Amp design attempt number 2 (simpler)

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

Another couple of ideas to try, thank you.

When you say inject the square wave via a 1K resistor do you mean taking the output of the amp and using that as the source.

Just so I understand why I'm changing the transistors (Q5,6) which characteristic is likely/possible to cause a change in behaviour?
 
Magic,

I shall be having a raid of components here at work.

Always feel much happier when I have things to try. Expected to get problems somewhere along the way with this amp. To have a problem like this seems like a minor one compared to what could have happened. At least the amp works, it's just "tweaking" really.

Thank you

Paul
 
Well I have made some more progress.

It's not the driver transistors and its not a local oscillation. Slowing the amp down does not help either.

Have worked out that the problem is with the control loop U11,12,13,14. Have also discovered that the situation is helped but not solved with a capacitor between the emitters of the VAS/TIS outputs.

Just got to work out how to stabilize this loop.
 
Last edited:
Toni,

Thank you for taking the time to look at my schematic. No I haven't removed the diodes. Another thing to try.

One thing I noticed was that if the power rails come up at the same time it oscillates. If you delay one of the power rails by a second or more it does not oscillate. Then I decreased the gate stoppers again and the same oscillation reappears regardless of power rail power up timings.

Paul
 
What a rookie error!!!!!

Messed up the ground scheme. Had the signal in ground sharing a trace with the feedback shunt resistor ground. The parasitics of the track allowed a feedback to happen via the signal in shunt capacitor and pulldown resistor back into the amp.

one track cut and new ground added and no more oscillation. Now to rebuild the amp back to its correct state.

:headbash::headbash::headbash::headbash:

This is a great learning curve. Loving it!
 
What a rookie error!!!!!

Messed up the ground scheme. Had the signal in ground sharing a trace with the feedback shunt resistor ground. The parasitics of the track allowed a feedback to happen via the signal in shunt capacitor and pulldown resistor back into the amp.

one track cut and new ground added and no more oscillation. Now to rebuild the amp back to its correct state.

:headbash::headbash::headbash::headbash:

This is a great learning curve. Loving it!
Beautiful! That's fun stuff (after finding it, not before obviously). Great you got it to work :)
 
Andrew,

Here's a diagram. Got some noise minor issues but two smps supplies are not helping.

It was the parasitics of the track that was causing the problem. The current flow of the feedback was changing the voltage at the signal input ground and feedback back into the amp.

Whether this grounding system is correct, I'm not sure. For the real thing the signal ground will go to its own star point away from the PSU grounds.

Magic,

Yes, it gives a sense of satisfaction.
I'm a stubborn type of person. It was only a matter of time.
 

Attachments

  • Draft4.jpg
    Draft4.jpg
    195.7 KB · Views: 167
Last edited:
What you have done is not what I would do.

Instead, are you willing to try a further change?
Take R11 direct to R12. The bottom of R12 becomes your Signal Ground reference point. All traces around here MUST have lowest loop area: the input pair and the NFB pair. The input references there, R12 references there and R11 references there. NO OTHER signal uses any of that trace arrangement.

Now take a reference from your SIGNAL GROUND to your main Audio Ground (I think you are calling this your "Star point".
As a further experiment insert a small resistor value into this last connection, maybe around 10r.
But you must bypass this resistor (if added) with a pair of inverse parallel diodes that can pass a fault current from Input Interconnect to the Chassis Earth.

Both (Rinsert=10r & Rinsert=0r0) these arrangements should work equally well when the amp is not connected to other equipment.
When you do connect other equipment, you will probably find that the resistor=0r0 or 10r will have a measurable and possibly an audible effect. Choose which ever is better. This can be optional, with a switch, or shorted pins, across the resistor.
 
Andrew,

I am always willing to try ideas out. The loop areas are already very small around this area. The layout was done with loop areas in mind.

Will implement your suggestion over the weekend. How would you physically place the PSU star ground and the signal star ground in relation to each other?

Thank you

Paul
 
The PSU output should never be used as a Star Ground (Main Audio Ground).

The Main Audio Ground serves the Speaker Return and PCB Power Ground with links running to the Signal Ground and PSU Zero Volts.
Because the MAG serves primarily Speaker and PCB, it should be placed close to these two locations. If the PCB output end were very close to the speaker terminals, then in my view the MAG can be on the PCB or on the Speaker Return terminal or in the short wire connecting the two. But NEVER at the PSU.
 
