Hi Mark and Sangram,
Here is the corrected schematic and a draft layout (much simplified as per Sangram's Suggestion). Any views on the same?
regards
Prasi
PS: O/p Terminal block can be replaced with a 3-pin block as the ground connections are common. Two wires from this ground connection will run separately to the amplifier left and right signal return connection .
Ofcourse, I will try to reduce the size of PCB and also number of jumpers in the coming days, this is just a concept layout that satisfies all the requirements of nano-watt hum that D. Joffe proposes.
Attachments
Last edited:
Hi prasi
I would still relook at the input return arrangement, I normally prefer to keep it almost exactly as per the schematic. You have the input return first going to the 'audio ground' and then going to various places like the pot etc.
In practice I would prefer this 'point' to be at the non-inverting input of the first amplifier. I agree with previous comments that trying to keep two separate grounds is going to be difficult, unnecessary and probably problematic. Seeing as how you have combined them at the output anyway, it may make more sense to keep them together all the way through the circuit.
If you persist with this kind of schematic though, you should keep the input return away from the audio ground until it 'gets' to the first half of the opamp.
I notice now that decoupling caps have moved very far away from the chip, if you have an issue with using these caps because of track pasta, try to use SMDs on the track side of the board. They work better than any TH component in this situation and are easier to route for.
Never use signal tracks to carry decoupling current. C18/20 are culprits, there may be more. These have basically placed a dead short across the lift resistors and rendered them useless. In general supply and decoupling is PGND and signal references - input, output and volume/balance etc - are AGND, and the two are never supposed to meet except at one well-defined point or through a lift resistor.
I'm terrible at tone controls and am not sure if that part of the board looks okay. Otherwise the tracks look a lot neater, less spaghetti, more sauce.
I would still relook at the input return arrangement, I normally prefer to keep it almost exactly as per the schematic. You have the input return first going to the 'audio ground' and then going to various places like the pot etc.
In practice I would prefer this 'point' to be at the non-inverting input of the first amplifier. I agree with previous comments that trying to keep two separate grounds is going to be difficult, unnecessary and probably problematic. Seeing as how you have combined them at the output anyway, it may make more sense to keep them together all the way through the circuit.
If you persist with this kind of schematic though, you should keep the input return away from the audio ground until it 'gets' to the first half of the opamp.
I notice now that decoupling caps have moved very far away from the chip, if you have an issue with using these caps because of track pasta, try to use SMDs on the track side of the board. They work better than any TH component in this situation and are easier to route for.
Never use signal tracks to carry decoupling current. C18/20 are culprits, there may be more. These have basically placed a dead short across the lift resistors and rendered them useless. In general supply and decoupling is PGND and signal references - input, output and volume/balance etc - are AGND, and the two are never supposed to meet except at one well-defined point or through a lift resistor.
I'm terrible at tone controls and am not sure if that part of the board looks okay. Otherwise the tracks look a lot neater, less spaghetti, more sauce.
Hi Sangram,
Lets try to look at the schematic from a different perspective. All reference points are connected to the "GNDAR" and "GNDAL" by the shortest route possible respectively for right and left channel. and hence the layout considers the same (except pin 3 and 5 of IC which can not be laid out separately without introducing additional jumpers. what say?
reg
prasi
PS Also modified output ground so that it connects directly and separately to the HBRR and HBRL junction.
Attachments
Last edited:
First, please look back to the last post and agree that the Star is the single point ground of the whole system. This is very important. There is no direct ground connection between preamp to IPS/VAS, and there is no direct ground connection from IPS/VAS to OPS. You need to be convinced on this before we can go any further.
Fig. 3 is the very simplified concept of the system, I am using only +V to draw the current loop, -V rail is the same. Point A and B in Fig.3 has noise due to the rectified ripples, also the current return from the speaker. So we should agree point A and B are both noisy with hum and signal current.
But since the ONLY ground reference is at the Star that connect to the earth ground. All the noise is dropped across the wire resistance R3. You can follow the RED loop on the right side and see that. Same analogy can be made for the left side where the current loop is in light BLUE. Again, since the Star is the only earth ground reference point, the noise voltage of the left side is across R2 at point A.
Notice, the +V at point C is noisy also because of R1 in series.
Now let's look at Fig. 4. This is a simplified circuit of one side of the signal chain from preamp to IPS/VAS, to OPS, to speaker. The whole signal chain is referenced to the GND of the Preamp as shown. GND is connected to the Star. If you look at the IPS/VAS and OPS, it is just an opamp with +ve and -ve inputs. The GNFB loop is formed by the 20K and 1K resistor to get gain of 21. BUT notice where the reference for the 1K resistor? It is referenced to the GND from the preamp. Any noise on the ground of the IPS/VAS becomes common mode signal at the power supply and is mostly rejected by the circuit. You can view the preamp is driving the power amp DIFFERENTIALLY and is explained in p380 section 18.4 in Cordell's book. He use a ground break resistor, but being in the same chassis here, you don't even need the resistor to connect between the ground of the preamp and the IPS/VAS.
Then if you look at the interface from IPS/VAS to OPS, it's only the + and - signal shown. There is no ground connection from IPS/VAS to OPS. The signal is a CURRENT signal. The noise on B in Fig.4 has no effect to the signal path. The noise on +V and -V are on the power supply and is mostly rejected. This is the very basic theory of how all the IC opamps reject the noise on the supply rails. You can count on at least -40dB rejection.
Notice in both preamp and IPS/VAS block, I put 22ohm on all the supply voltages and followed by something like a 1000uF filter caps to the quiet GND to the Star. The 22ohm limited the current and won't introduce much noise to the Star.
