Genuine question: will that help with the oscillation though if I've identified the other opamp as the source? Or is it a good idea in general?
OK, that's 'good' to know, thanks. And frustrating; this is actually rev2 and the first had other problems but didn't oscillate 🤦♀️
First you have to solve the power supply, it's not good as it is. Then reduce the resistors in the buffer stage. After that comes the tone generator and oscilloscope and possibly compensation.
Thats why i didnt try to spread my basic knowledge about fundamentals, it was already posted by mooly and other respective members. Now dont shout, it hurts my ears. Thank you.
I'm pleased to hear it worked 🙂
Lots of unknowns with board layouts tbh and particularly where you have tracks on both sides running together. That is enough to allow capacitive coupling although I think that coupling would have to be from the second opamp output to the inverting input of the first to cause instability (positive feedback).
Adding small caps as we have done is fine and empirical methods often work well. You could keep reducing the value until there was a problem and then increase it by say 50%. You might find even 6.8pF (or lower) is enough... or maybe not... you just have to try them.
If the pots are metal cased then that can cause issues with stray pickup as well.
Lots of unknowns with board layouts tbh and particularly where you have tracks on both sides running together. That is enough to allow capacitive coupling although I think that coupling would have to be from the second opamp output to the inverting input of the first to cause instability (positive feedback).
Adding small caps as we have done is fine and empirical methods often work well. You could keep reducing the value until there was a problem and then increase it by say 50%. You might find even 6.8pF (or lower) is enough... or maybe not... you just have to try them.
If the pots are metal cased then that can cause issues with stray pickup as well.
Some learn better by doing. Why discourage that? OP did prototype the circuit on a breadboard first and said it worked well, so he moved on to a PCB design. That seems like a pretty logical step to me. In this case, however, it appears that the parasitic capacitance of the breadboard – which can be significant – prevented oscillation. The parasitic capacitance of the PCB is much lower, so the circuit oscillated. Live and learn.
PCBs are so cheap these days that it makes complete sense to move onto a PCB pretty early in the design cycle. I used to make prototype boards in my garage. But with the low cost of PCBs I've gotten rid of all the chemicals and equipment. Don't miss it.
You're right that stacking traces results in capacitive coupling between them, but unless we're talking really wide and long traces it is unlikely to impact circuit performance for audio. You can do some back-of-envelope math on it if you wish. The resistors look to be maybe 150 mil apart so some of the larger coupling capacitances will be something like 20 mil (0.5 mm) trace width by maybe 6-7 mm. FR4 has a dielectric constant of 3.8-4.8, so I used the average (4.3). That gives me 87 fF for a 0.5x7 = 3.5 mm2, d = 1.53 mm capacitor.
It wouldn't surprise me if the fringe capacitance of two traces running in parallel was higher than the plate capacitance, but few seem concerned about the fringe capacitance. 🙂
I would still prefer that the circuit had an actual ground connection at the power supply. But that's me.
Tom
That's a good point when you think about it. With dual opamps you have pins 1 and 2 and 7 and 6 with parallel conductors on the breadboard (the opamp output and inverting input). It must easily be several pF if not more like 10's of pF.
I wonder if you could elaborate on that point. I haven't been able to get a clear feeling for the practical connection of grounds, hence leaving off the DC ground connection here in the interests of space saving.
My understanding of star grounding from what I've read (= high noise to signal ratio on the information..!):
- Each board should have a ground connection that meets at the star point.
- Loops are/can be antennae so should be avoided.
- Assuming the PCB in question is itself not the star point, it follows that each board should only have one connection to ground in order to avoid 2.
Granted in hindsight if I have only one ground connection on the PCB it perhaps should be the DC ground as the one likely to have to carry the highest current, but I'm not sure that's what you mean.
Last edited:
Indeed. ~AU$10 for 5 of these, incl. delivery within about a week and a half, really doesn't take much thinking about.
We call it a ground, but it is not, what if you are in a spaceship? Let's refer to it as the common return path of all signals, maybe this becomes clearer. You can decouple the AC return signals from the DC return signals by means of capacitors and resistors to avoid any loop. Ground actually refers to a safety point that is connected to anything that you can touch (like a metal chassis) that would not allow a mains fault to cause you harm but a low impedance return to the ground in your house. Never mix safety ground with signal or DC ground, they are not interchangeable in my view.
