1176 Motorboating/Oscillating

[sircletus]

Greetings, all:

I finally got around to doing my own 1176 PCB layout, after tinkering with mnats' for awhile.  This is my first PCB for discrete transistor design, though I've done several for IC-based designs and two for tube-based projects.

It works just fine and dandy and does what it's supposed to.  But if I crank the Output pot up to about "12" or higher on the dial, I get sub-audio motorboating AND an oscillation around 27kHz.  This is without an input signal, btw, so position of "Input" control makes no difference here.  Here's some additional info:

- Based on Rev. B, so no "low noise" circuit.
- Using Hairball transformers.
- All coupling caps are as specified in original schem, as are local PS bypass caps.
- Probing with a 'scope, VOM and/or my fingers can at some points completely eliminate both.
- Closing up the chassis makes no difference (so probably not local EM radiation)
- Doesn't matter where device is plugged in.  Tried in studio (its own circuit with power conditioning) and in my dining room "lab."
- Source seems to be coming from input amplifier and NOT the output amplifier, though I need to test a bit further.
- PCB layout is similar to actual 1176 PCB layout, at least in terms of component placement.
- Regulated +30V power supply with 2x 2200uF filter caps and specified cap on regulator output.

I understand that motorboating is often related to devices communicating through the power supply and that can be rectified (see what I did there?) through regulating the power supply and using adequate filter caps.  Am I understanding that correctly?

Could the LF and HF oscillating be related?

Any ideas for starting points?

Thanks!

 

[gyraf]

Power rail and/or ground stiffness? - Try beefing up your grounds, specially to areas eating significant current, by soldering a copper wire to the tracks? Compartment-alize your grounds and star them?

Jakob E.

 

[Majestic12]

Check if any of your cables are running near the input section of the pcb. I once also had HF oscillation problems on my Rev A when I was running some of the wires from the output straight across the pcb. Move your cables around a bit, maybe that helps.

 

[sircletus]

Jakob:

My PCB has a ground plane (top layer, same layer as power traces - is that the problem?), and is tied to chassis ground at one of the screws holding the AC power inlet to the chassis.  Would a 4-layer board with a power plane be an improvement?  Shame they're so expensive!


Majestic 12:

My XLR input and output jack are almost immediately next to their respective pads on the PCB, which are right next to each other, though they do not cross at any point.  But, the output traces run from the far right side of the board back to the upper left side where the bypass relay and XLR output connector are placed.

Both input and output connections from PCB to XLR are Mogami twisted pair.  Pin 1 on the input is tied to the XLR chassis connector via a series 51Ω resistor and .01µF disc cap.  Output pin 1 is tied directly to the output XLR chassis connector.


Thanks for responses, gentlemen!

 

[sircletus]

Taking MJ12's advice, I re-routed the wiring between I/O XLRs and their respective PCB pads, placing more distance between them (the wiring).  That took care of one oscillation. 

As for the motorboating, I cheated a little.  Probably not the "best" way of handling the situation, and I still might look for a better way to solve the problem, but I put a passive, first-order filter in between the input and output amplifier sections.  There's a 1µF cap before the output control, and I put a shunt resistor after that, tuning the value so that its effect on frequency response is minimal, but it still eliminates the motorboating.  So I'm about 3 dB down at 30Hz.

Far from ideal, but it works.  For now.

Any better ideas for solving this problem now or for avoiding this in the future?  PCB layout suggestions?

Suggested reading for PCB layout is also welcome!

Thanks, all!
IZ

 

[sircletus]

Never mind.  It was gone last night, but now it's back.

Basically, once I connect ANYTHING to the +output PCB pad, it oscillates.  Place my thumb near it, oscillation goes away.  Disconnect output XLR from PCB output+ pad, oscillation goes away.  Solder wire to output+ pad, leave other end unterminated, oscillation shows up.  Driving me NUTS.

