Randy Slone's "Fig 11.4" (Self's "Blameless"?): PCB layout

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Jump to Latest
A

AndrewT

R.I.P.
Joined 2004
Hi,
the 80W JLH mosfet push pull amp schematic is in this Forum.
There is doubt about the voltages marked on the schematic, but otherise it is a repeat of the original.
It is based on Source Follower.
It has about 4 or 5 compensation components spread around the PCB.
 
T

tcpip

Member
Joined 2003
tcpip.dhandanought.org/
pinkmouse said:
One track from R4, one track from C7, one track from input gnd, all leading separately to a pad where you connect to the main star ground. ;)
Click to expand...
I was thinking that my INGND would be the point which would connect to the off-board star-ground. So I'd do one track from R4 to INGND, and one from C7 to it, and then take a wire (a physical copper cable) from the INGND pad to the off-board star ground point. Wouldn't that do? Both the input RCA's ground and the copper wire to the star-ground point could come to the INGND pad. Would that be okay?
 
T

tcpip

Member
Joined 2003
tcpip.dhandanought.org/
MikeBettinger said:
I’m not trying to tantalize. I'm just looking at your layout and know from the experience that mixing anything that going on in the output stage with the input stage, which is exactly what you're doing with your power grounds, will get you chasing your tail for a long time.
Click to expand...
Okay... so, given my inexperience, I'm trying hard to translate what you've just summarised here into a set of specific pointers. Tell me if I'm right or wrong:
  • One case of mixing input and output stage grounds is because of decoupling caps. Ideally, the grounded ends of the decoupling caps need separate solder pads which will have wires leading to star-ground.
  • Another case of mixing grounds of different stages is the RC Zobel. Ideally, the Zobel, even if kept on the PCB, should have its ground lead going by a separate solder-pad and separate copper wire to star-ground.
Are these the only two? Do I need to look at the ground points of the VAS too?

The bypass caps (if needed) need a Low-Z path around the circuitry back to the supply ground. I suggested the chassis as an easy after the fact tweak that normally satisfies this if the need arises. A solder tag tapped and screwed into the grounded heatsink, next to where the supply connects to the PCB (or better yet, next to the transistor leads) allows short leads on the caps.
Click to expand...
Okay.... I think I got it. Whatever way I choose to do it, I'd need to have the ground-ends of these caps connected by a direct, separate copper path to the central star-ground point. Is this right?

My comment on not needing the bypass caps is based on my approach of building a design without the bandaides and adding them only if there is no other way to get it to the listening part.
Click to expand...
Got it. Will try it and see. And in the mean time, if I can separate out the ground path for the output-stage decoupling caps (C11 and C12 in the first schematic I'd posted), I'll do so.
 
T

tcpip

Member
Joined 2003
tcpip.dhandanought.org/
John,

I think I understood a lot of the things you're describing, though it'll take time to sink in.

john_ellis said:
Sorry if there was a lot to read through ... I'll see if I can help with the bits you did not follow.
Click to expand...
You're one of that rare breed of experienced diyers who actually encourages a confused beginner instead of asking him to "go do some reading up first." :D

Increasing the twopole resistor is possible if the capacitors are made smaller, as this just keeps the time constants constant. Reducing the capacitors are possible if the input resistors are increased to keep these time constants tracking.
Click to expand...
This makes sense. Randy Slone's book mentions a rule-of-thumb formula to correlate the current in the input LTP legs with the value of the Miller cap. I'm going to re-visit that with your notes in front of me.

You can use a CF OPS, but there are only about two advantages over the EF: quiescent current stability and greater efficiency because the "saturation" voltage when one of the pair is on is lower than that of the EF pair.
Click to expand...
Isn't lower distortion another reason to go CF? I thought the local feedback had something to do with it.

I'll try to undo the confusion over the CCS/VAS: of course, the CCS and VAS work together, but in your circuit,you had a transistor which looked as though it was trying to define a constant current in your VAS stage.
Click to expand...
No, that transistor was put there for protecting the VAS transistor from too high a current. Q17's job is to limit the base voltage supplied to the Q5+Q6 Darlington if the emitter current through Q6 rises too much.

