: sorry
LTSpice simulation of my grounding scheme
I just input my grounding scheme into LTSpice to verify my grounding design works. Attached is the .acs file. From my testing, it works, but I only spend about an hour testing it, so if anyone is interested, they can play with the file.
1) V1 and V2 is used to simulate the transformer with center tap. I simulate to proof it worked as a CT secondary.
2) I split into L and R section with the main filter caps C1 and C2 for +V and -V of the R section resp. C3 and C4 for the L section. These are for the OPS.
3) Notice I use 10mohm resistor to simulate resistance of a few inches wire. Notice every ground from the Star has the 0.01ohm to simulate the original circuit I drew.
4) The ground is split into L and R preamp-GND, IPS/VAS-GND, main filter ground and speaker return all coming back to the Star as the single point ground.
5) I use generator V7 and V8 to simulate the preamps driving the power amps. They are reference to the L and R preamp-GND resp. to be consistent with my original design.
6) I use 1ohm R4 to connect the Star to the earth ground(the triangle). That's the ONLY point that connected to the ground. I use 1ohm to proof that it is not even important as there is no current going from the system to the earth ground. In real life, the resistance is lower.
I tried running the R with 2KHz 2V input to the power amp and 3KHz 2V to the L side and monitor the voltage across R33 and R72 which is 4ohm as dummy load. I verify the amps work. The rest of the test use this to get very large signal at the output to draw a lot of current from the supply rails to simulate the worst case. this is way way over 1/3 power test people use.
I probe the +V and -V to verify there are ripples just like the real thing.
Then I shut off one of the generator and look at the output to see any power supply ripple and the signal coupling from the other channel that still running at 40Vpeak( large signal). I did not observe any power supply ripple. I can see about 7uV peak to peak of feed through from the other side. I think that is mighty good. Make sure you read the voltage ACROSS the dummy load.
I probed the Star reference to the earth ground ( triangle symbol) and see some glitches of only in nV amplitude. This shows there is NO NOISE ON THE STAR.
Attached also is the two model txt files I used for convenience if you want to play with it. I have not tested for hours to verify, maybe you can spot some problem with it. I like to hear about it. So far, seems to works as designed.
I just input my grounding scheme into LTSpice to verify my grounding design works. Attached is the .acs file. From my testing, it works, but I only spend about an hour testing it, so if anyone is interested, they can play with the file.
1) V1 and V2 is used to simulate the transformer with center tap. I simulate to proof it worked as a CT secondary.
2) I split into L and R section with the main filter caps C1 and C2 for +V and -V of the R section resp. C3 and C4 for the L section. These are for the OPS.
3) Notice I use 10mohm resistor to simulate resistance of a few inches wire. Notice every ground from the Star has the 0.01ohm to simulate the original circuit I drew.
4) The ground is split into L and R preamp-GND, IPS/VAS-GND, main filter ground and speaker return all coming back to the Star as the single point ground.
5) I use generator V7 and V8 to simulate the preamps driving the power amps. They are reference to the L and R preamp-GND resp. to be consistent with my original design.
6) I use 1ohm R4 to connect the Star to the earth ground(the triangle). That's the ONLY point that connected to the ground. I use 1ohm to proof that it is not even important as there is no current going from the system to the earth ground. In real life, the resistance is lower.
I tried running the R with 2KHz 2V input to the power amp and 3KHz 2V to the L side and monitor the voltage across R33 and R72 which is 4ohm as dummy load. I verify the amps work. The rest of the test use this to get very large signal at the output to draw a lot of current from the supply rails to simulate the worst case. this is way way over 1/3 power test people use.
I probe the +V and -V to verify there are ripples just like the real thing.
Then I shut off one of the generator and look at the output to see any power supply ripple and the signal coupling from the other channel that still running at 40Vpeak( large signal). I did not observe any power supply ripple. I can see about 7uV peak to peak of feed through from the other side. I think that is mighty good. Make sure you read the voltage ACROSS the dummy load.
I probed the Star reference to the earth ground ( triangle symbol) and see some glitches of only in nV amplitude. This shows there is NO NOISE ON THE STAR.
