PCB Layout

Are you sure that +16V should be connected to +ve signal input of the op-amp.
Even as a bias point, if it has a low impedance supply attached to it, the signal will be shunted to a.c. Gnd?
if it is a low impedance supply, simply add a series resistance in the line to the op-amp input, like you did on the bias for the first op-amp.
 
Are you sure that +16V should be connected to +ve signal input of the op-amp.
Even as a bias point, if it has a low impedance supply attached to it, the signal will be shunted to a.c. Gnd?
if it is a low impedance supply, simply add a series resistance in the line to the op-amp input, like you did on the bias for the first op-amp.
I've got it working on my current setup (Albeit 12v bias / 24v supply) Currently going from a DC power brick through LM317t's to regulate it and in.

The LM317 would mean its a low output Z supply if im not correct, but the prototype on the breadboard seems to be working fine. If I were to use a high value resistor like I did on the TL071 it would work regardless wouldn't it? Worse case I can throw on a jumper pad around the Resistor if I did need to use it without one.
 
Round 3.... ;)

  • Changed the Relay config as per Bigun's suggestion. I hope below if what you meant! Let me know if not :)
  • Increased Track widths to 3mm for power / OP and 1mm for signals
  • Generally moved bits about.
Any further thoughts?
Relay.png
Combined3.png
Top3.png



Bottom3.png
 
Yes, that's exactly what I meant regarding the relay, it's an old habit from using amps that require output protection relays, with a dc blocking cap like you have there shouldn't be any dc risk but it keeps the output cap from charging up from static pick up by the output cables etc. (I had further thought about what if you reverse pins 1 and 3 so that the output capacitor is held at ground when the amp is not powered up, but I'm not sure if this would put undue stress on the amplifier if it tried to drive a signal into a low a.c. impedance.)

V. interesting observation on the +16V bias. Certainly can't argue with the fact it's working in practice, something I don't understand. I'm guessing that the almost zero current draw on the voltage regulator you are using results in an anomalously high output impedance which would drop if you pulled decent current through - are you using the +12V supply for anything else or is it dedicated to the bias of the op-amp input pin?
 
Yes, that's exactly what I meant regarding the relay, it's an old habit from using amps that require output protection relays, with a dc blocking cap like you have there shouldn't be any dc risk but it keeps the output cap from charging up from static pick up by the output cables etc. (I had further thought about what if you reverse pins 1 and 3 so that the output capacitor is held at ground when the amp is not powered up, but I'm not sure if this would put undue stress on the amplifier if it tried to drive a signal into a low a.c. impedance.)

V. interesting observation on the +16V bias. Certainly can't argue with the fact it's working in practice, something I don't understand. I'm guessing that the almost zero current draw on the voltage regulator you are using results in an anomalously high output impedance which would drop if you pulled decent current through - are you using the +12V supply for anything else or is it dedicated to the bias of the op-amp input pins?
No just the bias so you are probably correct. Would having a large value resistance on that bias have any negative effect and what value would you recommend?
 
I can't imagine there being a negative impact from adding a resistor. I'd want it to be at least 10x the impedance of the circuit driving a signal into the opamp so that any signal loss at this point was 10% worse case. You have a 10k 'volume' pot driving the signal and that actually means worse case (highest) drive impedance is roughly 2.5k [1] so a 25k resistor min. value for the bias. (Hope I got my circuit math correct !). I don't anything much about these regulator chips, maybe they don't like being used without any current draw, you could put a resistor to ground on the output of the regulator to enforce some current flow - check the data sheet ????

What I can't say is what would be the optimal bias circuit for lowest noise. Modern regulators are usually pretty good.

[1] When the volume is set at min or max the wiper is connected to either the low impedance output of the preceding op-amp or to gnd which is also low impedance so worse case is when the volume is at half-way and then you have the wiper connected to gnd through half the value of the pot (10k/2 = 5k) and in parallel the wiper is also connected through the other half of the pot to the low output impedance preceding op-amp and those two 5k resistors are in parallel so the effective resistance is 2.5k.
 
I can't imagine there being a negative impact from adding a resistor. I'd want it to be at least 10x the impedance of the circuit driving a signal into the opamp so that any signal loss at this point was 10% worse case. You have a 10k 'volume' pot driving the signal and that actually means worse case (highest) drive impedance is roughly 2.5k [1] so a 25k resistor min. value for the bias. (Hope I got my circuit math correct !). I don't anything much about these regulator chips, maybe they don't like being used without any current draw, you could put a resistor to ground on the output of the regulator to enforce some current flow - check the data sheet ????

What I can't say is what would be the optimal bias circuit for lowest noise. Modern regulators are usually pretty good.