This is a COMPLETE system design, not just grounding. It is a combination of separating the ground, fixing the reference of the signal path to the very front from the preamp, preamp driving the power amp differentially and make use of the power supply rejection ability of the power opamp to attenuate the ripple noise. Then current limit the ripple with the 22ohm resistor in series with the supply rails for the IPS/VAS and preamp.
Attachments
Last edited:
I have to change the connector of the speaker return. I return them directly back to the Star instead of the ground of the big filter caps.
I modified the original diagram and Fig.3 and Fig.4 to reflect that.
The description of the last post still applies. Still the Star is the single point ground reference. Now there is a ripple loop that have the rectifier ripple and the signal loop that is the signal to the speaker. BUT you can see they still go through the Star back to the supply. This does not change that the fact there is only one point ground and it connected to the earth ground as the noiseless ground. No NET current will be flow from the system into the earth ground to create noise.
I added R6 and R7 on the speaker return path. This just create a slight resistance loss for the speaker signal no different from the speaker cable.
As you can see in Fig.3, the loops from the left side and right side are totally separated with only the battery source in common, they don't cross each other and therefore noise from one side will not affect the other side.
I modified the original diagram and Fig.3 and Fig.4 to reflect that.
The description of the last post still applies. Still the Star is the single point ground reference. Now there is a ripple loop that have the rectifier ripple and the signal loop that is the signal to the speaker. BUT you can see they still go through the Star back to the supply. This does not change that the fact there is only one point ground and it connected to the earth ground as the noiseless ground. No NET current will be flow from the system into the earth ground to create noise.
I added R6 and R7 on the speaker return path. This just create a slight resistance loss for the speaker signal no different from the speaker cable.
As you can see in Fig.3, the loops from the left side and right side are totally separated with only the battery source in common, they don't cross each other and therefore noise from one side will not affect the other side.
Attachments
Last edited:
Those connections are incorrect. They don't follow the schematic. You can see that the decoupling caps do not directly connect to either input or output ground, and the lift resistors are not placed between input and output ground.
'Reference' points carry very low current and are basically signal currents. Those would be the pot terminal 3, the balance terminal 2, and pins 3 and 5 of each opamp.
Your layout should attempt to connect these points together and to the output ground with the shortest possible path. Also respect the direction of the current. I would route to a signal ground reference only after the balance control wiper, instead of taking the signal ground to a central point and then pass in it to the V/B controls.
Decoupling currents need to be kept well away from this path, which is the way shown in the schematic.
It looks a lot better than before, I'll say that. A bit more elbow grease and you could have a good layout. Follow it up by trying to reduce crosstalk and thickening the supply tracks a bit. You want of aim for low parasitics.
Maybe you are just beginning to understand. Eventually you will move the charging pulse loop away from the star and then begin to have a quiet amp. In the meantime I am not going to nitpick my way through every wrong design you put up for comment. If you showed more humility I might be more willing to put in the effort.Alan0354 said:
I really don't think you get what I am trying to do. I spent a lot of time today drawing and typing since you asked me to explain. If you don't want to talk, don't ask. If you ask and I reply, this is your comment? What humility? I wasted all the time today, I bet you didn't read a word I wrote.
My scheme WILL work.
Hey there, I am still humble and eager to learn from anyone which is out there to give advice, be it Alan, Andrew, Mark, Sangram. Pl let us make it a forum for discussion and exchange of ideas (although in my case its receiving of ideas
)In continuing for my quest for a buzz free pre-amp amp combo, here is the layout and sch for the pre-amp, ( I am persistent!!).
Here I have separated the power gnd and audio gnd completely as per the layout shared by Mark.
Pay attention to the opamp in schematic, individual op-amps are now used for left audio signal and right audio signal. Hope its OK.
Attachments
Hi P, you are putting a power of work incorporating suggestions into your sch. Well done.
I saw your last sch and did not comment because I thought it was right.
Then Mark point out an error.
Now you have a new sch with Mark's change.
They look identical to me.
HBRR and HBRL are between the Signal Return and the Power/Decoupling Ground.
The signal current passes around the signal circuit and does not flow through HBRx
The speaker current passes around the speaker circuit and does not flow through HBRx.
Both sch seem to show the same. Both fit with Joffe's sch.
And Alan's does not match any of these !
I saw your last sch and did not comment because I thought it was right.
Then Mark point out an error.
Now you have a new sch with Mark's change.
They look identical to me.
HBRR and HBRL are between the Signal Return and the Power/Decoupling Ground.
The signal current passes around the signal circuit and does not flow through HBRx
The speaker current passes around the speaker circuit and does not flow through HBRx.
Both sch seem to show the same. Both fit with Joffe's sch.
And Alan's does not match any of these !
Hi Mark,
you are right on spot. The correct wiring is attached herewith.
For other things, I hope I have done everything right this time.
@ Andrew,
Thanks for your encouraging words.
regards
prasi
Attachments
Last edited:
Move R6 to the right of C15 so that you can use the same layout as R10/C16 and correct the text above X2.
The layout of IC1 looks better than IC2. Can you rotate IC2 and copy the IC1/R3 layout.
http://sound.westhost.com/pots.htm
Figure 9 looks interesting and should give you a better balance control.
The layout of IC1 looks better than IC2. Can you rotate IC2 and copy the IC1/R3 layout.
http://sound.westhost.com/pots.htm
Figure 9 looks interesting and should give you a better balance control.
Last edited:
Hi Mark,
Incorporated your suggestions.
The terminal block text "OUT-R/RGND/LGND/OUTL" remains same. did you mean something else to be corrected?
Balance control also improved as per ESP.
Attached schematic and layout. The board is ready for home etching. Min clearance of 24 mils is maintained everywhere -copper-dimension-pads-vias.
Regards
Prasi
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.