I think that the filter resistors on the power rails to the input stage are intended to stop power pulses reaching the input stages. The resistor is really the only effective component. As discussed in other threads, you may be better to use a separate supply voltage slightly higher than the output. But to keep you on track for a simpler circuit, try missing out the diode as you say.
Click to expand...
Yes, I'll most certainly try to build my first version of the amp without the diode.
The current source I recommend works well at filtering noise from the rail it connects to for the input stage. I've attached the input current source for you. Originator of this was Bailey. (NPN version).
Click to expand...
What's the difference between this one and the two-transistor current source in increasing the PSRR? (Or have you described it already in your last post?...)

AndrewT has said he uses 649's at 58V so maybe they're fine, you could also use 649A's; and I've used 968's.
Click to expand...
Yes, maybe I'll stick to 649 (availablve here) or 649A if I can get them for my v1 amp. Later I may think of switching.

The non-Miller approach is to wire the feedback capacitor from the VAS collector to the feedback transistor base....
Click to expand...
Okay, got it. I've seen this somewhere too, but that design had (what appeared to be) a Miller cap, in addition to this extra cap.

Thanks for the patience. :)
 
M

MikeBettinger

Member
Joined 2006
VPIindustries.com
tcpip said:

Okay... so, given my inexperience, I'm trying hard to translate what you've just summarised here into a set of specific pointers. Tell me if I'm right or wrong:
[
Click to expand...

My specific points were to place the zoebel network across the speaker terminals so that any energy being passed through it would return directly to the power supply return/sink (however you want to view it); avoiding any unintended coupling or paths back through the input circuitry. The same reason applies to the bypass caps.

The reason I do not advocate using separate ground wires back to the supply ground (from the board) is because it will act like a pretty good transmitting antennas and couple high frequency noise into any sensitive node in your design.

I realize what you are trying to do, but with any amp design, especially because of you inexperience, you will need a few itterations to produce a properly functioning/stable end product. Do your self a favor and don't try to hit every possible improvement on the first try. Produce a working prototype and then start the refinements and scaling down of the board. This way when the gremelins sneek in you have a working version to help you analyze where the problems have come from.

Just trying to help.

Regards, Mike.
 
T

tcpip

Member
Joined 2003
tcpip.dhandanought.org/
MikeBettinger said:
My specific points were to place the zoebel network across the speaker terminals so that any energy being passed through it would return directly to the power supply return/sink (however you want to view it); avoiding any unintended coupling or paths back through the input circuitry. The same reason applies to the bypass caps.
Click to expand...
I understood it... I'm not disagreeing with your suggestion. My last post was just to verify my understanding. I think if one has to do what you've said, it's better to do these things off-board.

The reason I do not advocate using separate ground wires back to the supply ground (from the board) is because it will act like a pretty good transmitting antennas and couple high frequency noise into any sensitive node in your design.
Click to expand...
Okay. I hadn't worried about this. But in any case, less wires carrying HF surges will be less noisy, so it makes sense.

I realize what you are trying to do, but with any amp design, especially because of you inexperience, you will need a few itterations to produce a properly functioning/stable end product. Do your self a favor and don't try to hit every possible improvement on the first try. Produce a working prototype and then start the refinements and scaling down of the board.
Click to expand...
Thanks a lot. And yes, a lot depends on having something working. Then, the modifications can be tried step by step.

After these two steps are moved off-board, what else do you suggest about the board layout itself?
 
T

tcpip

Member
Joined 2003
tcpip.dhandanought.org/
v6: with separate VAS CCS, higher VAS current, PCB changes

In this version, I've separated out the VAS CCS, using a red LED for the biasing device. Assuming that R14 now drops 1V, I've set its value to allow a fairly high CCS current. On the board, I've separated the tracks to the ground points (PGND1, PGND2, and INGND) from various subsections so that they run separately to the ground point. I've also replaced smaller heatsinks with larger ones. I know the final component layout can only be done after I see what sort of TO220 heatsink I get in the market, and arrange components to make space for them.