A start point is a a point that you create for all signal paths to return to. A ground plane serves the same purpose because it is a natural star point since the signal will find the shortest path to the common lowest impedance single return connection. Star grounds and power points are a common practice for low frequency. In RF and microwave, we use planes, and trackwidths and lengths calculated exactly using transmission line theory, even track thickness becomes an issue and physical track star points are avoided because of the adding parasitic anomalies. RF design engineers always use planes, and thereby avoid loops or uncontrolled parasitic anomalies.
Last point, if you want to avoid a loop, you insert a resistor to "cut" it. For instance, you can see this in many competent designs that the input connection "ground" is lifted by say a 10 ohm resistor that will cause a ground loop, not to "exist". In short, when you have a stereo signal return path from the outside equipment entering the amplifier each of the L and R signal returns (ground) is "buffered" from the amplifier common signal return "ground" by a resistor so that a loop low impedance loop is not present or created. I am not a lecturer, maybe someone can explain the physical constraints better than me. There are literature available on transmission line theory available but this is a in depth subject on its own. Just follow the common practice and you will be okay
A start point is a a point that you create for all signal paths to return to. A ground plane serves the same purpose because it is a natural star point since the signal will find the shortest path to the common lowest impedance single return connection. Star grounds and power points are a common practice for low frequency. In RF and microwave, we use planes, and trackwidths and lengths calculated exactly using transmission line theory, even track thickness becomes an issue and physical track star points are avoided because of the adding parasitic anomalies. RF design engineers always use planes, and thereby avoid loops or uncontrolled parasitic anomalies.
Last point, if you want to avoid a loop, you insert a resistor to "cut" it. For instance, you can see this in many competent designs that the input connection "ground" is lifted by say a 10 ohm resistor that will cause a ground loop, not to "exist". In short, when you have a stereo signal return path from the outside equipment entering the amplifier each of the L and R signal returns (ground) is "buffered" from the amplifier common signal return "ground" by a resistor so that a loop low impedance loop is not present or created. I am not a lecturer, maybe someone can explain the physical constraints better than me. There are literature available on transmission line theory available but this is a in depth subject on its own. Just follow the common practice and you will be okay
Last edited:
Star grounding is a good way to confuse yourself.
I'm much more a fan of this line of thought. More below.
I think, technically, that's actually called "Protective Earth" (PE), but that's another matter. We agree either way.
Thinking about ground in terms of return currents is more helpful that trying to keep all grounds flowing to a single point. Most audio circuits are single-ended, i.e., they sense the voltage from a signal pin to some ground reference. This means the ground return path is in series with the signal, so any unwanted current (error current) that also flows in that return path will introduce an error voltage that's in series with the signal.
There are basically two ways to resolve this: 1) Use differential signalling and 2) minimize the error current.
Star grounding is a (in my view) misguided attempt at minimizing the error current. I say 'misguided' because star grounding also maximizes the impedance of the ground nets, which works against the overall goal of minimizing the error voltage.
A ground plane offers the lowest impedance you can get, but it makes it harder to control where the current flows, which is likely why many audiophiles shy away from them.
The other aspect of grounding, which relates to my concern about floating ground references, is that the input voltage of the opamps must be kept within the opamp's input voltage range. For modern opamps that's within the supply rails, but for older types this could be more restricted. If the input voltages stray beyond the limits some opamps, especially the older ones, can have really unpredictable behaviour, including phase reversals. If the ground reference is floating there's nothing that limits the input voltage to the opamps, so you risk undesirable behaviour. The way to address this is to ensure that the power supply ground and signal ground are connected at the board level.
Tom
I believe your self oscillation comes from the positive and negative rail voltage.
I have had 5532 not sounding good in a few trials. But with separate decoupling of + and - to each op via 100 ohm and 10 uF the problem is gone.
I have had 5532 not sounding good in a few trials. But with separate decoupling of + and - to each op via 100 ohm and 10 uF the problem is gone.
My immediate concern is the neat array of resistors - layout is about making the electrons happy, not about making it look pretty. Typically the layout should follow the schematic, with separate parts in separate areas so reduce cross-talk between independent parts of the circuit.
Also resistors 1,2,5,6 are way too large, so the input of those opamps is very susceptible to capacitive pick-up. I understand you want high input impedance and an inverting stage to cancel the Baxandall stage's inversion, but that's a compromize that isn't great - I would buffer the inputs with followers, then invert at low impedance (5k or so - this would reduce the noise levels too). Often adding more components makes better performance (despite many people obsessively believing the opposite). It's usually good to use a stage for each function - one to buffer, one to invert, one to filter - straightforward and each can be optimized independently of the others that way.