Hi Jens,
I'm trying to find the NFB trace without the schematic.
Can you direct me to it's ends?
I'm trying to find the NFB trace without the schematic.
Can you direct me to it's ends?
Jens,
The negative feedback path is not clear, but if it's directed like I think it is, it might be better changed. See section 5.7 of this URL:
http://www.dself.dsl.pipex.com/ampins/dipa/dipa.htm#5
I think the ground line should have a separate path from the output back to the feedback resistors.
I also can't tell if the power supply grounds and the signal grounds are separated. They should be decoupled as shown in Section 5.5 of the above URL.
The negative feedback path is not clear, but if it's directed like I think it is, it might be better changed. See section 5.7 of this URL:
http://www.dself.dsl.pipex.com/ampins/dipa/dipa.htm#5
I think the ground line should have a separate path from the output back to the feedback resistors.
I also can't tell if the power supply grounds and the signal grounds are separated. They should be decoupled as shown in Section 5.5 of the above URL.
Hi Jens,
the section Pooge is referring you to is 5.5 to 5.7 (very near the end of the doc.).
my pdf download is working again, maybe an ISP problem.
the section Pooge is referring you to is 5.5 to 5.7 (very near the end of the doc.).
my pdf download is working again, maybe an ISP problem.
"something odd happening when I try to download the schematic.
the PDF starts to download (dial up) and then after about 20 seconds locks my computer. When I (ctrl+alt+del) & end the download the computer closes the internet explorer as well??"
I always right click and save to my desktop pdf files so the server doesn't time out on me.
the PDF starts to download (dial up) and then after about 20 seconds locks my computer. When I (ctrl+alt+del) & end the download the computer closes the internet explorer as well??"
I always right click and save to my desktop pdf files so the server doesn't time out on me.
Hi Jens,
1. You must change the NFB trace coming back to R13 labelled "out" on PDF sheet1/4. This label should change to "out-2" and the trace must connect straight to the output. If you can provide an alternative tapping point for the NFB return do so between the Thiel cap & resistor R57/C34 (as Dr. Cherry). A wire link here would do just fine.
2. The DC input bypasses the RF filter. Disconnect trace from R11/C8 and connect to C6 & C7/R11.
3. The ground returns for D2 & D3 are shown as the same symbol as the dirty grounds for the decoupling. Rubbish on the dirty ground will modulate the cascode reference.
I think these two grounds should be to clean ground, star connected to;- input grnd, C8, R17, C9, C13, C12 and all these should have one separate wire to the main ground off board and not shared with the dirty ground to the main common ground. Comments on this last point please.
1. You must change the NFB trace coming back to R13 labelled "out" on PDF sheet1/4. This label should change to "out-2" and the trace must connect straight to the output. If you can provide an alternative tapping point for the NFB return do so between the Thiel cap & resistor R57/C34 (as Dr. Cherry). A wire link here would do just fine.
2. The DC input bypasses the RF filter. Disconnect trace from R11/C8 and connect to C6 & C7/R11.
3. The ground returns for D2 & D3 are shown as the same symbol as the dirty grounds for the decoupling. Rubbish on the dirty ground will modulate the cascode reference.
I think these two grounds should be to clean ground, star connected to;- input grnd, C8, R17, C9, C13, C12 and all these should have one separate wire to the main ground off board and not shared with the dirty ground to the main common ground. Comments on this last point please.
1) The FB is taken at the common section where the emitter resistors connect. The node is called "out"
2) Done
3) The diodes refer to gnd as the original.
The input and FB sections refer to gnd through a 10 ohm resistor
I removed the caps C36-C39 and C40-C43 after re reading the leach pages.
\Jens
2) Done
3) The diodes refer to gnd as the original.
The input and FB sections refer to gnd through a 10 ohm resistor
I removed the caps C36-C39 and C40-C43 after re reading the leach pages.
\Jens
Hi Jens,
I am not sure if we are talking about the same thing in point 1.