Attached also is the two model txt files I used for convenience if you want to play with it. I have not tested for hours to verify, maybe you can spot some problem with it. I like to hear about it. So far, seems to works as designed.
Last edited:
This problem can easily be solved by moving the ground reference to the source as covered in Cordell's book also. Just look at how I drive the IPS/VAS pseudo differentially by the preamp, by referencing the 1K R18 to the preamp ground. You can easily move the reference to the source ground using the exact same technique to get rid of this problem. The simulation file is getting too big, and OP is asking about problem between preamp and power amp. That's why I took the short cut and assume the interface between the source and preamp is taken care of already.
BTW, Your test is really not a grounding issue, this is actually a COMMON MODE REJECTION issue of the power amp ( opamp). You are driving the preamp ground that has the generator to the input also. This particular IPS/VAS design does not have good common mode rejection as you can see R7 at the collector of the LTP is the main problem. You want much better performance, you need to have CCS as the collector load for Q4. This is covered in Cordell's book and other IC opamp design books. I just pull up an old circuit I simulated before and modified it for this.
Attached is the file with the generator modified to 1.5V noise.
you generator is only set to 0.01V amplitude. You have to set the voltage on the generator side also. I went and change both to 1.5V and I observed 17mV at the output. The is rejection ratio of 0.017/1.5=0.0133 or -39dB or COMMON MODE rejection ratio of 38dB.
It is a mistake to go round and round just looking at the grounding only, I don't want to get into debate grounding scheme suggested here. This is a SYSTEM design, not just grounding. I started out with pseudo differential interface between preamp and power amp to make noise common mode. Then isolate the ground. It is a complete design.
Did you try injecting noise on the other designs suggested here?
Last edited:
I don't want to comment on OP's preamp board before because many people helping him already, it would be too confusing for OP.
But look at post #82. I assume that is the preamp where R-GND at X1-2 is the source ground from the source. Let's look at what if you inject a 1.5V 60Hz noise source just like you injected into my design. This will be common mode that will be in R-IN at X1-1 because the signal from the source is REFERENCED to the source ground. So the 1.5V noise is on both X1-1 and X1-2. There is no other way around this. Now just look at what happens.
The +ve input of IC1A and IC1B are all referenced to the R-GND at X1-2 which is good. This is what I would do. The output of IC1B should have NO NOISE REFERENCED TO X1-2 R-GND. BUT the volume potentiometer R1 has the bottom tied to the PREAMP GND referenced to GND4 to X2-2. I assume this is the preamp ground. This is the ground that you put the noise source between X1-2 and X2-2.
This is where the problem starts. You have the output of the IC1B referenced to X1-2, but the volume pot ground is referenced to GND4. This will cause a 1.5V 60Hz noise at the output of IC1B referenced to the GND4 and X2-2 ground.
You see where the problem is, at least my design without even doing anything will attenuate 38dB to 17mV at the output of the speaker terminal. You have 1.5V noise input to the power amp and to be amplified by the power amp. I don't care what you do to the ground inside the amp, the preamp in post #82 is going to have a big problem with your generator injecting noise between X1-2 and X2-2.
Noise is a system issue, not just grounding. The fact that you have separate modules in the system, you are going to have grounding problem, all the grounding in the world can only get down to a certain point. That's the reason people use DIFFERENTIAL transceivers for interfacing. What I am doing and what Cordell suggestion is only half way "pseudo differential" only. You want to nitpick to this point with a generator in between grounds, you need to go full differential receiver......which is absolutely possible. Just two more resistor.
Last edited:
The editing time has past, but this is what I would do if I have to deal with so much ground hum. See attached file. This is a simplified drawing of the preamp from post #82. I go full differential to get rid of the noise right at the input using full differential amp. The whole preamp pcb is going to be one ground only connected to X2-2. They use what I do on the interface between preamp and IPS/VAS. Keep it simple. The power rails for the preamp board are all referenced to X2-2, nothing to do with X1-2. The noise on X1-1 and X1-2 ENDED at the input of IC1A.
I include the noise generator like Mark suggested. But if you simulate, you have to induce the noise to X1-1 as this is from the source. Noise has to be common mode on both X1-1 and X1-2.
