Post #4 shows a noninverting unity gain buffer using the OPA134 / OPA2134 that has never exhibited any instability.
What's going on in that circuit? It looks like the power supply is modulated by the input. Is that how the designer prevents the 2134 from latching up? Seems drastic.
I was prototyping a speaker compensation circuit. I was using the 2134 in the non-inverting configuration. It was stable until I applied a signal. Then it latched up. Powered it down and back up, rinse lather repeat. Then I configured it as an inverter and it worked great. I've been using that circuit in my preamp for a couple of years without a hiccup.
I was prototyping a speaker compensation circuit. I was using the 2134 in the non-inverting configuration. It was stable until I applied a signal. Then it latched up. Powered it down and back up, rinse lather repeat. Then I configured it as an inverter and it worked great. I've been using that circuit in my preamp for a couple of years without a hiccup.
This is the local bypass capacitor that I always solder directly to the pins. https://www.mouser.com/ProductDetail/594-K104M15X7RF53L2/
I'm wondering if there is a better choice. I have less than 20 left so I'm going to have to order more soon.
I'm wondering if there is a better choice. I have less than 20 left so I'm going to have to order more soon.
Hah! That's almost exactly the one I've been using (A104M15X7RF5TAA) except that mine might be just a tad smaller---see my picture in post #58. Your choice is a little bit less expensive in large (100) quantities, though.......methinks either one will be an excellent performer in this PS bypass application,
OK that's what I figured.
Is there any virtue in ordering 100V versions? Higher voltage will (theoretically) have less capacitance change with voltage.
Is there any virtue in ordering 100V versions? Higher voltage will (theoretically) have less capacitance change with voltage.
That may be true, but the higher voltage cap is going to be bigger, and that for me rules it out. 50 volts is more than enough for anything at or below line level. I usually run my line level equipment power supplies at +/- 17 volts. The X7R MLCCs are the best for power supply bypassing in as small a package as possible.
Last edited:
I put a 4562 in the first filter and it oscillated. I put a NE5532AP back in and it worked.
Magic combo is 5532 on the input, 4562 on the output. Without scrutinizing datasheets I surmise that the 5532 has a better phase margin.
I will further evaluate. If it sounds good (should be totally unobtrusive) it's a win. 45 mV offset in first stage is inconsequential. 3 mV on the output is what matters.
Magic combo is 5532 on the input, 4562 on the output. Without scrutinizing datasheets I surmise that the 5532 has a better phase margin.
I will further evaluate. If it sounds good (should be totally unobtrusive) it's a win. 45 mV offset in first stage is inconsequential. 3 mV on the output is what matters.
When replacing with the LM4562, I am wondering if you perhaps used the old tried-and-true method when upgrading to better, faster opamps: 100nF cap directly across the rails, and 22pF cap across the feedback resistor. That should cure any oscillation problems.
Project 99 - Subsonic Filter
Yes, thank you. I have used this method before. The problem is, there is no feedback resistor; the outputs are connected to the inverting inputs. And I always solder a 100 nF capacitor directly to the supply pins on the bottom of the board.
It is not clear to me whether adding a feedback resistor would change the transfer function of the circuit. If not, then using a resistor would address two issues; it would reduce offset as well as address stability problems.
Reducing the value of R3 and R6 should increase stability as well, at the expense of optimum Q and the likely addition of a few ripples in the low frequency response. This might or might not be consequential in this application.
It is interesting that I use a 2134 in noninverting configuration in my buffered volume control. Its input goes to the pot wiper. It is configurable for a gain of 1.4 or 2. There is a 22 pF capacitor across the feedback resistor. I've been using it for years with no hiccups or complaints.
Yes, thank you. I have used this method before. The problem is, there is no feedback resistor; the outputs are connected to the inverting inputs. And I always solder a 100 nF capacitor directly to the supply pins on the bottom of the board.
It is not clear to me whether adding a feedback resistor would change the transfer function of the circuit. If not, then using a resistor would address two issues; it would reduce offset as well as address stability problems.
Reducing the value of R3 and R6 should increase stability as well, at the expense of optimum Q and the likely addition of a few ripples in the low frequency response. This might or might not be consequential in this application.
It is interesting that I use a 2134 in noninverting configuration in my buffered volume control. Its input goes to the pot wiper. It is configurable for a gain of 1.4 or 2. There is a 22 pF capacitor across the feedback resistor. I've been using it for years with no hiccups or complaints.
