Summing amp?

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
I'm not a fan of onboard regulators in mixers, particularly 78/79's. They inject noise in the "ground" path that can too easilly pollute the whole system. My first mixern when I was wet behind the ears, had a 7824 on each channel (yes, no bipolar PSU at the time). I was not pleased with the noise performance.One day, as an experiment, I replaced the 7824 with a 10 ohm resistor. The noise improvement of the summing amp was significant. After that, no more on-board regs. Saves heat too.
I never tried with 317/337 though.
+1 as I've shared before, I once black-listed 3 terminal regulators from a major US manufacturer when they refused to provide a maximum noise spec for their 3-terminal regulators. In a big console everything can matter, and noisy 3 terminal regulators can matter. 🤔

JR
 
+1 as I've shared before, I once black-listed 3 terminal regulators from a major US manufacturer when they refused to provide a maximum noise spec for their 3-terminal regulators. In a big console everything can matter, and noisy 3 terminal regulators can matter. 🤔

JR
Does a 3 terminal regulator produce significant noise in that case that the input only gets pre-cleaned DC? Is there a correlation between the voltage the regulator has to drop and the noise it produces?
 
Does a 3 terminal regulator produce significant noise in that case that the input only gets pre-cleaned DC? Is there a correlation between the voltage the regulator has to drop and the noise it produces?
I accept responsibility for a large console design that was susceptible to regulator output noise (hiss). I suspect it was related to power supply decoupling and grounding systems.

I wouldn't over think the design what-ifs... I'd focus on objective measures to allow you to quantify what you are hoping to improve.

JR
 
I could write a small book on the subject, but what most circuit designers neglect to consider is that every bypass capacitor forms a path for rail noise to be injected into the very local ground network. What's missing in 99% of designs I see is decoupling impedances in the power rails to filter/attenuate the path from one stage of the signal chain to the next. We have to think in terms of physical current pathways and consider that the "ground" doesn't have zero impedance. It's an enlightening, but time-consuming, exercise to model a circuit board where the model includes trace resistance in ground and power rails. And, if you want to go even deeper, model their inductances as well, including the mutual inductance (coupling) between traces near and parallel to each other. Those little schematic symbols that simply point to ground and power rails can lead us to believe that parasitics are negligible. Often it can mean the difference between stability, marginal instability, or outright oscillation.
 
Physical ground trace resistance and impact of current flow becomes readily apparent when laying out power supplies with reservoir capacitors being charged by rectifiers. The real world trace resistance and peak charging currents make themselves known with visible voltage drops. 🤔

If you write a book I might be inclined to read it, but don't make it too long. ;)

JR
 
I could write a small book on the subject

The way I usually put this is (for clarity and to have something in mind when looking at a circuit):

1) Current always flows in loops and along the path of least impedance.
2) Ground isn't! (Ground is a signal line).
3) Voltages are commonly the consequence of currents flowing in a non-zero resistance/impedance.

And keeping these rules in mind leads me to draw schematics often like this:

1697188622853.png

In other words while most EE's work on a "Voltage" focused paradigm (because all active devices are voltage input devices), but it is current flow that actually matters and the impedance's the current flows in and even the coupling between the (power/ground) lines current flows in and the signal lines.

This is as valid in Audio, as it is in industrial and military electronics or digital (non-audio).

Thor
 
The way I usually put this is (for clarity and to have something in mind when looking at a circuit):

1) Current always flows in loops and along the path of least impedance.
2) Ground isn't! (Ground is a signal line).
3) Voltages are commonly the consequence of currents flowing in a non-zero resistance/impedance.

And keeping these rules in mind leads me to draw schematics often like this:

View attachment 115597

In other words while most EE's work on a "Voltage" focused paradigm (because all active devices are voltage input devices), but it is current flow that actually matters and the impedance's the current flows in and even the coupling between the (power/ground) lines current flows in and the signal lines.

This is as valid in Audio, as it is in industrial and military electronics or digital (non-audio).

Thor
Love the schematic!! De-coupling in the supply lines, too! This should be the norm for circuit designers - and it has the added benefit of guiding the PC board layout. Thinking in terms of current, as you say, is the key!
 
I've conducted more tests and got the best results in the console decoupling NE5532/5534 with op amps driving no external loads by decoupling with 100nf ceramics directly at the power pins and connected to the nearest ground, with the addition of a high-ESR 10uf electrolytic rail-to-rail.

Regarding the question of injecting noise into the audio ground - is this a problem when decoupling with small caps, say 100bf? Isn't any noise far outside of the audio band in this case? Would low level out-of-band noise dumped into the audio ground affect proper functioning of an op amp (I'd guess that higher up noise could cause a problem even at lower levels because of loop margins, slew rate and other concepts I do not mathematically grasp)? This is inside a console with twice regulated power and lot's of shielding.
 