The feedback MUST come from the end that feeds the speaker output. It must NOT come off the common between the emitter resistors.
Re 3. is there a separate ground for decoupling from the signal ground? Most papers published demand that dirty and clean grounds are separate until the star ground point.
Have you achieved that?
I am not sure if we are talking about the same thing in point 1.
The feedback MUST come from the end that feeds the speaker output. It must NOT come off the common between the emitter resistors.
Re 3. is there a separate ground for decoupling from the signal ground? Most papers published demand that dirty and clean grounds are separate until the star ground point.
Have you achieved that?
AndrewT said:
The FB is always taken before the output inductor (on the inside of it) and NOT on the speaker terminals.
I use a gnd plane.... different thing that the star gnd theory.
\Jens
Hi Jens,
Yes, your right just before the inductor.
Not sure about the ground plane in lieu of star earthing and separate dirty & clean grounds.
Any comment anyone?
Yes, your right just before the inductor.
Not sure about the ground plane in lieu of star earthing and separate dirty & clean grounds.
Any comment anyone?
This is interesting.
Any issues i may encounter if i connect the speaker GND to the mains power GND (kinda partial star) instead of the out GND ? This way the out power won't flow through the ground plane (except the power required to charge/discharge the 2 10K electros).
Oh, it just
me
What happens if you have 2 channels connected to the same dc supply and the input GNDs get connected somewhere in the signal source ?
Any possibilities of mains or parasitic injection ?
I may have spoken garbage
Any issues i may encounter if i connect the speaker GND to the mains power GND (kinda partial star) instead of the out GND ? This way the out power won't flow through the ground plane (except the power required to charge/discharge the 2 10K electros).
Oh, it just
meWhat happens if you have 2 channels connected to the same dc supply and the input GNDs get connected somewhere in the signal source ?
Any possibilities of mains or parasitic injection ?
I may have spoken garbage
Hi Jondoe,
yes, connect the speaker return off board but not direct to either the mains safety earth and not to the PSU capacitor common.
Create a master star ground(MSG) next to the PSU common and take all your grounds here (both channels) and also connect your PSU common here. One exception;- the input RCA return should probably go to the Input ground on the driver(VAS) board and from there to the MSG. Each of your driver(VAS) boards should have a clean AND a separate dirty ground wire to the MSG
When you interconnect the clean and mains grounds with R4 you are isolating the source equipment mains grounds from the power amp MSG. This should ensure no parasitics feeding through.
The purpose of keeping the dirty grounds separate is to reduce the risk of half rectified modulation from the dirty(power & decoupling) ground contaminating the clean ground and then feeding into the input along with the signal.
yes, connect the speaker return off board but not direct to either the mains safety earth and not to the PSU capacitor common.
Create a master star ground(MSG) next to the PSU common and take all your grounds here (both channels) and also connect your PSU common here. One exception;- the input RCA return should probably go to the Input ground on the driver(VAS) board and from there to the MSG. Each of your driver(VAS) boards should have a clean AND a separate dirty ground wire to the MSG
When you interconnect the clean and mains grounds with R4 you are isolating the source equipment mains grounds from the power amp MSG. This should ensure no parasitics feeding through.
The purpose of keeping the dirty grounds separate is to reduce the risk of half rectified modulation from the dirty(power & decoupling) ground contaminating the clean ground and then feeding into the input along with the signal.
JensRasmussen said:
Jens, I don't think you are getting the picture, here. Please read the Self Url regarding this. You shouldn't have the FB path coursing through all those high current emitter resitor areas. Take it directly, via its own path, from the inductor to the feedback resistors.
JensRasmussen said:
I don't get your point, here. A ground plane is fine on the board. You should have a separate power supply ground and signal ground path from the board to the star ground. That way, the ripple currents won't be coursing through the signal ground, causing voltage variation in your signal ground. The wires from the board don't care whether the ripple came from a ground plane or not. You don't want I2R losses due to ripple currents in your signal ground.
Hi all,
Pooge, well put.
