After the whole power supply fiasco, I got a couple of 120VA transformers so I could have one regulated supply per channel and a lower voltage preamp supply all working correctlly. Indeed, on the power supply side of things it now outputs the correct voltages.
But, as my luck goes, I have now run in to another issue - oscillation!
The sequence of events:
- Powered on, checked power supply voltages. All OK.
- Connected LM3886 amp boards to power supply, powered on. No input signal (not shorted either). No load. No explosions, and no visible oscillation on the scope. Great stuff!
- Connect input signal. Play some basic test tones + music. No load (dummy or speaker) Seems to be working great. Ignoring my DIY preamp for the moment and using the old Cambridge pre out.
- Connect ancient worthless test speaker. "Dual cone", 4 ohm nominal. Playing music! Leave for ~30 minutes playing music.
- All seems well. Decided I would play a test tone because the bass sounded kind of wrong / distorted at higher volumes. Figured it was probably because it wasn't in an enclosure and it's a bit of a crap speaker, but I was going to look on the scope. Upon turning the volume up, probably about half way....... the distortion I was seeing looked sort of like clipping. I think it may very well have been SPIKE. There was a strange little notch, like the SPIKE picture in the datasheet. Seconds later, the speaker made some fairly horrid noises and went quiet / faint white noise. Oh no....
Quickly turned things off. Lesson #1, use a dummy load
Not too bothered, it was the test speaker after all, but it might have saved a test speaker.
So, in comes the dummy load to see what's happened. A 3.3R 2W resistor (a tad low I know). Switch on, and I see some quite nasty oscillation. My first real experience with it. It's rather obvious to see on the scope
On 2v/div setting it was about 2.. uh...div's, so I assume this means it was 2v in amplitude (or is it really 20v, because it's a 10X probe?). I think it was very likely in the MHz range, as with time/cm set to 1uS I could only barely make out how many times the wave was repeating! Probably somewhere between 10-20 times for the short time I was able to look at it.
Some observations:
- A 330R dummy load doesn't seem to oscillate, or if it is then it's no where near as bad. The amplifier still produces at least sine waves too. The resistor didn't seem to get hot.
- Apparently only oscillates when the input is shorted or has an input connected, even if it's not playing anything. I can use the 3.3R dummy load without an input and it's OK.
I have 100nF bypass caps as close to the chip as possible, and then just after I have 100uF, so I don't believe it is a bypassing issue... though I don't really know how I would check to see if it was.
I tried adding a 470pF cap between the input pins, but this doesn't seem to have helped either.
I have a Zobel on the output at 4.7R / 100nF. The schematic I used has 2.7R / 100nF, but I have seen 10R / 100nF recommended elsewhere so I figured 4.7R is probably even a bit "safer" than 2.7R..
I dont have a LR filter on the output. I will be adding one when I get around to buying some copper winding wire stuff though, if only to stop that damn mobile phone interference!
It seems like it was working fine, and then just suddenly... broke! On both channels... maybe it was oscillating the whole time, but the speaker somehow managed to survive for quite a while before giving up, I don't have another blowupable speaker to tell.
The only thing I can really think of is perhaps something to do with a poor PCB layout (quite likely) or grounding. I'm not sure if grounding could cause oscillation though. Grounding at the moment looks like this:
Should give a fairly good idea of how lousy I am at grounding The only thing that is hard to illustrate in my simplified pictures is that on the PCB the NFB ground is connected to the zobel ground via a thin trace, which is in turn connected to the power ground by a slightly thicker trace. I didn't seem to be getting any humming at all though, so I figured it was actually OK. Would the input ground connected to the Zobel via a thin trace be able to cause oscillation?
Current PCB layout (not sure how helpful this is..):
Note I am not running separate wires for SPKR_GND and SIG_GND at the moment. I didn't want to run a huge number of wires about the place really!
I would be very grateful if someone could help me track down what's causing this to oscillate so merrily!
