I'm just starting out playing with circuits and I wanted to play with the OPA1656 and I took a circuit from the application notes found in the op amp datasheet. it has a Baxandall treble and bass controls with a little bit of gain and thought it would make a nice headphone amp that would run from two 9v batteries so I drew the circuit into a circuit SIM (Tina TI) and it appeared to me to work and I could see the top and bottom of the frequency range adjusting as I would expect with a fixed gain. I then copied it into KICAD and designed a small PCB and built it (see attached PNG) on power up I hear
d music and it sounded reasonable at first, I was flushed with success ! but there was a lot of hiss and when I adjusted the treble it got considerably worse yet the bass adjusted as I expected. I then looked at the output with a 33 ohm load on a scope and could see hundreds of mV of oscillations at around 1 MHz that went away if I had the treble full up (Fully Clockwise) but any adjustment made it spring into oscillation, I had an input signal of 1 KHz at around 1v RMS. I have sat and thought about it and can only think it is capacitive either to much or too little but no other thoughts really other than I have made a fundamental error and there is a lot missing, bypass caps could be closer to the chip but they are there and its batteries anyway. I am not sure if I'm in the right part of the forum but I'm hoping for some advice please.
Cheers
Andy
I'm just starting out playing with circuits and I wanted to play with the OPA1656 and I took a circuit from the application notes found in the op amp datasheet. it has a Baxandall treble and bass controls with a little bit of gain and thought it would make a nice headphone amp that would run from two 9v batteries so I drew the circuit into a circuit SIM (Tina TI) and it appeared to me to work and I could see the top and bottom of the frequency range adjusting as I would expect with a fixed gain. I then copied it into KICAD and designed a small PCB and built it (see attached PNG) on power up I hear
Cheers
Andy
I agree with all of rayma observations/suggestions.
Random thoughts:
These opamps are pretty fast, so bypass cap lead lengths could be an issue.
As a diagnostic experiment, try putting 470 ohm in series with output and see if this helps tames behavior a bit.
1 Vrms may be a bit aggressive with input gain of 11 when powered by +/- 9V.
When probing with a scope, be sure to use x10 probes to ensure low load capacitance--- helps to preclude oscillation.
Random thoughts:
These opamps are pretty fast, so bypass cap lead lengths could be an issue.
As a diagnostic experiment, try putting 470 ohm in series with output and see if this helps tames behavior a bit.
1 Vrms may be a bit aggressive with input gain of 11 when powered by +/- 9V.
When probing with a scope, be sure to use x10 probes to ensure low load capacitance--- helps to preclude oscillation.
The 18V total supply means that no more than about 6V rms output is possible.
An input of 1V rms x 11 = 11V rms output, so severe clipping is guaranteed.
An input of 1V rms x 11 = 11V rms output, so severe clipping is guaranteed.
Last edited:
You'll want beefier decoupling on the OPA1656. I'd add 100 nF in parallel with the 10 nF you have in there already. I'd also add 10 uF of bulk supply capacitance on the board.
The OPA1656 can deliver 100 mA of output current, so it should be able to drive a headphone load. It'll struggle a bit with 33 Ω load, though. I use two OPA1656 in parallel in the HP-2.
Note that while the OPA1656 is unity gain stable, it does have an internal compensation circuit that causes a phase wobble in the 1-10 MHz region. With a gain of 10 V/V (20 dB) that phase wobble will result in low-ish phase margin. I would add a little capacitance (10-22 pF) across R2. You can also try adding an R+C series combination in parallel with R2. Maybe something like 56 pF + 1.5 kΩ or 100 pF + 1.5 kΩ. You may have to figure out something similar for the second opamp.
Tom
The OPA1656 can deliver 100 mA of output current, so it should be able to drive a headphone load. It'll struggle a bit with 33 Ω load, though. I use two OPA1656 in parallel in the HP-2.
Note that while the OPA1656 is unity gain stable, it does have an internal compensation circuit that causes a phase wobble in the 1-10 MHz region. With a gain of 10 V/V (20 dB) that phase wobble will result in low-ish phase margin. I would add a little capacitance (10-22 pF) across R2. You can also try adding an R+C series combination in parallel with R2. Maybe something like 56 pF + 1.5 kΩ or 100 pF + 1.5 kΩ. You may have to figure out something similar for the second opamp.
