Shredhead, could it be, if connecting all the grounds to chassis lowers the noise floor, it is a bad PCB layout that is to blame?
You use the word "grounded", but in most cases we are talking about connections for audio (return) signals. Do you want the chassis to be part of your circuit?
More food for thought. If instead of using the chassis to connect the grounds, you could place the inputs and output close together on the PCB, then connect the grounds together. Would this not be a better solution?
You use the word "grounded", but in most cases we are talking about connections for audio (return) signals. Do you want the chassis to be part of your circuit?
More food for thought. If instead of using the chassis to connect the grounds, you could place the inputs and output close together on the PCB, then connect the grounds together. Would this not be a better solution?
Many here use their ears as measuring instruments.
This gives the rest of the Membership bad to very bad data from which we sometimes try to draw conclusions.
Instead we should use intrumentation to measure noise.
State what kind of instrumentation one is using and the measurments obtained and then let the Membership interpret the data and draw conclusions.
Stating: " I did it this way and I can't hear any hum" is completely worthless when trying to compare results of other's experiments.
I completly agree with that, and it is not in contradiction with the connection of ground to the chassis.
You can (must) put the output on one side of the box (near the pcb output), near the supply input and the input on the other side (pcb input).(I think) It is the best way to do when possible.
Do you know the impedance of a 2mm thick copper plane in audio frequency domain ?
The impedance between the most distant 2 points in one of my amplifier chassis is ~10µohm.
That's not the way H.Ott explains it.
He suggests that in/outs should all be grouped together at the entry to the box/PCB
and pass 10Apk across that and you get 100uVpk.
Compared to a 1Vac signal that is just -83dB
Yes. Some people put outputs near inputs, or inputs near power supplies, then wonder why they get problems.herve00fr said:
To me, it has been a very efficient way to avoid problems.
As well as making circuits of small dimensions, so that they have short loops, I think it was the good idea of the "Gainclone" amp, the other ideas behind it being dubious.
I do like that when the other way ids not possible or not necessary (not on the pcb).
1) There will never be 10A flowing in the chassis, except if the output stage blow
2) The 2 most distant points are note used at least on the same amplifier
3) Using the chassis as ground does not avoid making a reasonable placement of sensible circuitry
For a low gain low impedance circuit you can get away with putting inputs and outputs near each other. For higher gain or higher impedance situations you get capacitive feedback and the potential for instability.
^^^This x2
Right on the money DF, I'm boosting 20Hz and down for sealed subwoofer EQ up to 12dB so thanks for this advice. As most of you know, there is some gross noise that hides down there (for folks reading who don't know, this is why most companies hide behind weighted specs for their S/N ratio). Anything that I can do to push that noise floor down, I am game for. I will provide you all with more details of my project soon. I'm still in protoboard stage with this but I try my hardest to layout with small loops and I'll report the measurements at the end of next week. I appreciate all of the discussion in this thread, thank you guys.
2.54mm?herve00fr said:
Opamp pins are small so have small stray capacitance and it is likely that the input is either a virtual ground or fed from a lowish impedance. If not, stability problems may need investigating.
Your point is?
When often experimenting, you may meet unexpected hum from time to time. When a particular grounding scheme markedly diminishes the times this occurs, you can think it's a preferable way.
Hearing phones carefully connected to the output of a power amp without signal at its input already tells a lot of the story.
Looking at the spectrum of the noise floor tells a lot of the story. Any kind of listening test is subjective and therefore impossible to accurately describe to others or use for comparisons.
This used to be sometimes mentioned in the articles I read when I started electronics. I never met instability in audio due capacitive feedback.
You are right. Spectrum reveals much more than what I called a lot of the story. It puts almost all defaults in light which, despite many of them being inaudible, I am absolutely in favour to fight. Currently, spectrum images to support arguments in the present discussion are lacking.
Listening a system having a null signal at its input through phones is a good primary test to be sure there are no gross unwanted signals. It's more tough than listening with the ear very near the drivers as many people do, but I never saw anybody claiming to have done it.
At die level, it's 500µm max, and you have to consider trace capacitance.
Input is low impedance only if you have nothing between previous stage and input, not the case in filter for example.
With RCA connectors, distance is 3cm min.
Connections have to be made with coax cable.
Source driving the input have to be low impedance.
So my point is that you have no chance to have capacitive coupling between in and out.
Many Members have posted the suggestion to use headphones to help detect low level Hum+Noise signals. Even I have suggested this subjective method.
Shred summs it up.
ANY subjective description cannot be used by others for effective engineering comparison.
Good data is needed. Bad data is worse than useless.
A complement of what I already said.
It must not be forgot that the noise signal of the circuit to analyze travels to the analyzer through cables and connectors and has to be amplified by a very low noise amplifiying circuit before being submitted to the analysis hardware. So one has to be sure that the whole measuring apparatus does not introduce errors itself.
EDIT : an idea to circumvent this difficult could rely in temporarily increasing the closed loop gain of the device under test in order to have a large noise signal.
Last edited:
I have. This wasn't an input-output problem, but a valve LTP phase splitter. Polypropylene coupling caps did not help as they are so large. I had to add a screen/shield so the in-phase stage output could not 'see' the input.forr said:
- Status
- Not open for further replies.