I already have a design modification in my head for a hardwired NFB wire on the back of the board and a nest of wires again on the back, interconnecting all the decoupling returns back to master star ground.
It will look terrible but mostly hidden when installed.
Shame we cannot get it right at manufacturing stage. The opportunity was there but deadlines are approaching.
Pooge, well put.
I already have a design modification in my head for a hardwired NFB wire on the back of the board and a nest of wires again on the back, interconnecting all the decoupling returns back to master star ground.
It will look terrible but mostly hidden when installed.
Shame we cannot get it right at manufacturing stage. The opportunity was there but deadlines are approaching.
I have just been reading through Self’s k Chapter 6. The extra distortion from a “wrong” NFB take off point is in the 0.002% range. If the FB point is moved, I would have to cut the ground plane in half, effectively reducing its value. I think it is better to keep the NFB path the way it is, because it is not a practical layout solution to move it as it would require a third layer in order to get the NFB signal take from the outmost point of the out node. Andrew you already suggested this when you talked about air wire.
I have added an input ground connector on the board; this can be used to connect the input ground to central ground somewhere off the board. I’ll post data ASAP. I am still convinced that the ground connection to the 20V zener diodes should be the ground plane and not the small signal ground.
AndrewT, I think you should try the design before making a lot of modifications, otherwise you will never know if it actually worked or not. Before we decide how we could “get it right”, may I suggest that you show how you would implement the ground and feedback solution on a two layer PCB?
The PCB is the third generation of layout, if any of you think you have a better way of implementing the design, please step up and offer your layout to the public. This layout is just the way I wanted to do it. I will not make changes to the layout that compromises my own layout plan in any way. The design is offered the way it is, if you don’t like it, I suggest you make your own or start a redesign thread where the design is reworked. I will be happy to participate in any redesign but as one of my first post of this thread says, “this is the reason why I was a bit reluctant in releasing the design for a group buy in the first place”
I will get boards made for my self with the layout pretty much the way it is now, with or without the group buy.
\Jens
I have added an input ground connector on the board; this can be used to connect the input ground to central ground somewhere off the board. I’ll post data ASAP. I am still convinced that the ground connection to the 20V zener diodes should be the ground plane and not the small signal ground.
AndrewT, I think you should try the design before making a lot of modifications, otherwise you will never know if it actually worked or not. Before we decide how we could “get it right”, may I suggest that you show how you would implement the ground and feedback solution on a two layer PCB?
The PCB is the third generation of layout, if any of you think you have a better way of implementing the design, please step up and offer your layout to the public. This layout is just the way I wanted to do it. I will not make changes to the layout that compromises my own layout plan in any way. The design is offered the way it is, if you don’t like it, I suggest you make your own or start a redesign thread where the design is reworked. I will be happy to participate in any redesign but as one of my first post of this thread says, “this is the reason why I was a bit reluctant in releasing the design for a group buy in the first place”
I will get boards made for my self with the layout pretty much the way it is now, with or without the group buy.
\Jens
Hi Jens,
thanks for the reply.
I think you could achieve most goals without splitting the ground plane and without going 3 layers.
Consider the NFB route;- from inductor under D15 out to top edge of board along to vicinity of C21/F1/Vcc route down towards C16 then use R13 to jump across bottom copper tracks. A number of options for that final part of the route to keep your board as neat as it is at present.
You can run the zeners off the ground plane IF you take the decoupling out on a separate grounding system(Power ground). Then the ground plane becomes the signal ground with only clean signals on it.
I have not examined the difficulty in separating the decoupling from the ground plane. Having seen your skills in board layout I am absolutely sure you can do a better job than me and probably most others in the industry, but do you have the time?
Considering finally your extraction of the quote only "0.002% distortion" this was for a short 6mm lenngth of track between the emitter resistors and it showed a 3 fold increase in distortion when taken from a power resistor wire about 10mm long. In your layout the shared track length is about 70mm or 80mm.
Please try to make your time available to finalise this final part of what has been a very successful redesign. I believe the benefits would be worthwhile.
thanks for the reply.