Hmmmm....yes, I have heard that, in non-inverting buffers, the method of putting a 100 ohm resistor in the feedback rather than a direct short does help. I've never had to use it myself, but it certainly won't hurt anything.
Oh, really? Hmmmm....how much bigger? Remember that it is supposed to be UNITY gain, is it not?
Yeah, what is going on with that circuit in post #4, with the opamp supplies being modulated by the input signal??? Seems very strange to me......
Yeah, what is going on with that circuit in post #4, with the opamp supplies being modulated by the input signal??? Seems very strange to me......
22 pF? That's rather a lot. You only need to equal the input capacitance of the amp, divided by the closed loop gain, to perfectly cancel the input pole. 5 pF is usually plenty, but it depends on the gain and the op amp. This is known as cancelling the input pole with a zero.
If your feedback resistor is small, this is less of a problem, but a little feedback C can still enhance stability with extra gain margin. Don't go overboard, or you will harm stability by pushing the feedback to a higher ratio.
Yes, but then what value capacitor do I use for compensation? All the other impedances are much higher. It's not that simple.
I build buffers with the 5532 that have well under a millivolt output DC offset. I do this by using matched, precision, low noise resistors (CMF50) for the feedback and input bias resistors. A typical practical value for audio would 100K. Add a small capacitor (10-22 pF) across the feedback resistor, a 47-100 ohm resistor on the output, and you have a super stable buffer that can drive anything within reason. My point is that this (inconvenient) configuration enhances both stability and overall performance of the circuit.
The moral of the story is that a 100 ohm resistor will not do much to enhance the stability of this circuit, in my opinion. It is my hunch that adding a reasonable value feedback resistor (47-100K) and a small capacitor in parallel will enhance the stability, as well as greatly reduce DC offset. A larger resistor will add noise, and the situation calls for mitigation, not ultra precision. It's all a tradeoff.
At least that what I'm thinking at this point.
A 100K resistor is going to add noise. Root(4kTR*BW)
At that point, if you're already unity gain, the capacitor cannot hurt stability.
At that point, if you're already unity gain, the capacitor cannot hurt stability.
Yes. I use low noise, high stability resistors in the audio path, Dale CMF50 or CMF55 series. But they add noise.
Sometimes, like when you're driving a potentiometer, you need DC offset as low as you can get it. In a real world circuit, the potentiometer is going to dominaste the noise if you have DC current flowing through it. The practical solution is to lower DC offset voltage as much as possible.
The practical thing to do in this circuit is to continue to evaluate the circuit in its present state. If it's practical and doesn't degrade the sound, then I will move forward with it.
Sometimes, like when you're driving a potentiometer, you need DC offset as low as you can get it. In a real world circuit, the potentiometer is going to dominaste the noise if you have DC current flowing through it. The practical solution is to lower DC offset voltage as much as possible.
The practical thing to do in this circuit is to continue to evaluate the circuit in its present state. If it's practical and doesn't degrade the sound, then I will move forward with it.
If you're building a buffer for audio and you need 47K input impedance, then the best you can do is to use 47K resistors. If you need an input cap, then you can use a 470K bleeder resistor for an input impedance close enough to 47K.
If you don't need 47K input impedance, then you have a lot more leeway. But when there's specs to be met, you do what you gotta do. Any device that has a 47K input impedance has a 47K (or two 100K, etc) resistor on the input.
I'm telling you, the resistors I use make an audible difference. https://www.mouser.com/datasheet/2/427/cmfind-1762971.pdf I built my buffered volume control with cheaper resistors (and a cheaper volume control pot) and while it was perfectly usable, it was clearly inferior to the ones I build with the fancy resistors and Alps Blue Velvet.
If you don't need 47K input impedance, then you have a lot more leeway. But when there's specs to be met, you do what you gotta do. Any device that has a 47K input impedance has a 47K (or two 100K, etc) resistor on the input.
I'm telling you, the resistors I use make an audible difference. https://www.mouser.com/datasheet/2/427/cmfind-1762971.pdf I built my buffered volume control with cheaper resistors (and a cheaper volume control pot) and while it was perfectly usable, it was clearly inferior to the ones I build with the fancy resistors and Alps Blue Velvet.