I am with Andrew and Edmond on this one. Any radiation is very small, current about a factor of a thousand less than the output traces, linear anyway and any errors it induces will be reduced enormously by the feedback. Whereas any pick-up by the feedback line is not corrected at all. This would seem to dominate by about 40 dB.
I asked Edmond for data because it is nice to check, but I can't see how such an uneven trade-off could be tipped by your second point about error in the earth reference. I suspect that error is solvable with careful earth layout, will think about this.
Best wishes
David
Last edited:
Well, the signal currents are in the mA range (typically) so the radiated magnetic fields are small, and the trace between the output pick-off point the the top feedback resistor HOT side is a few CM's. My conclucion is that if you locate the FB resistors close to the inverting input and do not use screened cable
1. magnetic induction into surrounding circuitry will be low (low signal current flowing into resistor)
2. There may be some capacitive coupling, but given the fact that most circuit nodes in a power amp are running mA and the components quite far apart, this is unlikely to be as big a problem as one would assume (either into the f/back circuit, or from the feedback resistors out to other circuits)
3. There are no very fast rising edges in a typical music signal, again, capacitive coupling minimal.
So, I place my feedback resistors close to the inverting input and just run a track back to the output pick-off point. No need for any screened cable.
1. magnetic induction into surrounding circuitry will be low (low signal current flowing into resistor)
2. There may be some capacitive coupling, but given the fact that most circuit nodes in a power amp are running mA and the components quite far apart, this is unlikely to be as big a problem as one would assume (either into the f/back circuit, or from the feedback resistors out to other circuits)
3. There are no very fast rising edges in a typical music signal, again, capacitive coupling minimal.
So, I place my feedback resistors close to the inverting input and just run a track back to the output pick-off point. No need for any screened cable.
The point of twisting the supply rails together is to minimise radiation and to cancel the harmonics. When the positive rail conducts this "induces" harmonics in the negative rail, but when the negative rail conducts this produces harmonics in anti-phase with the induced harmonics; the cancellation isn't perfect because the cables don't physically occupy the same space. The only way to stop the rail connections radiating is to twist them together with the return path; your approach means that the rail connections could radiate the half-wave nasties to sensitive areas of the front-end of the amplifier.
Yes, I should have drawn the twisted pair connection from the speaker going further; however, this still means there is a conductor that carries the return current of the loudspeaker without being next to its "hot" connection. This inevitably leads to larger loop area, higher inductance and more radiation.
So the connection goes directly to the PSU but isn't connected to the PSU
I can't see how what I've drawn is different from what is shown in your e-amp write up, page 13 and page 15: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.
Last edited:
Hi Dado,
You may have spotted earlier that I mentioned that layout issues hurt my brain more than I would like. Figuring out if you have a problem is usually not hard*; the head-hurty bit is trying to figure out how to solve the problem! The bad news is that with your TT amp layout, I have done (some of) the easy bit in that I can confirm there are issues with it. On the image below, I have traced the current pathway for current from the +ve decoupling capacitor. Note how the loop area is large and that the loop partly encloses the front-end of the amplifier; not good!
An externally hosted image should be here but it was not working when we last tested it.
Now you've got to do the hard bit, which is figure out how to make those issues go away. Have another look at my "improved layout" image which hopefully will help; start by bringing the power connection in at the horizontal midpoint between the PNP and NPN devices at the top of your board, together with the ground.
* To see if there are problems, trace out the current loops for tracks that will be carrying large currents. Remember when you do this that current always travels in complete loops; it can't just suddenly stop or start at a point (on the current path illustrated above, the loop is completed via the loudspeaker). Also, identify any tracks on the circuit that connect to a high-impedance node and try to keep these as small/short as possible, preferably protected by a reference plane or track nearby.
Harry,
your pic of the layout/schematic in post865 shows multiple tappings to feed the V+ and similar for V-.
If one considers that each of the links from cap to cap has some inductance then removing all the links to V+ from caps 1, 2, 3 & 4 (reading from the left) then the link inductance helps attenuate the interference/pulsing coming in from the left side.
I consider that by only connecting V+ to the 5th cap pin the output is cleaner than using the layout as you show.
