I'm trying to identify the source of parasitic oscillation. The amp topology is fairly standard "Blamless Amp" (CF) Except that I'n using separate CCs for the IS and VAS section. Simulation says it should be stable so I suspect something less than obvious.
For test purposes input and output are shielded (BNC) and are at come off opposite sides of the PCB. Grounding is standard Star ground that has always worked for me in the past. On the PCB the IS and VAS are at opposite ends of the PCB with as much distance as I can put between then in a 6"x6" PCB. Heat sinks are mounted on the PCB but are grounded. Signal ground and power ground are separate.
One thought: This is my first shot at a double sided PCB with the top layer devoted 90% to being just the Ground Plane. Could it be an antenna??
There is no sign of oscilation when no signal is present, i.e nothing connected. However, there are small signs when anything is connected - even if the input is shorted to signal ground. It is present unloaded and without a signal but very small (6mV AC) at the outputs. Attaching a dummy load doesn't change this.
On the 'scope the oscillation increases as the signal increases. All this is with the compensation caps doubled in value (reduces the problem but doesn't make it go away.) Sine waves seem to be behaving normally except for the oscillation fuzz on them. Infact, an audio signal sounds normal (the crap is a ways above 20kHz.)
A second thought: I've laid this out in LT-Spice using MJE340/350 pre drivers. An FFT plot show very small artifacts in the band where the oscillations seem to concentrate. Changing to 2sd669/2sb649 makes them go away leaving a very smooth plot. I'm doubtfull these means much given the uncertain quailty of the models.
More interesting is that a small change to the OPS topology also cleans up the FFT plot. Instead of connecting the emmitters of the predrivers directly to the collectors of the (parralleled) output devices (between the collectors and the usual .22-ohm power resistors), --I terminate the emmitters with 4.7-ohm resistors and then connect those to the output/NFB point between the two power resistors.
I don't expect a solution but any suggestions on where to look. I'm 95% sure the usual suspects - grounding, unshielded inputs, IS to close to OPS are not the problem. I'm leaning toward use of the ground plan on top layer or the topology sitiatuion describe above.
All thought are welcome.
For test purposes input and output are shielded (BNC) and are at come off opposite sides of the PCB. Grounding is standard Star ground that has always worked for me in the past. On the PCB the IS and VAS are at opposite ends of the PCB with as much distance as I can put between then in a 6"x6" PCB. Heat sinks are mounted on the PCB but are grounded. Signal ground and power ground are separate.
One thought: This is my first shot at a double sided PCB with the top layer devoted 90% to being just the Ground Plane. Could it be an antenna??
There is no sign of oscilation when no signal is present, i.e nothing connected. However, there are small signs when anything is connected - even if the input is shorted to signal ground. It is present unloaded and without a signal but very small (6mV AC) at the outputs. Attaching a dummy load doesn't change this.
On the 'scope the oscillation increases as the signal increases. All this is with the compensation caps doubled in value (reduces the problem but doesn't make it go away.) Sine waves seem to be behaving normally except for the oscillation fuzz on them. Infact, an audio signal sounds normal (the crap is a ways above 20kHz.)
A second thought: I've laid this out in LT-Spice using MJE340/350 pre drivers. An FFT plot show very small artifacts in the band where the oscillations seem to concentrate. Changing to 2sd669/2sb649 makes them go away leaving a very smooth plot. I'm doubtfull these means much given the uncertain quailty of the models.
More interesting is that a small change to the OPS topology also cleans up the FFT plot. Instead of connecting the emmitters of the predrivers directly to the collectors of the (parralleled) output devices (between the collectors and the usual .22-ohm power resistors), --I terminate the emmitters with 4.7-ohm resistors and then connect those to the output/NFB point between the two power resistors.
I don't expect a solution but any suggestions on where to look. I'm 95% sure the usual suspects - grounding, unshielded inputs, IS to close to OPS are not the problem. I'm leaning toward use of the ground plan on top layer or the topology sitiatuion describe above.
All thought are welcome.
sam9 said:
A while back I built one of Self's Blameless type designs (what he refers to as load-invariant.) Had low level oscillations around the 600 kHz region. Tried a few things suggested by forum members including compensation cap values and o/p stage emitter resistor values. Also removed the output protection circuit but couldn't eliminate the oscillations and so the amp is parked on the shelf until I have time to go back to it. Unfortunately the thread I was discussing this in (semi-hijacked unfortunately - sorry!) was locked by the moderators due to a flame war and a conclusion wasn't reached. In the maintime I've been working on another design.
As a point of reference I made a double sided pcb layout but didn't use a top surface ground plane which may suggest that this is not the source of your troubles. Have you tried a single output pair?
James
I built one of Slones amps that was clearly related to the Blamless configuration. It had only one output pair. There were some oscillation problems with it too until I revised the PCB to move the OPS away from the input - hence my attention to that in the current layout. Discinnecting one pair is not something I've thought of but will giv it a try. At least that will tend to identify the approximate location of the problem.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.