Regarding the question of injecting noise into the audio ground - is this a problem when decoupling with small caps, say 100bf? Isn't any noise far outside of the audio band in this case? Would low level out-of-band noise dumped into the audio ground affect proper functioning of an op amp (I'd guess that higher up noise could cause a problem even at lower levels because of loop margins, slew rate and other concepts I do not mathematically grasp)? This is inside a console with twice regulated power and lot's of shielding.
There will always be noise currents dumped into sundry grounds.

Signals need to be handled differentially, grounds are not 0V

Do you have a noise problem ?

JR
 
I've conducted more tests and got the best results in the console decoupling NE5532/5534 with op amps driving no external loads by decoupling with 100nf ceramics directly at the power pins and connected to the nearest ground, with the addition of a high-ESR 10uf electrolytic rail-to-rail.

What are the parameters wrt "best results" ? Or to put it another way - what were the deficiencies with alternative schemes ?
 
What are the parameters wrt "best results" ? Or to put it another way - what were the deficiencies with alternative schemes ?
Bad sonics and high THD with the original layout (insufficient decoupling). Oscillation with certain brands / production dates of NE5534 in certain spots (in a circuit designed for the NE5534). Higher THD of a nearly identical stage at a different location on the PCB.

After the added decoupling better nulling with the signal not travelling through the circuit. With the 2*100nf to ground and a 10uf R2R THD got a little bit better than with - say - a couple 10uf to ground.
 
Thanks. What degree % or dB of increase in THD ? Any figures for IMD ? Any comparison with something like 2 X 10u from rails to 0V somewhere on the PCB then either 2 X 100n from rails to 0V or a single 100n R2R near the NE5534 ?
 
Thanks. What degree % or dB of increase in THD ? Any figures for IMD ? Any comparison with something like 2 X 10u from rails to 0V somewhere on the PCB then either 2 X 100n from rails to 0V or a single 100n R2R near the NE5534 ?
Unfortunately not, sorry. I jump around a lot with levels and frequencies to find differences between the channels decoupled in different ways. I have not currently connected equipment that enables me to sensibly measure absolute THD levels with. And the signal travels through the rest of the console afterwards still, so there is no way to test circuit parts in isolation at this time.

I do think that THD in itself doesn't matter much once its below a certain level (maybe - 80 db, maybe a little more). But in an anlog circuit it can be a good indicator that something is wrong / can be improved.
 
I've conducted more tests and got the best results in the console decoupling NE5532/5534 with op amps driving no external loads by decoupling with 100nf ceramics directly at the power pins and connected to the nearest ground, with the addition of a high-ESR 10uf electrolytic rail-to-rail.

Regarding the question of injecting noise into the audio ground - is this a problem when decoupling with small caps, say 100bf? Isn't any noise far outside of the audio band in this case? Would low level out-of-band noise dumped into the audio ground affect proper functioning of an op amp (I'd guess that higher up noise could cause a problem even at lower levels because of loop margins, slew rate and other concepts I do not mathematically grasp)? This is inside a console with twice regulated power and lot's of shielding.
Briefly - the noise "far outside the audio band" is a big problem for op-amps because the op-amps have no open-loop gain at those frequencies to correct their open-loop non-linearity (which is often quite horrible). This non-linear behavior turns the amplifier into a classic "mixer" with sum and difference spectral components being created. Many of the difference components are in the audio band and are harmonically unrelated (discordant) with the original signal. The late Deane Jensen described this in his 1988 AES paper about "Spectral Contamination". Band-limiting between stages is one way to reduce the accumulation of the contamination, but controlling contamination of "ground" is also necessary. Adding series impedances in power rails greatly reduces ground contamination by attenuating noise from other stages feeding from the same rail - having bypass caps right at the supply pins of op-amps is a good thing (maintains stability) but all the benefit of using "star" connections in the grounding scheme can be undone by forgetting to either add the series impedances or "star" connecting the supply rails as well.
 
Mr CMRR's point about the need for isolation between the power supply lines of amplifier sections to reduce noise is a very good one.

One thing I haven't seen mentioned in this thread is that when OP amp designs oscillate because they have inadequate phase margin, any oscillations well above the audio frequencies are still problematic because they consume available system current and power. So its not just a problem if they only oscillate in the audio spectrum. Of course any unwanted oscillation, even high ones can intermodulate with audio frequencies too, to produce audible frequencies that are probably not very musical.
 
Having op amps oscillating at high frequency is a good way to roast your tweeters (monitors) so having an unrestricted bandwidth inside gear CAN be problematic (I Know from a costly trip to Scotland many years ago on behalf of someone else).
 
Back
Top