DIY LA-2A oscillation at 24kHz. Tried the usual...missing something?

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

sircletus

Well-known member
Joined
Nov 19, 2008
Messages
165
This is on my own, 2-channel LA-2A-style PCB layout, with a few key differences:

- Cinemag 1:1 in AND out transformers  (so no huge step-up on the input, no step-down on the output)
- Power and audio are separate PCBs, each with its own separate path to star ground point
- Heater supply is 12VDC regulated
- 6AQ5 replaced with 12AQ5
- "variable capacitors" on both channels have been replaced with fixed 470pF
- neons replaced with zener diode
- All switches have been replaced with relays controlled via the front panel (separate, regulated 9v supply for relays and meter LEDs)


Circumstances under which oscillation (always right around 16kHz) occurs:


- output and gain reduction turned all the way up
- changing pre-emphasis sidechain filter setting has no effect on oscillation

I've tried much of the "standard" tube stuff to eliminate the oscillation:

- Tried 1K grid stopper resistors in many places:
      - both sections of 12BH7A
      - input of sidechain triode (second section already has one as per original LA-2A schematic)
      - input of sidechain pentode

- Tried anode bypass caps (100pF - 220pF)  everywhere but the 12BH7A
- Swapped T4s just in case, to no avail.
- Re-sanded star ground point to maximize chassis contact, to no avail.


The weird/Voodoo Stuff:


- It doesn't happen on channel 1, only on channel 2.
- Probing with 'scope on sidechain triode anodes makes it go bye-bye
- A combination of grid stoppers and a 220pF anode bypass cap on the sidechain triode solved the problem in channel 1; same bypass cap doesn't help with channel 2.

Finally...[/b]

- DISABLING the sidechain (a la an 1176 bypass) makes the problem disappear, which tells me the problem is cropping up in the sidechain somewhere.
- Further, the oscillation is measurable on the input of the pentode; it goes away when the GR control is reduced OR the sidechain is disabled.  This tells me the problem is coming from the sidechain triode.
- BUT, removing ANY tube, including the output amp tubes eliminates the oscillation....
- Oscillation also seems to be load-dependent.  When the output is terminated into a 10k load, the HF oscillation goes away and it just starts motorboating.  Again, this ONLY channel 2!

This is DRIVING ME INSANE.  I've wasted two full days trying to chase this damn thing down.

What the hell have I missed?!
 
The enclosure is grounded, right? I'd look for miswires, and a wiring layout issue that might bring output and input signal a little too close to each other. Also, you might disable the entire sidechain circuit (pull the driver tubes, disconnect stuff) and see whether the oscillation goes away, which could be a clue. Your "switches replaced with relays" comment makes me a bit nervous, not sure of the benefit, but the additional wiring to accomplish that might have added some unintended coupling between input and output. Are there places where an input and output signal are in a cable with a common shield?
 
Thank you for the response, Mr. Kulka.  Much appreciated.  The signal PCB, the power PCB, and the IEC safety ground are all grounded to a central point on the chassis.  No signal-carrying cable shares a shield.  Pin 1 of each input is floating with a 51R/.01uF filter to chassis tab on the XLR.  Pin 1 of each output is connected directly to their respective XLR chassis tabs.

I'm using 9V relays for:
- comp/limit switch
- sidechain disconnect
- true bypass

I was doing this in an effort to keep signal from leaving the board as much as possible and to keep off-board wiring to a minimum.

Here's what I've tried since yesterday.

Test 1:

I thought maybe the proximity of the input/output points on the PCB was causing the HF oscillation via capacitive coupling, so I lifted some components and swapped output between channels 1 and 2; channel 2's output is now on the other side of the board from its input.

The oscillation is still there.  Hmmm.

Removing the sidechain tubes eliminates the 24kHz oscillation (which, by the way, is a bit closer to 30kHz today, for some reason) altogether.

Using the "sidechain disconnect" switch, which simply disconnects the ouput of the sidechain pentode from the T4, also eliminates the HF oscillation.

But now there is an infrasonic oscillation of about .25Hz.  So the HF oscillation has something to do with the sidechain, it would seem, which brings me to

Test 2:

The input to the second grid (from the 1k grid stopper) of the sidechain triode is VERY close to the pentode output on the PCB.  I lifted a component and re-routed the input to the triode as far from the pentode output as possible.  No change.



UPDATE: the problem is now on both channels 1 and 2.  The same HF oscillation and then the .25Hz oscillation when the sidechain is disconnected.

FWIW, plate B+ on channel two tubes is consistently a few volts low, but nothing particularly crazy.

Ripple on B+, which is the full-wave rectified supply taken straight from the schematic, jumps around from about 20mV up to about 40mV.

At this point, I'd almost rather just pay someone to do the layout for me and learn from that.  I really feel like I'm spinning my wheels here and getting nowhere.

Thanks again for the response!
 
Well, I'm starting to see your point about the relays.  When I flip switches on one channel, I often see (and measure) spikes on the other channel, odd oscillations and other strange things.  I have protection diodes (1N4148) across each relay, sharing the same ground plane as signal.  I'm sure this doesn't help.

BUT, an earlier version of the same board had no relays yet exhibited the same HF oscillation issues at the same frequency.  SO...yeah.  No idea.
 
Hi,
Long ago I had the same problem with volkers dual la2a. I checked out everything and was totally desperate. A few weeks later I realized by accident, that the shielded wires to the input pot were not really wired wrong, but the signal was connected to the shield. That was indeed a stupid mistake.
Cheers
Bernd
 
It happens to the best of us, Bernd!

UPDATE:  The ultimate source of the problem seems to be coming from the sidechain triode; as a test, I placed a .0068uF cap from the 1k grid stopper resistor to ground, creating a first-order RC filter with ƒc @ 23.4kHz.  No more oscillations of any type.


It's fine for now, but clearly not the "correct" answer.
 
Back
Top