I wrote to Harry about this problem because I was not sure if it was my browser or his links. In the meantime I can open them in a new tab so try that. They are too useful to miss.
Best wishes
David
updated images
Sorry about the images. They are PDFs so you will only see them if your browser will render them "inline". You should be able to right-click to load in their own window or download, whichever browser you are using. For convenience I've reproduced the images below in PNG format:
e-amp layout as I understand it from written description:
current pathway from +ve reservoir cap:
current pathway from +ve decoupling cap:
current pathway for Zobel:
issue for order in which grounds are teed on PSU board:
An improved layout approach:
Sorry about the images. They are PDFs so you will only see them if your browser will render them "inline". You should be able to right-click to load in their own window or download, whichever browser you are using. For convenience I've reproduced the images below in PNG format:
e-amp layout as I understand it from written description:
An externally hosted image should be here but it was not working when we last tested it.
current pathway from +ve reservoir cap:
An externally hosted image should be here but it was not working when we last tested it.
current pathway from +ve decoupling cap:
An externally hosted image should be here but it was not working when we last tested it.
current pathway for Zobel:
An externally hosted image should be here but it was not working when we last tested it.
issue for order in which grounds are teed on PSU board:
An externally hosted image should be here but it was not working when we last tested it.
An improved layout approach:
An externally hosted image should be here but it was not working when we last tested it.
I add my thanks to Keantoken's.
The layout is more or less what I planned but your pictures are much better than my description.
One point on the position of the feedback resistors.
Your layout reduces the current in the feedback line so it will radiate less but the increased impedance will make it more vulnerable to pick up. Probably not a major issue but would it be better to put the resistors at the input end?
Best wishes
David
Thank you. That was my belief too so I was a bit surprised by that part of Harry's layout. As I wrote, probably not major, but the improvement is free so why not. Do you have any actual data on this from the PGP project or elsewhere?
Best wishes
David
>would it be better to put the resistors at the input end?
Yes!
Cheers,
E.
BTW, what about this question?A matter of 'splendid isolation'?
Why not design the PCB in such way that output is close to the - input and no need for long twisted screened pair. For Gain Card all philosophy is based on very short feddback path.
dado
Is it not better to connect cold side of the loudspeaker directly to the ground between big power caps and not to the PCB ground?
dado
I'm glad you asked this question! I was hoping this might stimulate debate; actually I have never put the feedback network at the output side as illustrated but it is an option. Putting the feedback network at the output does reduce current in the wires taking the feedback to the front end - this is good for two reasons: it radiates less and it reduces the error voltage generated in the ground reference at the input side relative to the ground reference of the loudspeaker connection. However, as you said it does mean this wire will now be more susceptible to pick up and this is why I drew a screened twisted pair where only one end of the screen is connected to ground.
As the errors produced when placing the feedback network at the input side are linear ones (error voltage generated is linear and any radiation from the connection is "linear") but pick-up is potentially non-linear, it may indeed be best to place the feedback network at the input side. But the use of a screened cable should minimise any pickup in the feedback connection if you put the feedback network near the output. Clearly there's a tradeoff here and it's another case where doing a controlled experiment comparing the two approaches would be beneficial. There is some discussion of these two approaches in Bob Cordell's book.
I used ISIS which is part of the Proteus schematic capture/PCB layout CAD package from Labcenter Electronics. Many of my colleagues are very rude about this suite but I actually quite like it; perhaps a case of Stockholm syndrome?
The images were exported from ISIS by printing to PDF with PrimoPDF; the PDFs where converted to PNG using Graphic Converter on my Mac.
This is certainly an option, but this runs the risk of placing sensitive nodes of the front-end close to traces that are carrying large currents and potentially radiating nasties which will be picked up. Using my layout approach above minimises this radiation but it's still a concern.
As shown in this image, this would create a high inductance pathway from PSU to amplifier board (note that the -ve current pathway has a similar large loop area):
An externally hosted image should be here but it was not working when we last tested it.
The only way to reduce the inductance of these loops would be to twist the loudspeaker return with the +ve/-ve supply links from PSU to Amplifier board. You would also have to take the ground reference of the feedback link from the PSU board, but the +ve side of this feeback link from the speaker output port on the amplifier board. This increases the loop area of the feedback connection, increasing inductance and possibility of radiation/pickup.
Last edited:
A Thread discussed this a year or so back.
I suggested that the screen should be connected to the output, as would be done with "guard ring" topology.
I always locate the feedback at the input and minimise the loop area of the feedback and input filtering components. Effectively I treat the NFB & the Signal Ground as an INPUT to the amplifier. Thus it needs protecting from external influences/interference. This arrangement would still suit a guard ring style of feedback wire from output to input.
Hi Damir,
It's a sound idea. Of course you can't put the output devices close to the input, but probably you mean to put the output speaker connection close to the input. Well, that's exactly what Edward Cherry did. see: EW+WW, Jan. 1995, p. 18.
Cheers,
E.
I did it with my TT amp, see post 804 http://www.diyaudio.com/forums/soli...one-seen-front-end-before-81.html#post3151755.
I would apreciate if Harry points out what is wrong with this layout, and I can make next one better.
dado
Hi Edmond,
Although this topology yield roughly 135db OLG , the baxandall pair
high output impedance will forcibly see its gain collapsing as soon as
it is loaded ,so much that dozens megohms input impedance is required
from the following current gain stages.
Moreover , the loading due to the compensation network will
drasticaly reduce OLG at high frequencies even if the current
gain stages have high enough input Z.
The output stage in the schematic in post 1 is modeled as having
0.97 voltage gain irrespective of frequency , wich yield flawed
simulations in respect of OLG.
Overall , for the few sims i did , the results are no better
than the same topology using a cascoded VAS once an output
stage is attached to the front ends.
Cheers,
Wb
I will need to get a piece of A3 this weekend and draw up the layout. Here are some interim comments.
In your improved layout you have twisted the rails together. The trade off here is noise coupling (half wave rectified signal) through the non conducting half of the amplifier or minimum loop area and I chose to avoid the noise coupling problem.
The speaker return runs in parallel with the hot connection in my amp, then branches off shortly before going onto the board off to the PSU.
Feedback resistors. These are placed right next to the input diff pair. I don't want noise getting in between the feedback junction and the inverting input. The signal current is low, so I feel running a track back to near the front end to the feedback resistors is ok and prefer this to the screened cable approach.
My Zobel return runs directly back to the PSU and is not connected to the main PSU ground as you show. This does create a loop, but I want to avoid HF currents other ths those driving the load in the main board ground.
In your improved layout you have twisted the rails together. The trade off here is noise coupling (half wave rectified signal) through the non conducting half of the amplifier or minimum loop area and I chose to avoid the noise coupling problem.
The speaker return runs in parallel with the hot connection in my amp, then branches off shortly before going onto the board off to the PSU.
Feedback resistors. These are placed right next to the input diff pair. I don't want noise getting in between the feedback junction and the inverting input. The signal current is low, so I feel running a track back to near the front end to the feedback resistors is ok and prefer this to the screened cable approach.
My Zobel return runs directly back to the PSU and is not connected to the main PSU ground as you show. This does create a loop, but I want to avoid HF currents other ths those driving the load in the main board ground.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.