I would use a solid ground plane on this preamp board, forget all the star on the pcb, the more you mug around, the more trouble you have. Look at noise as a system issue, not just grounding.
I include the noise generator like Mark suggested. But if you simulate, you have to induce the noise to X1-1 as this is from the source. Noise has to be common mode on both X1-1 and X1-2.
I would use a solid ground plane on this preamp board, forget all the star on the pcb, the more you mug around, the more trouble you have. Look at noise as a system issue, not just grounding.
Attachments
Last edited:
I am seeing a 1.5 V signal at the right output and I am convinced that with a more correct simulation of the leads between the source and power amp the output signal would increase. Although the 70 A current through the interconnect is probably a bit high.
This with the sim from post #85
This with the sim from post #85
Last edited:
I did the simulation with your file already, it's 17mV at the output. It is important for you to read the voltage ACROSS THE DUMMY LOAD, not from the power amp out to the earth ground. Of cause if you read from the output to earth ground, you see 1.5V!!! the dummy load return is NOT the earth ground. If you don't know how to read voltage across the dummy load, I can show you.
Also read my post I just posted, In my simulation I don't deal with the source as OP's problem is between preamp and power amp. I just gave suggestion in dealing with the source in response to the possibility that you have so much ground noise between the source and integrated amp by using full differential input to get rid of the noise once and for all.
Last edited:
No, I triple checked your file, it is 17mV.
When you want to read the voltage across the R33, you first hover on the top connection with the mouse pointer until you see the RED probe appears. Then you left hold the mouse button, move the pointer towards the lower connection of the R33. You'll see the probe turn from RED to GREY. When you touch the lower wire of the R33, you'll see the probe turns BLACK. Release the left mouse button and you'll see the wave form of the voltage across the R33 only.
I just learned this on Yahoo LTSpice forum. You can read through the thread:https://groups.yahoo.com/neo/groups/LTspice/conversations/messages/87315
Ha ha, I just learn this as I go!!! You should verify I am doing it right!!! You make me question my measurement!!!
EDIT: I verify in another way. If you move the mouse pointer on top of R33, the little bird shows up to read current. You click that and the graph will show the current waveform. It showed 4.2mA peak. So with 4ohm, gives you about 17mV of noise.
At this point, I am sure I am right.
Last edited:
We must be looking at different files. I attached it again so you can test it.
I did it two ways to verify it. I read the voltage across R33, I read the current through R33, all gave me 17mV. I don't know what to tell you.
I even verify by setting V7 back to 2Vpeak and I got output of 10A peak using the little owl, AND I can read 40V peak when I do the differential probing.
You really need to double check your reading method.
It is important to note that the Star is bouncing 1.5V!!! That's what you are driving it. The whole preamp and power amp are all bouncing up and down 1.5V!!! BUT we are only interested in what comes out of the speakers.......which is across the R33.
The whole speaker is also bouncing 1.5V!!!! but you should not hear it as both terminals of the speaker is bouncing 1.5V, not across the two terminals.
I did it two ways to verify it. I read the voltage across R33, I read the current through R33, all gave me 17mV. I don't know what to tell you.
I even verify by setting V7 back to 2Vpeak and I got output of 10A peak using the little owl, AND I can read 40V peak when I do the differential probing.
You really need to double check your reading method.
It is important to note that the Star is bouncing 1.5V!!! That's what you are driving it. The whole preamp and power amp are all bouncing up and down 1.5V!!! BUT we are only interested in what comes out of the speakers.......which is across the R33.
The whole speaker is also bouncing 1.5V!!!! but you should not hear it as both terminals of the speaker is bouncing 1.5V, not across the two terminals.
Last edited:
We are using the same method to measure the voltage over R33. Only I had increased the input loop voltage. Also the phase between the two ground loops will alter the result.
The amp first measured zero noise but with ground loops does not.
A better design needs to apply controlling the ground loop currents and voltages.
What I also see as a problem is the long leads/traces used. Will the high inductance and large loop areas not cause problems. This is also not a part of the simulation.
The amp first measured zero noise but with ground loops does not.