Tom
I agree with Tom that the decoupling should be beefed up. For more precise recommendations a picture of the PCB would be helpful. Before bothering with compensation in too much detail, maybe it should be confirmed wether the gain really needs to be all that high.
Absolutely agree, and a 33 ohm reactive load like a headphone is probably worse, likely pulling the circuit out of feedback lock on the peaks...
This opamp is able to drive 600 ohm loads well, so don't load it much heavier than that. Many opamps only handle 2k loads, note.
Someone suggested a 470 ohm series resistor into the headphones would achieve this - but if you want to drive headphones you really should use a suitable output buffer in the feedback loop.
I doubt you need gain of 10 in front with modern sources and headphones. I had a gain of 50 headphone amp for critical recording but I had it turned above gain of 5 only a few minutes in 5 years of often daily recording sessions. (To check for hiss in room, mikes, preamps.) High gain helps oscillations.
When you have the oscillation tamed:
R8 R16 1.5k is very low for hi-fi. 10k is more nominal and I'm sure 22k or 100k can work with these opamps.... hey, they are FET input, you can use 1Meg. Then you can use a 0.1uFd or 0.01uFd input capacitor which has very much less leakage than a 10uFd electrolytic, or much less bulk (RFI antenna area) than 10uFd film.
Lovely clear drawing. (And I do not like colors, but they hardly distract in this one.)
When you have the oscillation tamed:
R8 R16 1.5k is very low for hi-fi. 10k is more nominal and I'm sure 22k or 100k can work with these opamps.... hey, they are FET input, you can use 1Meg. Then you can use a 0.1uFd or 0.01uFd input capacitor which has very much less leakage than a 10uFd electrolytic, or much less bulk (RFI antenna area) than 10uFd film.
Lovely clear drawing. (And I do not like colors, but they hardly distract in this one.)
Hi , in case of oscillation i would try to shunt each opamp output to inverting input with small capacitor like 10-47pf , just for test . Feedback path delay can cause oscillation ,longer wires to potentiometer and so . Try to add 10ohms at least in series with load .
The noted supply bypass cap are only 10nF, so they present impedance of about 17 ohms at 1MHz where Andymouse describes oscillation. Much heavier bypassing seems in order, especially since relatively heaving loading is the eventual goal.
I advised a series 470 ohm as a diagnostic experiment, but not a cure. Pics of PCB and setup may help.
I advised a series 470 ohm as a diagnostic experiment, but not a cure. Pics of PCB and setup may help.
Thanks to everyone who posted some advice ! not sure if I should thank everyone individually or not so this is to all of the above, Cheers !
Here is a picture of the PCB and I don't think there is enough capacitance and what I do have is placed incorrectly, I didn't think it was critical when using batteries. I am wondering if I can try the tests mentioned on one channel or should it be both ? one thing I noticed is the oscillations stopped if I lower the supply voltage to about 2v. I am going to lower the gain that was suggested and also place capacitors as a shunt also I
will add some series resistance as well, hopefully I will have some time to try the test in the next few days.
Thanks once more
Andy
Here is a picture of the PCB and I don't think there is enough capacitance and what I do have is placed incorrectly, I didn't think it was critical when using batteries. I am wondering if I can try the tests mentioned on one channel or should it be both ? one thing I noticed is the oscillations stopped if I lower the supply voltage to about 2v. I am going to lower the gain that was suggested and also place capacitors as a shunt also I
Thanks once more
Andy
As they say:
Place 0.1uf+10uf decoupling capacitors as close to the power pins as possible
And estimate place for capacitors pin1 to pin 2 , pin 6 to pin 7, you may not solder them ,if not needed .
If the oscillations are caused by improper decoupling, the issue is the trace inductance from the power supply to the ICs. That doesn't change if you change to battery power.
It wouldn't surprise me if the output impedance of a battery is somewhat gnarly, especially as the battery discharges. You do need some bulk capacitance on the board and 100 nF X7R by each IC.