[1] When the volume is set at min or max the wiper is connected to either the low impedance output of the preceding op-amp or to gnd which is also low impedance so worse case is when the volume is at half-way and then you have the wiper connected to gnd through half the value of the pot (10k/2 = 5k) and in parallel the wiper is also connected through the other half of the pot to the low output impedance preceding op-amp and those two 5k resistors are in parallel so the effective resistance is 2.5k.

That's very helpful information actually because I was worried that adding the resistor would have a negative effect. I think i've found out which snippet of information i've neglected to mention.....

Below is the board for the regulators coming in. The output impedance from the voltage dividers would be acting as a 11k resistor feeing into the bias if im not mistaken. this is something ive overlooked without much thought. Would that be the reason its functioning currently? (although not optimally)
PSUSchem.png
 
Bingo, that explains it indeed. It’s probably not far from optimal too because there’s a tradeoff between wanting higher impedance at the bias generator and also not adding too much (resistor thermal) noise generated by that voltage divider.

Have you thought about the mechanical design, e.g. making sure you can access the pots to adjust them once the board is assembled and installed, will the i.c.’s be in sockets and if so, can you get a screwdriver blade (perhaps this is not recommended but it’s how I do it!) under them to remove them from the sockets or are you blocked by other components, etc.

Current paths - I try to analyse traces in my designs to see if currents (that always flow in loops) have small physical loop areas, which is a way of saying you want current going ‘out‘ along one trace to come ‘back’ along an adjacent trace or have a return path in a ground plane that follows the outgoing trace above it and doesn’t have to jog around some obstacle on the way. It’s as if the whole circuit was built out of twisted pair wiring. I’m not explaining it well, it’s something that’s painfull to spell out in a short post but if this strikes a chord with you then good! It’s only really important for a.c. signals (more so at high frequenciesj, bearing in mind that the largest a.c. currents will be through the p.s.u. - output stage - external load.
 
Last edited:
Bingo, that explains it indeed. It’s probably not far from optimal too because there’s a tradeoff between wanting higher impedance at the bias generator and also not adding too much (resistor thermal) noise generated by that voltage divider.

Have you thought about the mechanical design, e.g. making sure you can access the pots to adjust them once the board is assembled and installed, will the i.c.’s be in sockets and if so, can you get a screwdriver blade (perhaps this is not recommended but it’s how I do it!) under them to remove them from the sockets or are you blocked by other components, etc.

Current paths - I try to analyse traces in my designs to see if currents (that always flow in loops) have small physical loop areas, which is a way of saying you want current going ‘out‘ along one trace to come ‘back’ along an adjacent trace or have a return path in a ground plane that follows the outgoing trace above it and doesn’t have to jog around some obstacle on the way. It’s as if the whole circuit was built out of twisted pair wiring. I’m not explaining it well, it’s something that’s painfull to spell out in a short post but if this strikes a chord with you then good! It’s only really important for a.c. signals (more so at high frequenciesj, bearing in mind that the largest a.c. currents will be through the p.s.u. - output stage - external load.
No it definitely makes sense, especially the twisted pair analogy. I assume the ground plane would have the same effect as there is usually a convenient path to ground?

As for the bias impedance, the voltage divider has 11k out z on the amplifier, so I believe I should increase it, not not at the expense of extra noise.

As for the bias of the buffer, this would be reduced because of the divider I assume? I'm at work so I'll look when I'm home, but I'm assuming the voltage divider and the 1m resistor are in parallel?
 
the risk with relying on a ground plane is that sometimes it gets 'cut' to allow another trace across the board - creating a discontinuity in the plane that forces current to flow around the cut. Not saying this applies to your board, just making a generic observation that I've learned from reading about such things and from a presentation at my work by a talented designer a few years ago.

Given that the circuit is working the bench and that you are happy with it, there may be no compelling reason to adjust voltage divider values except out of a desire to optimize the design - whether such optimizations are worth it is up to you, any changes have the risk of 'unintended consequences' :)
 
the risk with relying on a ground plane is that sometimes it gets 'cut' to allow another trace across the board - creating a discontinuity in the plane that forces current to flow around the cut. Not saying this applies to your board, just making a generic observation that I've learned from reading about such things and from a presentation at my work by a talented designer a few years ago.

Given that the circuit is working the bench and that you are happy with it, there may be no compelling reason to adjust voltage divider values except out of a desire to optimize the design - whether such optimizations are worth it is up to you, any changes have the risk of 'unintended consequences' :)
Yes, Worse case I'll be just replacing a resistor of the same footprint anyway :)

Thanks for the feedback! :)