>everyone must be busy with details of the PCB layout for the superTIS
Not yet. I'm still too busy with perfecting the software of my distortion analyzer.
Now dual sound cards are supported and a lot of bugs are resolved, though not all of them. Also numerical noise and spectral leakage is a concern. At the moment it's around -250dB. It would be nice if I can break the -300dB barrier (not that I really need it but it's cool).
Cheers,
E.
Not yet. I'm still too busy with perfecting the software of my distortion analyzer.
Now dual sound cards are supported and a lot of bugs are resolved, though not all of them. Also numerical noise and spectral leakage is a concern. At the moment it's around -250dB. It would be nice if I can break the -300dB barrier (not that I really need it but it's cool).
Cheers,
E.
Hi Arthur
As I wrote, it is just doodles so far. But for what it's worth...
It seems common to use a symmetrical layout and feed the +ve and -ve power to opposite sides of the board. This doesn't cancel the currents in the OPS very well. Some layouts try to keep the power cables paired up until the board but on the "Honey Badger" (for instance) this can lead to asymmetric layouts.
This circuit cries out for a symmetrical layout so I plan to put the power feed to the middle of the OPS to minimize the current loop areas.
Some people like to layout the OPS transistors on opposite sides of a narrow board but my available space would best fit all the OPS transistors in one line.
This is not quite equalized for lead inductance etc. between transistors - Probably not important but could one do better? Say with the transistors in a circle and central feeds.
We have star earths, why not star power supply? Feed the IPS from a separate, star tapped power lead.
I would like some sort of thermal coupler for the LTP transistors to keep them balanced, to place the TO-92s flat to flat is tidy but complicates layout.
Bob Cordell and D. Self both have a selection of circuits within the bounds of common practice. But I haven't seen a selection of PCB layouts with notes for comparisons and contrasts. So I have some ideas but few norms or benchmarks. I hoped some one would comment.
Best wishes
David
![DADO-TT-TMC-7-4-d-double.LAY].jpg](/community/data/attachments/284/284419-86346734fbe3ff1af4af8c5ce72274aa.jpg)
I reckon you are wise to expend your energies in this direction.
The vast majority of PCB layouts presented on this Forum make the mistake of including high loop area in the power trace lines.
Look at post84 layout and work out the loop areas and trace lengths of the MF and HF decoupling. Bonkers!
Last edited:
You mean post 804 I suppose??
What's so wrong with this layout, high current is on the top, low current is after C multipliers. OPS grounding is separated from input circuit ground. 470uF caps are close to the output transistors. Always is possible to make it better, but I never seen any layout made by you. Could you show how it should be made, with a drawing not just with words.
Thanks in advance.
dado
What's so wrong with this layout, high current is on the top, low current is after C multipliers. OPS grounding is separated from input circuit ground. 470uF caps are close to the output transistors. Always is possible to make it better, but I never seen any layout made by you. Could you show how it should be made, with a drawing not just with words.
Thanks in advance.
dado
On class AB coupling the supply rails together has its problems since they carry half wave rectified facsimilie of the music signal. So, if you lay the supply rails and the ground return together, you actually can end up coupling garbage into your amp front end and TIS via the non- conducting rail. This is another reason why rail decoupling is important. Also, in most layouts, the speaker return line is taken from the output terminal straight back to the PSU complicating the matter. I generally keep the + and - rails apart after the smoothing caps. I run the speaker cable return line straight back to the PSU. For the ground on the PCB, best is to just keep is as short as possible.
I seem to recall Syn08 discussing somewhere the challenge of class AB layout for sub ppm distortion performance with some good practical tips.
No silver bullet answer here I am afraid. It's all about trade offs
Last edited:
you should want +,- rail traces on top of each other (or more precisely, common centroid layout) - then the close paired trace's half wave AB current's magnetic field sum looks linear from just a little distance away
then any coupling is mostly linear, adding small frequency dependency to the gains
this is much better than separated PS traces nonlinear fields
if the gnd return is also close by then you can have pretty rapid fall off of the nearly canceling load current's fields
then any coupling is mostly linear, adding small frequency dependency to the gains
this is much better than separated PS traces nonlinear fields
if the gnd return is also close by then you can have pretty rapid fall off of the nearly canceling load current's fields
Last edited:
Hi Andrew
What PCB layout software did you use and any comments?
I notice that there is no PCB layout in your E-amp documentation.
I understand if you want to retain your intellectual property but then could you show a redacted version?
A pity that peoples' natural desire to keep their PCBs proprietary means there is little established best practice.
Best wishes
David
What PCB layout software did you use and any comments?
I notice that there is no PCB layout in your E-amp documentation.
I understand if you want to retain your intellectual property but then could you show a redacted version?
A pity that peoples' natural desire to keep their PCBs proprietary means there is little established best practice.