Tom
It wouldn't surprise me if the output impedance of a battery is somewhat gnarly, especially as the battery discharges. You do need some bulk capacitance on the board and 100 nF X7R by each IC.
Tom
1/Decoupling.
2/ Small caps across feedback resistors.
3/ Keep feedback tracks short. 1mm = 1nH of inductance.
4/ Dont load output too much.
2/ Small caps across feedback resistors.
3/ Keep feedback tracks short. 1mm = 1nH of inductance.
4/ Dont load output too much.
1. An input RC filter is absolutely no luxury today. It should be standard like it was for decades. Stuff that can amplify HF/RF will amplify incoming HF/RF.
2. Decoupling is inadequate but that was already noted.
3. Ground reference resistors 470 kOhm are missing at the inputs. So connecting the mobile device will lead to the inputs caps charging and thus heavy plops.
4. The load is connected directly to the feedback loop. The circuit needs a buffer after each channel so an extra OPA1656.
5. The LEDs serve no purpose other than to drain you batteries faster. Design green and leave them away.
6. "KHz at around 1v RMS". It is kHz and V. If we already write stuff wrong the circuits probably will be of the same standard.
7. Why are R8 and R16 chosen so low? Many normal sources won't like 1.5 kOhm. It also causes C4/C11 to be an ugly electrolytic (despite what gurus say, film caps are better than electrolytic caps). C4/C11 should be quality 1 µF film caps and input impedance around 20 to 25 kOhm.
2. Decoupling is inadequate but that was already noted.
3. Ground reference resistors 470 kOhm are missing at the inputs. So connecting the mobile device will lead to the inputs caps charging and thus heavy plops.
4. The load is connected directly to the feedback loop. The circuit needs a buffer after each channel so an extra OPA1656.
5. The LEDs serve no purpose other than to drain you batteries faster. Design green and leave them away.
6. "KHz at around 1v RMS". It is kHz and V. If we already write stuff wrong the circuits probably will be of the same standard.
7. Why are R8 and R16 chosen so low? Many normal sources won't like 1.5 kOhm. It also causes C4/C11 to be an ugly electrolytic (despite what gurus say, film caps are better than electrolytic caps). C4/C11 should be quality 1 µF film caps and input impedance around 20 to 25 kOhm.
Last edited:
It is not precision IMHO but simple discipline. Sloppiness leads to sloppy circuits which saying proves itself over and over again. Also in this circuit if I may say so.
Last edited:
What is the best way to decouple such a fast oap ?
Datasheet says inductance is important, i.e, small leads gap and also close distance.
So in case of classic 5mm pitch mkp, one has to shorter the leg that fit on the oap rigth to the power pin and put the other leg in a not too far ground ? Or as far you have two close traces with a little gap of 2 mm, one power trace, the other the return to the ground, you still have a good inductance whatever the cap can be a little further from the oap power pin ?
Which conduct to the next question about layout decoupling : if one can not use rigth 0405 decoupling next the oap power pin to ground, what if a mkt 2.5 mm pitch leg, or ceramic trough hole VS a TH classic MKP ?
Is it more an inductance to the ground priority or a dielectric one about high frequency management ? What if you put a 22 uF lythic radial that has 1,5 mm pitch legs direct on the oap with short lead on the power pin and a longernother lead to the ground ? Ground beginns at the output of the cap, no ?
Datasheet says inductance is important, i.e, small leads gap and also close distance.
So in case of classic 5mm pitch mkp, one has to shorter the leg that fit on the oap rigth to the power pin and put the other leg in a not too far ground ? Or as far you have two close traces with a little gap of 2 mm, one power trace, the other the return to the ground, you still have a good inductance whatever the cap can be a little further from the oap power pin ?
Which conduct to the next question about layout decoupling : if one can not use rigth 0405 decoupling next the oap power pin to ground, what if a mkt 2.5 mm pitch leg, or ceramic trough hole VS a TH classic MKP ?
Is it more an inductance to the ground priority or a dielectric one about high frequency management ? What if you put a 22 uF lythic radial that has 1,5 mm pitch legs direct on the oap with short lead on the power pin and a longernother lead to the ground ? Ground beginns at the output of the cap, no ?