Rudi i will be the last one here to make suggestions but for the fun of it this typical layout suffers also from a number of problems depending on the pov ( point of view )
for example
---output trace is too long and might "spread" things around
---similar to that is the ground trace
---by the same logic if zobel is accommodated this way it might also spread dirt in your LTP or current mirrors
---There is no clean ground ( through 10R )
---Spreading the output trace so widely means that feedback resistor that has to be taken from the middlest point suddenly is too close to the zobel
---The way ground is spread supplying secondary circuits ( LTP /VAS )is getting to be a problem ( to many links to many long traces )
---while feedback as said needs to be in a middle point so is the ground of it and that actually cuts your pcb in half while the left side might accommodate the LTP parts and may be VAS then current mirror will have to be on the other side and then far away from LTP
and these are the things that can be seen with a naked eye
roll the dices to correct the above things and you loose symmetrics and looks
roll the dices for the symmetrics and looks and you will end up with a double layer pcb
Go for a double layer pcb and then you need to learn a holly lot about inductance and capacitance between traces and layers that you never thought off ...
Then again all the above have also another factor ..... cost !!!!
for example
---output trace is too long and might "spread" things around
---similar to that is the ground trace
---by the same logic if zobel is accommodated this way it might also spread dirt in your LTP or current mirrors
---There is no clean ground ( through 10R )
---Spreading the output trace so widely means that feedback resistor that has to be taken from the middlest point suddenly is too close to the zobel
---The way ground is spread supplying secondary circuits ( LTP /VAS )is getting to be a problem ( to many links to many long traces )
---while feedback as said needs to be in a middle point so is the ground of it and that actually cuts your pcb in half while the left side might accommodate the LTP parts and may be VAS then current mirror will have to be on the other side and then far away from LTP
and these are the things that can be seen with a naked eye
roll the dices to correct the above things and you loose symmetrics and looks
roll the dices for the symmetrics and looks and you will end up with a double layer pcb
Go for a double layer pcb and then you need to learn a holly lot about inductance and capacitance between traces and layers that you never thought off ...
Then again all the above have also another factor ..... cost !!!!
Hi,
Yes sakis, 'shematic style PCB' design suffers from several issues.
I'll might add,
Long ground traces to decoupling caps - big current loop area(s) that encloses small signal circuitry.
Instead, if power traces must reach the front of PCB, run them close together and outside of other circuits. It serves two purposes:
1. Very good decoupling - trace length to/from caps minimized.
2. Minimal exposure of magnetic fields that could influence low level and/or high impedance circuitry.
Look at commercial PCB design, it's revealing.
Yes sakis, 'shematic style PCB' design suffers from several issues.
I'll might add,
Long ground traces to decoupling caps - big current loop area(s) that encloses small signal circuitry.
Instead, if power traces must reach the front of PCB, run them close together and outside of other circuits. It serves two purposes:
1. Very good decoupling - trace length to/from caps minimized.
2. Minimal exposure of magnetic fields that could influence low level and/or high impedance circuitry.
Look at commercial PCB design, it's revealing.
Hi,
One example from a simple Rotel amplifier. It might not be optimal in every sense but at least the Rotel PCB designer avoided spacing the rails. The good old NAD 3020 is another classic example worth a review.
One can find quite a few PCB design threads here with good information. For example 'marce' do PCB design professionally and has provided good technical info on this forum.
One example from a simple Rotel amplifier. It might not be optimal in every sense but at least the Rotel PCB designer avoided spacing the rails. The good old NAD 3020 is another classic example worth a review.
One can find quite a few PCB design threads here with good information. For example 'marce' do PCB design professionally and has provided good technical info on this forum.
Attachments
Nico is right ....
there is no books about design like described above...my approach was in a way a teaser to a nice looking pcb that Roudy made ...
If you read closer to my post you will understand that in a way pcb design for audio amps is like playing a play station game .... You have to go by the rules and always give something to gain something ...
The P3a could cover lets say 80-85% of the given rules only in a double layer pcb which obviously i can design ( million other people also ) but actually cannot evaluate properly
when you go for a single layer less of the 85% of the rules will be achieved and which one is a winner and which one stays out of the picture will define a good amplifier from a very good amplifier
your lay out is fine ( couple of looks with naked eye i didnt spent more time on it ) but if you have higher or very higher targets this is not gonna happen with this pcb Of course you need to remember that in this article is a lot of snake oil and some of us talk about things that are hardly measurable ....
Go ahead ... choose your semis carefully check related threads about matching LTP choice of VAS and so on and you are going to have a very nice amp !!!
there is no books about design like described above...my approach was in a way a teaser to a nice looking pcb that Roudy made ...
If you read closer to my post you will understand that in a way pcb design for audio amps is like playing a play station game .... You have to go by the rules and always give something to gain something ...
The P3a could cover lets say 80-85% of the given rules only in a double layer pcb which obviously i can design ( million other people also ) but actually cannot evaluate properly
when you go for a single layer less of the 85% of the rules will be achieved and which one is a winner and which one stays out of the picture will define a good amplifier from a very good amplifier
your lay out is fine ( couple of looks with naked eye i didnt spent more time on it ) but if you have higher or very higher targets this is not gonna happen with this pcb Of course you need to remember that in this article is a lot of snake oil and some of us talk about things that are hardly measurable ....
Go ahead ... choose your semis carefully check related threads about matching LTP choice of VAS and so on and you are going to have a very nice amp !!!
Thanks guys. I now know that there are some basic guidelines to follow based on general electronics know-how, but that the "piece the resistance" lies within details.