Best wishes
David
Hi David, I use Sprint Layout software, it's very intuitive with only drawback a quite small component librery. You can make your onwn one.
BR dado
Thank you for that. Also I should have thanked you earlier for the display of your layout. I said that not many layouts were shown because people wanted to keep them proprietary, but perhaps it is also because of a fear of criticism. So thanks.
Best wishes
David
these two comments seem to be on opposite sides !
Which to believe?
I know which I want to believe.
Speaker return layout.
Hi Andrew
I am not sure I understand your point. Your run the speaker cable the same way as most layouts and complicate the matter?
It seems to me that the way to minimize the loop area is to run the speaker return line next to the speaker hot line all the way back to the board where they connect to an earth point and the output. The output connection is next to the power supply connection so the earth + power supply loop area is minimized too.
Best wishes
David
Hi Andrew
I am not sure I understand your point. Your run the speaker cable the same way as most layouts and complicate the matter?
It seems to me that the way to minimize the loop area is to run the speaker return line next to the speaker hot line all the way back to the board where they connect to an earth point and the output. The output connection is next to the power supply connection so the earth + power supply loop area is minimized too.
Best wishes
David
Perhaps I shouldn't say it here as I haven't had the time to converse directly with Andrew about it (sorry Andrew), but I believe his e-amp layout (as it is described in the e-amp document from June 2012) is sub-optimal; in particular the way ground connections are made is dubious. Andrew has separate PSU and amplifier boards, but all grounds are starred/teed on the PSU board, including connections for the speaker return, output stage zobels and output stage decoupling caps. This massively increases the effective inductance of the zobel networks and decoupling caps and means it's unlikely that they will be doing the job they should be doing (respectively: presenting a resistive load to the output stage and reducing the power supply impedance at high frequency). I'm also not convinced by the order in which the ground connections are teed (zobel, speaker return, decouple, signal; in particular I believe the speaker return and decouple connections should be reversed).
Speaker hot and return should be twisted together and connected directly to the amplifier output stage board. Feedback (signal and ground) should be taken as close as possible to these connections (I run a twisted pair back to the input stage instead of trying to lay out feedback tracks on a board).
I understand people's desire to run + and - power rails on top of each other throughout the output stage however I can't see how this can be done whilst simultaneously keeping said traces away from the output trace or transistor base/gate traces. So if you do run + and - power rails on top of each other throughout the output stage, you make the radiated magnetic field linear, but couple it better to places you don't want it coupled to. What I do is bring the +, - and ground in together to the centre of the board (double sided with one side as ground plane), then take the + to the NPN devices and the - to the PNP devices, trying to keep the + and - traces as symmetrical as possible. This keeps the supply traces as far from the output trace as possible, but also means that magnetic fields produced by the + trace and - trace should couple equally to sensitive parts (due to the symmetrical layout) and therefore look "linear" to the victim. Radiation should be minimised thanks to minimum loop area; out and return paths for high-frequency supply-rail currents lie on top of each other due to ground plane). Obviously what is required here is a controlled experiment with the two different layout approaches but sadly I have never had enough time to do this.
Speaker hot and return should be twisted together and connected directly to the amplifier output stage board. Feedback (signal and ground) should be taken as close as possible to these connections (I run a twisted pair back to the input stage instead of trying to lay out feedback tracks on a board).
I understand people's desire to run + and - power rails on top of each other throughout the output stage however I can't see how this can be done whilst simultaneously keeping said traces away from the output trace or transistor base/gate traces. So if you do run + and - power rails on top of each other throughout the output stage, you make the radiated magnetic field linear, but couple it better to places you don't want it coupled to. What I do is bring the +, - and ground in together to the centre of the board (double sided with one side as ground plane), then take the + to the NPN devices and the - to the PNP devices, trying to keep the + and - traces as symmetrical as possible. This keeps the supply traces as far from the output trace as possible, but also means that magnetic fields produced by the + trace and - trace should couple equally to sensitive parts (due to the symmetrical layout) and therefore look "linear" to the victim. Radiation should be minimised thanks to minimum loop area; out and return paths for high-frequency supply-rail currents lie on top of each other due to ground plane). Obviously what is required here is a controlled experiment with the two different layout approaches but sadly I have never had enough time to do this.
Last edited:
Nice to have some confirmation of my previous post
Completely concur on the centre feed power + earth and symmetry.
What I plan to do is to run the + rail out from the centre connection as a sub-plane (on the bottom of a double sided board) out to the NPN output transistors (collectors) and run the return (emitters) on top and similarly for the PNPs. This should produce about 99% cancellation (for Hfe of 100).
Make sense?
Best wishes
David
Nice to have some confirmation of my previous post
Indeed. Also, thanks for quoting as it meant I noticed a typo; I should of course have written "hot and return" or "live and return", as "hot" and "live" are in this case synonyms! Fortunately the edit timeout had not expired so I've been able to correct my original post.