I think you could achieve most goals without splitting the ground plane and without going 3 layers.
Consider the NFB route;- from inductor under D15 out to top edge of board along to vicinity of C21/F1/Vcc route down towards C16 then use R13 to jump across bottom copper tracks. A number of options for that final part of the route to keep your board as neat as it is at present.
You can run the zeners off the ground plane IF you take the decoupling out on a separate grounding system(Power ground). Then the ground plane becomes the signal ground with only clean signals on it.
I have not examined the difficulty in separating the decoupling from the ground plane. Having seen your skills in board layout I am absolutely sure you can do a better job than me and probably most others in the industry, but do you have the time?
Considering finally your extraction of the quote only "0.002% distortion" this was for a short 6mm lenngth of track between the emitter resistors and it showed a 3 fold increase in distortion when taken from a power resistor wire about 10mm long. In your layout the shared track length is about 70mm or 80mm.
Please try to make your time available to finalise this final part of what has been a very successful redesign. I believe the benefits would be worthwhile.
The way you propose to make the FB will create a loop that will pick up noise. The shared out node is indeed about 80-100mm long, but it is 10 mm wide and 110µm thick.
I still think it is better the way it is, and as the layout responsibility will be mine I will not make changes that incorporates unknown into the layout. If big changes are made I suggest a prototype run to ensure that the builders will get boards that work.
Are you talking about the decoupling of the zeners? Or the main caps?
\Jens
I still think it is better the way it is, and as the layout responsibility will be mine I will not make changes that incorporates unknown into the layout. If big changes are made I suggest a prototype run to ensure that the builders will get boards that work.
Are you talking about the decoupling of the zeners? Or the main caps?
\Jens
Leach design - PS decoupling
Hi Jens,
I went back to Leach V4.5 and found the section on Power Supply Decoupling and Grounding.
Para 4 states "The circuit board has two ground leads, both of which connect to the central power supply ground. One lead grounds the signal reference points for the diff amp input stage. The other grounds the power supply decoupling capacitors and provides a ground reference for the protection circuit. R51 (R4) connects the two ground leads together on the circuit board. This resistor is small enough to look like a signal short circuit between the two grounds but large enough to force the currents to flow to the central ground through the separate wires. This helps prevent hum induced by power supply ripple currents ( Pooges IsqrdR ) in the ground system."
This confirms the separate clean and dirty grounds and also confirms the reference voltage for the cascode should refer to the signal ground (ground plane).
I found Self's graph of distortion induced by decoupling connection error. Increases distortion throughout the frequency range but worst at low frequencies. eg. 0.01% to 0.02 % @ 20kHz, 0.001% to 0.018% @ 1kHz, 0.0006% to 0.12% @20Hz
conclusion distortion increases of between 6db and 46db in the audible spectrum due to incorrect connection of decoupling.
Hi Jens,
I went back to Leach V4.5 and found the section on Power Supply Decoupling and Grounding.
Para 4 states "The circuit board has two ground leads, both of which connect to the central power supply ground. One lead grounds the signal reference points for the diff amp input stage. The other grounds the power supply decoupling capacitors and provides a ground reference for the protection circuit. R51 (R4) connects the two ground leads together on the circuit board. This resistor is small enough to look like a signal short circuit between the two grounds but large enough to force the currents to flow to the central ground through the separate wires. This helps prevent hum induced by power supply ripple currents ( Pooges IsqrdR ) in the ground system."
This confirms the separate clean and dirty grounds and also confirms the reference voltage for the cascode should refer to the signal ground (ground plane).
I found Self's graph of distortion induced by decoupling connection error. Increases distortion throughout the frequency range but worst at low frequencies. eg. 0.01% to 0.02 % @ 20kHz, 0.001% to 0.018% @ 1kHz, 0.0006% to 0.12% @20Hz
conclusion distortion increases of between 6db and 46db in the audible spectrum due to incorrect connection of decoupling.
- Status
- Not open for further replies.