Yes the supply V+ will be slightly slower, but that slowness is tiny in comparison to the slowing introduced by the final output cable feeding the amplifier.
All of the combined "slowness" of the single V+ tapping and of the final output cable is overcome by using appropriate MF & HF decoupling on the amplifier PCB.
Slowness is the inability to respond to to fast transient demands for current.
your pic of the layout/schematic in post865 shows multiple tappings to feed the V+ and similar for V-.
If one considers that each of the links from cap to cap has some inductance then removing all the links to V+ from caps 1, 2, 3 & 4 (reading from the left) then the link inductance helps attenuate the interference/pulsing coming in from the left side.
I consider that by only connecting V+ to the 5th cap pin the output is cleaner than using the layout as you show.
Yes the supply V+ will be slightly slower, but that slowness is tiny in comparison to the slowing introduced by the final output cable feeding the amplifier.
All of the combined "slowness" of the single V+ tapping and of the final output cable is overcome by using appropriate MF & HF decoupling on the amplifier PCB.
Slowness is the inability to respond to to fast transient demands for current.
Last edited:
In a high-performance amp there ill alays be nodes of high impedance. The usual technique is to use indefinite or infinitesimal impedances to increase gain. Therefore even slight capacitive coupling from the output in the rong place ill certainly cause oscillation, and perhaps anomalous behavior that cannot be addressed once the PCB is made. So, a guard ring, or perhaps guard rings around high-impedance places passing by the feedback trace, could be a good investment. In any case, this is prototyping. e can disable the guard trace any time and then e'll kno for sure hether it is helpful.
* my key does not ork. Unless I press und very quickly. unwd
* my key does not ork. Unless I press und very quickly. unwd
Dadod,
see the little bit of unused space between the leads of the two middle output devices.
Separate the output devices slightly, maybe an extra 5mm.
Now use that space to insert: the +ve supply, the -ve supply, the Power Ground and the Decoupling Ground (particularly the HF decoupling).
If possible the Zobel ground should also come into that area.
This brings all the high current routes to meet in that one small area and also helps reduce trace lengths.
Move the R//L off the PCB.
This operates just as well when located in the cable pair feeding the output terminals.
This remote location removes/attenuates the effect that the inductor may have on nearby PCB circuitry.
see the little bit of unused space between the leads of the two middle output devices.
Separate the output devices slightly, maybe an extra 5mm.
Now use that space to insert: the +ve supply, the -ve supply, the Power Ground and the Decoupling Ground (particularly the HF decoupling).
If possible the Zobel ground should also come into that area.
This brings all the high current routes to meet in that one small area and also helps reduce trace lengths.
Move the R//L off the PCB.
This operates just as well when located in the cable pair feeding the output terminals.
This remote location removes/attenuates the effect that the inductor may have on nearby PCB circuitry.
This is another very good suggestion. I put the inductor right at the speaker terminals, well away from any sensitive circuitry or cabling of any kind (other than the speaker output cable, of course).
Last edited:
Hi Dado
If you feed V+ and V- in at the middle of the output transistors as Andrew T and Harry recommend then you can run the feeds to the collectors out and return the current from the emitters back on the other layer directly on top. This is what I recommended to Harry earlier - I don't see how the loop can be made any smaller than this, it is effectively only the thickness of the board.
As to the feedback circuit. It occurs that it would be possible to send the trace (or wire) from the centre point of the output transistors back to the input then have the feedback divider resistors there, with the earth return back to the earth point between the output transistors. This should cancel any inductive couple and not dump the feedback current into the quiet earth.
Best wishes
David
If you feed V+ and V- in at the middle of the output transistors as Andrew T and Harry recommend then you can run the feeds to the collectors out and return the current from the emitters back on the other layer directly on top. This is what I recommended to Harry earlier - I don't see how the loop can be made any smaller than this, it is effectively only the thickness of the board.
As to the feedback circuit. It occurs that it would be possible to send the trace (or wire) from the centre point of the output transistors back to the input then have the feedback divider resistors there, with the earth return back to the earth point between the output transistors. This should cancel any inductive couple and not dump the feedback current into the quiet earth.