 

[PRR]

> Place my thumb near it, oscillation goes away

So leave your thumb there.

Or find a dummy-thumb to install permanently.

Seriously: sure sounds like capacitive coupling from output to input around a high-gain circuit. This tends to provoke supersonic oscillation. In some circuits, the high levels de-bias some coupling cap until the amp stops working, then recovers at a subsonic rate (motoboating).

Think of it like house plumbing. You have water in, sewer out. You do NOT want sewage in the water. In plumbing, pipes don't leak much, but wires DO. Keep them apart. Shield too.

 

[simonsez]

> Place my thumb near it, oscillation goes away

So leave your thumb there.

Or find a dummy-thumb to install permanently.

Aha...you can't get those from mouser..

 

[sircletus]

> Place my thumb near it, oscillation goes away

So leave your thumb there.

Or find a dummy-thumb to install permanently.

Seriously: sure sounds like capacitive coupling from output to input around a high-gain circuit. This tends to provoke supersonic oscillation. In some circuits, the high levels de-bias some coupling cap until the amp stops working, then recovers at a subsonic rate (motoboating).

Think of it like house plumbing. You have water in, sewer out. You do NOT want sewage in the water. In plumbing, pipes don't leak much, but wires DO. Keep them apart. Shield too.

Thank you, sir.    I'll try a little experiment tomorrow:  I'll move the output XLR to the other side of the rear panel and see what happens.  I'll cut the traces from output trafo and run a twisted pair from output trafo straight to the output XLR.

If that doesn't work, where should I go next?

Looking at a modern 1176, I can see that there's a big ol' metal shield separating the barrier terminal from the rest of the PCB, and that the connection between output transformer and terminal/XLR is via twisted pair and goes nowhere near the input.  Sehr interessant.

I'd like to learn more "best practices" in terms of PCB design for circuits like this.  Any recommended reading would be appreciated.

 

[PRR]

> Aha...you can't get those from mouser..

Are you sure? Who can tell?

All Search Results for "thumb"  (1,295 Matches)

> big ol' metal shield
> output ... via twisted pair and goes nowhere near the input.

Same as laying out water and sewer ditches. When you can't *seal* your fluid (plastic pipes seal water well, coax and twist-pair leak some), then you also keep your output (toilet drain) FAR away from input (drinking water). In an imaginary water/sewer system that is all ditches (Romans ran big water in open aqueducts), then it might be wise to add a ditch in-between the water and sewer, drained to sewer, but nominally dry (or a "big ol' metal shield").

It all depends on overall gain and bandwidth (bandwidth because HF couples through universal stray capacitance). Ponder WWII radar IF strips. Long narrow box, input on one end, output on other end, 80dB gain at 30MHZ in 6 stages with partitions between stages. Audio is never that extreme. But it's not unusual to have 30dB-60dB in one box using amplifiers with gain out to 20MHz.

 

[sircletus]

[/quote]
Aha...you can't get those from mouser..
[/quote]

Now Craigslist on the other hand...

 

[sircletus]

Same as laying out water and sewer ditches. When you can't *seal* your fluid (plastic pipes seal water well, coax and twist-pair leak some), then you also keep your output (toilet drain) FAR away from input (drinking water). In an imaginary water/sewer system that is all ditches (Romans ran big water in open aqueducts), then it might be wise to add a ditch in-between the water and sewer, drained to sewer, but nominally dry (or a "big ol' metal shield").

It all depends on overall gain and bandwidth (bandwidth because HF couples through universal stray capacitance). Ponder WWII radar IF strips. Long narrow box, input on one end, output on other end, 80dB gain at 30MHZ in 6 stages with partitions between stages. Audio is never that extreme. But it's not unusual to have 30dB-60dB in one box using amplifiers with gain out to 20MHz.

PRR:

My overall PCB layout is a very "left to right" affair with sidechain and meter drivers oriented "below" the audio circuit.  I'll try to optimize that layout even more in the future for high gain circuits like this.