I've not deleted the output rail RC and LR components, figuring that I can fit them off-board and leave those holes blank on the PCB.

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.
 
M

MikeBettinger

Member
Joined 2006
VPIindustries.com
tcpip said:

I understood it... I'm not disagreeing with your suggestion. My last post was just to verify my understanding. I think if one has to do what you've said, it's better to do these things off-board.
Click to expand...


Board real estate seemed to be a concern to you. Your comments about leaving space for them indicated you seem to either: need to keep them on the board or that you believe that you can get them to work there. Hence my comments.


tcpip said:
After these two steps are moved off-board, what else do you suggest about the board layout itself?
Click to expand...

I would suggest you build the board with the layout that you've put your time and effort into.

I would lay the board out differently based on my experience, but then there are many knowledgable people on this site that would no-doubt lay theirs out differently, based on their experience.


Once you get past the fundamentals, PCB layout for audio is an art form and the highly subjective rules won't be conveyed, much less agreed on, in a handful of posts.

Regards, Mike.
 
T

tcpip

Member
Joined 2003
tcpip.dhandanought.org/
MikeBettinger said:
Board real estate seemed to be a concern to you. Your comments about leaving space for them indicated you seem to either: need to keep them on the board or that you believe that you can get them to work there. Hence my comments.
Click to expand...
I'd suspected you'd think this way. :)

Actually, the next time I need to add more parts on the board, and I run out of space, I will knock out those components. Till then, I'm retaining them. That's all. Till now, I've managed to add separate VAS CCS without knocking out anything, hence I'll just see if there's anything else to add. If not, I'll keep those RC and LR marks.

It's quite likely the space crunch will hit me if I need to add large heatsinks for the VAS CCS and VAS transistors and that's when I'll get rid of these parts.

I would suggest you build the board with the layout that you've put your time and effort into.
Click to expand...
Yes, I guess I have no choice but to do that now. :) Any show-stoppers you can see in the layout? Let me know.
 
T

tcpip

Member
Joined 2003
tcpip.dhandanought.org/
A question about a smoothing cap for CCS

If you see the CCS schematic below:

*link removed a request of poster*

I wanted to know what would I lose if I removed the capacitor and combined the values of R1 and R2 into one resistor? Someone mentioned this cap may be to reduce the turn-on thump of the amp. Someone else mentioned that it helps increase PSRR. I wanted to know whether there's a serious drop in CCS quality if I omit the cap? I'm wondering about this from the point of PCB real estate, if push comes to shove.
 
T

tcpip

Member
Joined 2003
tcpip.dhandanought.org/
MikeBettinger said:
Dump the output transistor bypass caps until you think you actually need them. Ground them separately to the heatsink, or anywhere to return them to whatever part of the supply you believe will absorb the energy; your current layout dumps it back into the input stage.
Click to expand...
I was reading Self's site over the weekend, and his analysis of eight different sources of audio amp distortion. And in that analysis, I saw that one of the eight sources of distortion was improper fitting of these output decoupling caps. Till then I hadn't realised that these were such a common and major source of adding distortion to an otherwise good design. He too says that these caps should be connected to star ground via separate paths so as not to disturb the grounding of the rest of the circuit. Wow! :)
 
lineup

lineup

Member
Joined 2005
tcpip said:

Till then I hadn't realised that these were such a common and major source of adding distortion to an otherwise good design.
He too says that these caps should be connected to star ground via separate paths
so as not to disturb the grounding of the rest of the circuit. Wow! :)
Click to expand...


This is a true.
Can be an issue also in op-amp circuits.

Whenever a supply rail Capacitor does filter something away
this HAS TO go away somewhere.
In fact, it is injected into the ground connection of this capacitor.

So the nightmare is to connect a power supply ripple filter electrolytic
to ground close to or in same rail as The Input of signal.

This is also perhaps one of the reasons, in some op-amp circuits, the designer
will recommend not the 2 usual decoupling caps at V+ and V- pins to ground
but instead
one cap from V+ directly to V-


lineup
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Top Bottom