The PSU output should never be used as a Star Ground (Main Audio Ground).

The Main Audio Ground serves the Speaker Return and PCB Power Ground with links running to the Signal Ground and PSU Zero Volts.
Because the MAG serves primarily Speaker and PCB, it should be placed close to these two locations. If the PCB output end were very close to the speaker terminals, then in my view the MAG can be on the PCB or on the Speaker Return terminal or in the short wire connecting the two. But NEVER at the PSU.

I still don't understand this very well. If everything runs there separately, then as long as it's not upstream of any reservoir cap, why could it not be near the PSU output?
 
Well the results are in...

AndrewT's scheme works. The amp was still conditionally stable though. Probing the output with the DMM sent it into oscillation. So I went back to the sims. Found a peaking that I had somehow missed in the closed loop response. So there was a control loop problem in there. Increased the shunt compensation at the output of the VAS/TIS. This has solved it. Appears stable now. No oscillation on the scope. Need to double check things still, like feeding a square wave back to the rails via a 1K resistor as keantoken suggested.

Its been a productive weekend so far.

I have a theory (could be wrong) about the MAG not being near the PSU ground and that would be voltage gradients caused by the potentially high currents in the PSU. Apparently you can get these in the star point itself. This would be the same reason why the speaker out and feedback take off point need to be on their own spur.

I have got to say making mistakes is the best way to learn.
 
Stereophonic sound...

Now in stereo and stable. No output inductor yet but happy driving the Coda 7's. They have never sounded so good. (Maybe some bias ;) But pleased that I haven't designed a dud).

No measurements yet but they will come later in the project.

Now I turn to thinking about the power supply requirements. Was thinking of running a second toroid for the front end at a higher voltage to allow some PSU filtering to be using. Some sort of "ripple eater"

In the mean time its time to do metal work and rework the case. Have bigger heat sinks to fit (330mm x 125mm x 40mm) One per channel.

Here's the latest schematic.

Cheers

Paul
 

Attachments

  • Prototype.jpg
    Prototype.jpg
    384.4 KB · Views: 303
The main "ripple" will be caused by playing music. The reservoir and decoupling caps' current IS THE MUSIC, directly. So the louder the music the bigger the ripple, since the cap VOLTAGE is proportional to the integral of the music current, between charging pulses. The best thing you can do, to prevent the music current from affecting the voltage of anything else that uses the same supply is to put lots of decoupling capacitance right AT the power rail connections to the active output devices. Use smaller caps REALLY close to the devices, and larger ones slightly farther away. A larger number of them in parallel will do slightly better than one (or a few) larger ones, for the same total capacitance. And separate parallel paths will do better than a buss topology, for the caps, especially the multiple smaller one that should be really, really close to the device, that handle the transient currents at up to a few hundred kHz. You want roughly at least 10 uF per watt, for those, but should also have a closer core of 1uF to 10 uF that are film or C0G ceramic type.

See the rather-enlightening image of the LT-Spice simulation plots of cap currents and output voltage, in the post at:

http://www.diyaudio.com/forums/power-supplies/240955-resevoir-capacitors-chip-amps.html#post3599692

This post might also be helpful:

http://www.diyaudio.com/forums/powe...evoir-capacitors-chip-amps-2.html#post3607549

Cheers,

Tom
 
Last edited:
Could you please share the spice file?

Thanks.

No problem. Hopefully you'll get some value from it ;)

It uses Bob Cordell's models for the smaller transistors and Exicon models for MOSFETs

To get it to run properly need to set:

Gmin = 1e-009
Abstol = 1e-006

Use GEAR solving method.

Paul
 

Attachments

  • exicon mosfet.zip
    9.3 KB · Views: 63
  • compdiff Prototype1.asc
    35.9 KB · Views: 64
Gootee,

Thank you for that info and the links.

Some of your suggestions can be implemented. The only part I don't think I can do is the inner core of film caps. I only have 330n per device.

Your spreadsheet looks interesting! The only part I haven't figured out is the Vclip parameter. What does this relate to? My limited mind can't process this part.

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