I was alarmed by the;
"This is horrible", "OMG!", "That PCB can never perform well"-comments.
So far, thanks for the shakedown and I wished there was something written about class AB power electronics layouts.
The thing is, I am new to this and have always followed an IC-datasheet on layouts. Now that I built a circuit from scratch, I want to understand what goes well and wrong. Since I could not find a decent piece of literature on such content, I decided to put it up for review by the more experienced guys in electronics.
So far, thanks guy.
I was alarmed by the;
"This is horrible", "OMG!", "That PCB can never perform well"-comments.
So far, thanks for the shakedown and I wished there was something written about class AB power electronics layouts.
The thing is, I am new to this and have always followed an IC-datasheet on layouts. Now that I built a circuit from scratch, I want to understand what goes well and wrong. Since I could not find a decent piece of literature on such content, I decided to put it up for review by the more experienced guys in electronics.
So far, thanks guy.
I redid a bit of the layout after reading your comments and this morning made another pcb. It performed similar, lots of 200-800 noise. I want to lose that noise. Since I have fairly efficient loudspeakers and would like to use more than 30% of my volume control, I tried to lower the gain. That resulted in a larger volume control action (obviously) and seemed to drop some of the noise. THS+N is now 0.02% And the low frequency noise has peaks to -90dBu.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
Can you also post a black on white traces only PCB layout?
I can't find the Power Ground, nor the Signal Ground, with all that noise (devices and notes & labels) superimposed.
The FFT shows a lot of mains hum and buzz. There is still something wrong.
Have you measured the noise on the amplifier output with the input shorted to signal ground? What is it?
I can't find the Power Ground, nor the Signal Ground, with all that noise (devices and notes & labels) superimposed.
The FFT shows a lot of mains hum and buzz. There is still something wrong.
Have you measured the noise on the amplifier output with the input shorted to signal ground? What is it?
Last edited:
The measurement is a 1kHz sinewave, going from my pc, through my pre-amp, to the P3A, to a load (around 10 Ohm), and scaled down to not overload the soundcards linein. Without the amplifier in the measurement-loop the noise is lower and the only low frequency peak is at 200 Hz.
By adding the amplifier 200-800Hz noise is added.
In the PDF is a black and white pcb layout. The jumper is not shown!
Powerground is between the two power transistors on the top and the signal ground is near the input, on the right-hand side.
By adding the amplifier 200-800Hz noise is added.
In the PDF is a black and white pcb layout. The jumper is not shown!
Powerground is between the two power transistors on the top and the signal ground is near the input, on the right-hand side.
The system logged me out again. This time the system did not lose my post. Able to copy and paste and repost.
This layout is much better.
The decoupling is where it is needed. The left most HF cap can be moved to between the PG plane to the emitter PIN of Q7. The right most HF cap could also be moved but the trace length saving is not as great.
The Signal Ground is separated from the Power Ground. Add a resistor location. There is space next to X1. 10r or 0r0 or open gives three choices to try when the amp is ready to test. You can add this to your present PCB using holes drilled through that wide trace and cutting the trace between those added holes.
I would suggest rearranging the input components to significantly reduce the loop area of all the input traces. 0.1" pitch instead of 0.4" pitch for the Signal Hot & Signal Ground would be better and could be moved over to the space below R1.
The output Zobel is injecting it's noise directly into the dirty ground.
I would like to see the right hand power input moved over to the left side so that it runs alongside. This reduces loop area of the current pulses in the power supply traces.
This layout is much better.
The decoupling is where it is needed. The left most HF cap can be moved to between the PG plane to the emitter PIN of Q7. The right most HF cap could also be moved but the trace length saving is not as great.
The Signal Ground is separated from the Power Ground. Add a resistor location. There is space next to X1. 10r or 0r0 or open gives three choices to try when the amp is ready to test. You can add this to your present PCB using holes drilled through that wide trace and cutting the trace between those added holes.
I would suggest rearranging the input components to significantly reduce the loop area of all the input traces. 0.1" pitch instead of 0.4" pitch for the Signal Hot & Signal Ground would be better and could be moved over to the space below R1.
The output Zobel is injecting it's noise directly into the dirty ground.
I would like to see the right hand power input moved over to the left side so that it runs alongside. This reduces loop area of the current pulses in the power supply traces.
Last edited:
something is imposing all that noise on the FFT results.
Measure the amplifier alone.
What is the mVac and mVdc using the 199.9mV scales of a narrowband DMM. It does not need to be an rms measurement. An average reading DMM is good enough to give us and you a good clue. A reading of 0.0mVac means that the narrowband noise is <=0.0499mVac
If that includes all the hum and buzz and all the 40Hz to 1kHz noise and a proportion of the 1kHz to 20kHz noise then that is good enough for first assessment.
PS.
I don't know how to interpret the post 78 plot.
If that includes all the hum and buzz and all the 40Hz to 1kHz noise and a proportion of the 1kHz to 20kHz noise then that is good enough for first assessment.
PS.
I don't know how to interpret the post 78 plot.
I don't have a narrow band DMM. It goes up beyond 20kHz according to the specs so the results might be a bit pessimistic. Shorted, well the end of the 20cm coaxial cable is shorted, gives me 0.3 mVac. Dc is a disappointing 30 mVdc. It was lower in my previous design, but it's a "Work In Progress" right?
Ps. I posted a very low noise picture earlier when the input is shorted, that one is not correctly measured. Noise is around -100 dBu
Ps. I posted a very low noise picture earlier when the input is shorted, that one is not correctly measured. Noise is around -100 dBu
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.