General PCB layout-related question for you: I'm using a faux "ground plane," basically a copper pour on one side of the PCB.  I tend to keep power traces on one side of the PCB with signal traces on the other.  Is it better practice to do a copper pour on the power side or the signal side?  I did this one on the power side.

I'm wondering if doing the pour on the signal side would help serve as a "shield" between sensitive parts of the circuit.  I seem to remember Walt Jung (in the IC Opamp Cookbook) advocating drawing ground traces as a protective measure around certain legs of the opamp.

Here's the part that's really biting me in the hindquarters: I'm using a relay true-bypass.  Two DPDT relays in close proximity to one another.  One steering the +/- input, one steering the +/- output.  Obviously the two have to share a few traces to route the input signal straight to the output connector when in bypass.  Could this be another issue?  I'm wondering if ground traces around pairs of relay pins would help with isolation and serve as a shield...

Thanks as always for sharing your experience and knowledge!

 

[PRR]

Bringing Out and In to a small relay is asking for leak-back.

I would bypass the bypass and see if it all gets happy.

True, a LOT of good gear has bypass on small switch/relay. Perhaps they run the In and Out wires in opposite directions? So only the tip of the contact is in proximity?

 

[sircletus]

Bringing Out and In to a small relay is asking for leak-back.

I would bypass the bypass and see if it all gets happy.

True, a LOT of good gear has bypass on small switch/relay. Perhaps they run the In and Out wires in opposite directions? So only the tip of the contact is in proximity?

Sure enough, I cut traces, bypassed the relay and moved the output XLR a few inches away from the input and - voilá! - I'm in business.

I figured relay bypass was a better bet than running I/O wiring to the front of the chassis then back again.  Relays make life easier in that regard, but this?  Oy.  I'm using fairly large relays (the bigger Omron G6As) instead of the teeny tiny guys, but I guess 5mm separation vs 2.5mm just isn't enough.  Who woulda thunk it.  I don't suppose my previous thought of using ground traces around sensitive pins as a "guard ring" would help at all...

I'm going to remove the relay from the PCB and make a small daughtercard on a perfboard, mount it to the chassis near the newly-relocated XLR connector, try some different signal routing options and see if I can get anything to work.

Thanks again, PRR.

 

[CJ]

that 1166 pc board seems to be sensitive to layout,

maybe post your artwork so we can tale a peak,

putting I/O jacks side by side is a standard practice, seems counter intuitive, but it actual shrinks the flux loop area which can result in less hum and noise,

so you integrate flux with respect to area to get total induction of noise,

loop on side simulates tighter layout>

 

[sircletus]

CJ:

Interesting.  I'll upload a screen of that corner of my PCB in the next day or two.

Really, though, I took the output leads of the transformer, ran a shielded, twisted pair straight to the output XLR (detached from chassis) and moved it around to get an idea of where coupling was occurring.  As soon as I got within about two inches of the input area of the PCB, oscillation like crazy...

 

[radardoug]

You only need one double pole relay to do the bypass.
Output from the relay wiper, normally closed to the input, normally open to the output of the device.
When power is off the unit is hard bypassed.

 

[radardoug]

CJ:

Interesting.  I'll upload a screen of that corner of my PCB in the next day or two.

Really, though, I took the output leads of the transformer, ran a shielded, twisted pair straight to the output XLR (detached from chassis) and moved it around to get an idea of where coupling was occurring.  As soon as I got within about two inches of the input area of the PCB, oscillation like crazy...

Are you using good quality cable like Belden or Mogami? Have you got the input shielded as well?

 

[sircletus]

CJ:

Here's a screen cap of the I/O corner of the PCB.  the blue traces coming from the upper right are the output traces from the trafo.

Radardoug:

I'm using Mogami twisted pair with shield grounded only at the XLR end.

 

[radardoug]

Does this mean the board itself is possibly not grounded to the chassis?