Seeking for advice: routing/physical layout

[Onvinyl]

Hi,
as a follow-up of this MC-Phono stage I'm in the process of building the attached schematic.
Currently, I can't make it work properly and I think I'm don't find the right routing. Though feeding it with batteries, I get excessive noise and/or bad oscillations. The input stage itself is no problem and works very stable, the problems arise when U1 comes into play.

I see several issues that make it hard:

• long feedback loop
• many parts = long traces for the RIAA-correcting part
• ultrahigh gain
• I'm uncertain where the servo should be placed
• Finding the right compromise between 'make the shortest signal path, and the shortest return current path, don't span a wide area between the live lines' is tough

I would take a photo of the current stage, but I'm prototyping p2p and I doubt one can see anything useful.

The circuit exhibits several desgin decisions that are debatable, if possible, I would like to concentrate on routing & layout.

I have read heaps of white papers concerning these issues, but I'm hanging anyway.

Thanks,
Rüdiger

[syn08]

Originally posted by Onvinyl

• long feedback loop
• many parts = long traces for the RIAA-correcting part
• ultrahigh gain
• I'm uncertain where the servo should be placed
• Finding the right compromise between 'make the shortest signal path, and the shortest return current path, don't span a wide area between the live lines' is tough
  
  [syn08]
• no frequency compensation
  [/syn08]
  

I don't think it's at this stage a layout and routing issue.

How to compensate this thing? I may have a few ideas but why don't you try yourself first using LTSpice that you already captured the schematic in?

As you said, there are quite a few debatable decisions in this design.

[Onvinyl]

Originally posted by syn08

[syn08]
# no frequency compensation
[/syn08]

You were of course right. I did a quick dominant pole compensation which needs a substantial 33pF-Cap (next smaller one I have in spare is 15pF which is to small). All noise, ringing, osc. issues solved so far. Sq-wave rounds a bit at 20kHz (without riaa-network, of course). It doesn't match the simulated situation too well, but I don't really know which options to use in LTSpice.

So it might be a good idea to research for the more fancy compensation-schemes which is what I'll do now.
Thanks,
Rüdiger

[Onvinyl]

Sorry, I meant lead compensation (cap over R4)
15pF should be the best value according to the sim, in reality there is some ringing left with this value.

Regarding routing: I have read different opions whether it's a good idea to route ground traces under opamps. In my case it helped not to do so.

Rüdiger

[djoffe]

take a look at the top of R33...it goes to the sources of the long-tailed pair (the 6 fets). That limits the current source's output impedance to something less than 22K. I don't think that's what was intended. The top of R33 should probably go to ground.

[Onvinyl]

Originally posted by djoffe
The top of R33 should probably go to ground.

Oops, I did see the forest for the trees not...
This is indeed a mistake.
Thanks,
Rüdiger

[Onvinyl]

I made progress: previously, I placed R2 as close as possible to the gates of J4-J6. Now its ground connection is as close to the bypass-C's ground connection as possible (and close to central ground return point) and everything works a lot better now, despite an adventurous routing back to the gates of J4-6.

I fixed the cascode issue. With the now working CCS the circuit is also less noisy. Does that point me to possible common mode noise interference?

This is definitly no 'let's swap the opamp and see what happens'-kind of circuit, as every chip needs its own compensation arrangement. At the moment I tried op27, LT1028 and OPA637. op27 is a bit (too) slow, LT1028 and OPA637 give nice squares. I'm surprised that the OPA gives more noise at the output, I thought the input stage noise would dominate. Another hint to suboptimal wiring?

I might test other op-amps as well at a later date.

Rüdiger

[Onvinyl]

forget the noise issue, it is something else...
Rüdiger