Best wishes
David
That is a return to the topic of this thread because it is my principle concern with Edmond's circuit. The hi gain is produced by the use of very hi impedance loads so I expect it will be fairly sensitive to stray capacitance and the like. I would like to use more current gain at lower impedance levels for a similar loop gain result. Not sure about how best to do that.
In ICs they scale the area of transistors to have current mirrors where the current in one path is a multiple of the other.
Anyone know if that would work here?
Best *ishes
David
BT* I think you try to use *ords *ith * so it *ill look funny
I have a current feedback design where for modest current increase (2-4) I use parallel transistors, also if available one can use different degeneration resistor values. Each of these come with their own drawbacks but nothing is free (except for the letter w, wwww)
Thanks
-Antonio
Don't take the NFB back to the Power Ground.
The +IN & -IN pins of the LTP (or other inputs) measure the Signal.
That signal comes in via the chassis located input socket (RCA, DIN XLR etc.).
The signal hot and the signal return are the two points/inputs about which the amplifier tries reproduce a copy at it's own output.
Give the amplifier the best chance it has of reproducing the input signal by connecting the input socket to BOTH of the amp inputs. That requires one to couple the NFB lower leg to the signal return, not to Power Ground.
Thanks to all for your help,
As this PCB is already produced I can't change to much. Here is what I think could help. I will cut(look the picture) main grounds(GND) from were it go around input stage and take from new GND by wires to the star ground. I removed the coil from the PCB too. Were to place the coil could be a problem. Often loudspeakers terminal are quite close and if the coils are fixded to the terminals there we have a cross talk.
For the next Layout I'll try to implement "all" (its not easy task) what was suggested here.
dado
Attachments
![DADO-TT-TMC-7-4-e-double.LAY].jpg](/community/data/attachments/279/279058-1d40a02716557e56654c43e1e2f1ecb2.jpg)
David,
There remains a small loop area due to the finite distance between the emitter and collector terminals of the O/P devices. The good news is that you can neutralize this loop too by an 'anti loop', i.e. a loop with the same area, but which rotates in the opposite direction. Think of lemniscate or figure of 8. One enclosed area rotes CW, while the other area rotates CCW. This way the radiated fields will neutralize each other.
Of course. you also can apply the lemniscate trick at the inputs. Then they will be less susceptible to magnetic radiation (courtesy of Edward Cherry).
Agreed, though with only two PCB layers you can't put the Vcc and Vee traces also on each other (correct me if I'm wrong).
Cheers,
E.
David,
You are right, the high impedance IPS output is of concern needs special attention: guard rings or planes. Not a serious problem, I think. Putting more current gain in the IPS is totally against the philosophy of the super TIS. If you do that, you will be confronted with other issues: an ill defined quiescent current of the VASes/TISes, which is precisely what I try to avoid with the super TIS.
BTW, did you read the recent comments on Bob's book?
I was really shocked by the ignorance of a well established audio guru regarding Iq issues with complementary VASes.
Cheers,
E.
To take all that e kno here and put it in a book ith no errors, all the calculations, all the configurations, and so on, ould be an unwending task (there it is!). As I looked through Bob's book I sa it more as a compendium of somehat unorganized knoledge, intending to get much of hat e've discussed here in documentation. The book still isn't big enough to even introduce everything e have learned through netorking. As a hole e have amassed tons of collective knoledge. To do this I think e ould need something like a textbook panel consisting of Ed, Harry, and so on, and it ould need to be continually informed by forum members to ensure nothing as forgotten.
Isn't this hy e have the DIYAudio iki? Each one of us probably knos more than could be contained in Bob's book. e just need to rite it don so it doesn't disappear ith us. Collaboratively, e could surely rite many chapters in a short time.
* I should really fix that button.
Isn't this hy e have the DIYAudio iki? Each one of us probably knos more than could be contained in Bob's book. e just need to rite it don so it doesn't disappear ith us. Collaboratively, e could surely rite many chapters in a short time.
* I should really fix that button.
You're right; in many ways my book is not big enough, although I do think it is well organized. I was lucky to get the 600+ pages from McGraw-Hill that I did. When I do a second edition I'll try to add more pages and add more knowledge. I'm always open to suggestions that would be good things to include there, or to better explain or elaborate on.
Cheers,
Bob
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.