A better design needs to apply controlling the ground loop currents and voltages.
What I also see as a problem is the long leads/traces used. Will the high inductance and large loop areas not cause problems. This is also not a part of the simulation.
who said I am using long leads? The components are so big it's unavoidable. I don't intentionally make it longer, I put 0.01ohm to make it the worst case.
I just went and simulated with the noise generator at 1.5V, at 0deg, 60, 180 and 270 deg phase. The result is EXACTLY the same. 17mV!!!
I don't follow you. You have to show me your .asc file how you get 1.5V noise. I can assure you I am right at this time how I measure on the original file you posted. You really need to going back and check you measurement. We are going no where now. There are only so many ways to spin this. Post the file again, label EXACTLY where you measure the voltage and how you measure it.
You cannot keep increase the noise generator beyond 1.5V. that's a lot of noise if you are supposed to plug the source and the amp in the same wall plug or even in the same house. Be real.
One thing is you are putting a lot of noise between the source ground and the amp ground. If you do it on the one in post #82, it is a whole lot worst, you'll have 1.5 X the gain of the power amp which is usually 20, that will give you 30V noise!!!
Last edited:
I am also measuring 17mv with a source ground loop of 0.0001 V. What I had done was to increase this voltage and then saved the file.
I do not know if the source ground loop of 0.0001 V is correct. What it does demonstrate is that the output is not 0 V and that some of the claims made in post #83 are not correct.
I do not know if the source ground loop of 0.0001 V is correct. What it does demonstrate is that the output is not 0 V and that some of the claims made in post #83 are not correct.
I don't know what is source loop. If you said there is a problem post the .asc file so I can look at what you are talking about. We are going no where at this point. You claim there is a problem in the simulation, then show me the .asc file, why are we still doing "yes it is, no it's not!!!"?
I did not claim anything in post 83 other than showing what I did to model the circuit and the result to let other people challenge it. You did challenge it, the more challenge the better. If you think my model on the system in the simulation is wrong, then tell me what is wrong.
Seems like people reading Self's APAD6 in page 600 Fig.25.1 that you want to the reservoir cap to be very close to the CT of the transformer to confine the large ripple current in D-E. Therefore anything else is WRONG!!! I cannot do that if I want separate banks of reservoir caps. There is a school people claimed the best is to separate the reservoir caps to L and R side. This is what I am doing. You cannot do what Self did if you want to separate them UNLESS you use separate secondaries or two separate power transformers. I am trying to show you don't have to, and you can let current flow in the Star as long as you make it a true one point ground.
Of cause, if you go back to the single bank of reservoir caps, then it's just common sense to do it like Self. If fact, I don't even think separate E-F is that important as the reference of F is from A and it's irrelevant whether E-F is a line or a point.
Last edited:
I made no comments or challenged the amp or PSU design and not a fan of Self.
All I did was join the source grounds together in you sim and created a little Hum by adding ground loop voltages. Mostly this is the cause of what we hear.
Your simulation is to perfect, it is missing component and layout parasitics.
All I did was join the source grounds together in you sim and created a little Hum by adding ground loop voltages. Mostly this is the cause of what we hear.
Your simulation is to perfect, it is missing component and layout parasitics.
I thought you said you see 1.5V of noise. that's the reason I kept asking.
I think I cover all the parasitic by adding wire resistance, you added the inductors. Tell me what am I missing that make it work so well?
The pcb has to be layout good, this is not the subject here. This wiring is the main subject and I added the pseudo differential at the input. The simulation is exactly the same as my original design drawing.
As I said, try simulate with your 1.5V generator in the other grounding scheme you guys talked about. I bet you'll see worst as I explained in post #86 already.
In fact, I make the resistance larger than in real life already. As I said, the 17mV of noise is because of the common mode rejection deficiency of the IPS that use a resistor load. If it is changed to CCS, you'll see a big reduction of the noise.
Also, If I lower the resistor between the Star and earth ground from 1ohm( I intentionally make it so high to show it does not matter) to 0.01ohm, and then you add wire resistance to your 1.5V generator to simulate real life situation, the noise will be much much lower.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.