David, all I know for sure is that layout issues hurt my brain much more than I think they should. It seems like basic stuff that should be easy! What you've said above makes me think I need to revisit slightly how I do output stage layout, of course in an emitter-follower stage the return path is the emitter leg not ground! You only need to keep the "output trace" away from the power supply traces once it is indeed the output trace - i.e. the + and - emitter paths after they have joined. Before they've joined, the emitter paths are carrying the nasty half-wave stuff. The ideal is for the output trace to be perpendicular to the power supply traces and the only way I can see to do that is 3D - on the board, bring the NPN and PNP emitter traces together at a point, which is the point at which you make the speaker live connection. Then, take the feedback (via twisted pair) after a few cm of the speaker live wire.
Last edited:
Well, I appreciate tHe feedback, but perhaps need to clarify.
Speaker wiring.
I in fact use very heavy twin core (actually its speaker cable) inside the amp. The hot goes from the board out to the +speaker terminal. The return comes back down alongside the hot and then branches off to the PSU 0V Tee. I don't bring return back to the PCB, and then send the combined ground back to the central ground point on the PSU which is what is proposed above. I have lost no sleep over doing I this way.
Zobel. This is taken directly from the board and is the connected closest to the PSU 0V T point (see below). The inductances involved are no more than a few 10s of nH - certainly not massive and only a small fraction of the total speaker cable inductance which is likely to be in the uH range. I do it this way because I want to avoid injecting HF currents into the ground return. A star connection approach here would also work. The main point is that this is a dirty return with HF, so termination location is important.
Feedback pick off point and location- no argument there. I do exactly this on the e-Amp.
Re the supply wiring (+and - and speaker hot). On the PCB layout, you can try to make things symmetrical, and I also did this. However, many people believe that by coupling the speaker and + - lines together you minimize radiated noise - which you do but at the probable expense of injecting noise into the non conducting rail. The answer is good decoupling (I used a ripple eater as well to remove PSU ripple, but it also helped with this issue) and careful layout.
Grounding Order. The e-Amp Uses a T grounding method. I have used both Star and T. The grounding order in the T is important in order to avoid injecting noise sources (caused by common impedance coupling) into more sensitive parts of the circuit. I T off from the PSU reservoir cap ground connection (you should do this in any wiring scheme BTW, even the Star) after which the first return is the Zobel- we don't want HF of any sort in any of the other grounds, then the speaker return, then the amplifier ground return. My input signal ground can be configured for returning via the amplifier ground connection, or it can come from to the T via the PCB. Right now, it comes via the Amp PCB, which is how the Ovation 250 is also setup.
Both amplifiers are very quiet.
Speaker wiring.
I in fact use very heavy twin core (actually its speaker cable) inside the amp. The hot goes from the board out to the +speaker terminal. The return comes back down alongside the hot and then branches off to the PSU 0V Tee. I don't bring return back to the PCB, and then send the combined ground back to the central ground point on the PSU which is what is proposed above. I have lost no sleep over doing I this way.
Zobel. This is taken directly from the board and is the connected closest to the PSU 0V T point (see below). The inductances involved are no more than a few 10s of nH - certainly not massive and only a small fraction of the total speaker cable inductance which is likely to be in the uH range. I do it this way because I want to avoid injecting HF currents into the ground return. A star connection approach here would also work. The main point is that this is a dirty return with HF, so termination location is important.
Feedback pick off point and location- no argument there. I do exactly this on the e-Amp.
Re the supply wiring (+and - and speaker hot). On the PCB layout, you can try to make things symmetrical, and I also did this. However, many people believe that by coupling the speaker and + - lines together you minimize radiated noise - which you do but at the probable expense of injecting noise into the non conducting rail. The answer is good decoupling (I used a ripple eater as well to remove PSU ripple, but it also helped with this issue) and careful layout.
Grounding Order. The e-Amp Uses a T grounding method. I have used both Star and T. The grounding order in the T is important in order to avoid injecting noise sources (caused by common impedance coupling) into more sensitive parts of the circuit. I T off from the PSU reservoir cap ground connection (you should do this in any wiring scheme BTW, even the Star) after which the first return is the Zobel- we don't want HF of any sort in any of the other grounds, then the speaker return, then the amplifier ground return. My input signal ground can be configured for returning via the amplifier ground connection, or it can come from to the T via the PCB. Right now, it comes via the Amp PCB, which is how the Ovation 250 is also setup.
Both amplifiers are very quiet.
Last edited:
I believe Syn08 experimented with bringing his power and/or speaker connections perpendicular onto the PCB. This was when he was doing his sub 1ppm amp. He remarks somewhere that he amost gave up trying to get the wiring layout right. So, this is not easy and 'cancellation' is a relative term.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.