But, as my luck goes, I have now run in to another issue - oscillation!
The sequence of events:
- Powered on, checked power supply voltages. All OK.
- Connected LM3886 amp boards to power supply, powered on. No input signal (not shorted either). No load. No explosions, and no visible oscillation on the scope. Great stuff!
- Connect input signal. Play some basic test tones + music. No load (dummy or speaker) Seems to be working great. Ignoring my DIY preamp for the moment and using the old Cambridge pre out.
- Connect ancient worthless test speaker. "Dual cone", 4 ohm nominal. Playing music! Leave for ~30 minutes playing music.
- All seems well. Decided I would play a test tone because the bass sounded kind of wrong / distorted at higher volumes. Figured it was probably because it wasn't in an enclosure and it's a bit of a crap speaker, but I was going to look on the scope. Upon turning the volume up, probably about half way....... the distortion I was seeing looked sort of like clipping. I think it may very well have been SPIKE. There was a strange little notch, like the SPIKE picture in the datasheet. Seconds later, the speaker made some fairly horrid noises and went quiet / faint white noise. Oh no....
Quickly turned things off. Lesson #1, use a dummy load
Not too bothered, it was the test speaker after all, but it might have saved a test speaker.So, in comes the dummy load to see what's happened. A 3.3R 2W resistor (a tad low I know). Switch on, and I see some quite nasty oscillation. My first real experience with it. It's rather obvious to see on the scope
On 2v/div setting it was about 2.. uh...div's, so I assume this means it was 2v in amplitude (or is it really 20v, because it's a 10X probe?). I think it was very likely in the MHz range, as with time/cm set to 1uS I could only barely make out how many times the wave was repeating! Probably somewhere between 10-20 times for the short time I was able to look at it.
Some observations:
- A 330R dummy load doesn't seem to oscillate, or if it is then it's no where near as bad. The amplifier still produces at least sine waves too. The resistor didn't seem to get hot.
- Apparently only oscillates when the input is shorted or has an input connected, even if it's not playing anything. I can use the 3.3R dummy load without an input and it's OK.
I have 100nF bypass caps as close to the chip as possible, and then just after I have 100uF, so I don't believe it is a bypassing issue... though I don't really know how I would check to see if it was.
I tried adding a 470pF cap between the input pins, but this doesn't seem to have helped either.
I have a Zobel on the output at 4.7R / 100nF. The schematic I used has 2.7R / 100nF, but I have seen 10R / 100nF recommended elsewhere so I figured 4.7R is probably even a bit "safer" than 2.7R..
I dont have a LR filter on the output. I will be adding one when I get around to buying some copper winding wire stuff though, if only to stop that damn mobile phone interference!
It seems like it was working fine, and then just suddenly... broke! On both channels... maybe it was oscillating the whole time, but the speaker somehow managed to survive for quite a while before giving up, I don't have another blowupable speaker to tell.
The only thing I can really think of is perhaps something to do with a poor PCB layout (quite likely) or grounding. I'm not sure if grounding could cause oscillation though. Grounding at the moment looks like this:
An externally hosted image should be here but it was not working when we last tested it.
Should give a fairly good idea of how lousy I am at grounding
The only thing that is hard to illustrate in my simplified pictures is that on the PCB the NFB ground is connected to the zobel ground via a thin trace, which is in turn connected to the power ground by a slightly thicker trace. I didn't seem to be getting any humming at all though, so I figured it was actually OK. Would the input ground connected to the Zobel via a thin trace be able to cause oscillation?Current PCB layout (not sure how helpful this is..):
An externally hosted image should be here but it was not working when we last tested it.
Note I am not running separate wires for SPKR_GND and SIG_GND at the moment. I didn't want to run a huge number of wires about the place really!
I would be very grateful if someone could help me track down what's causing this to oscillate so merrily!
Hi,
two things I can see:
#1: Your E$10 (10kOhm) mute pin resistor goes from pin 8 to the mute capacitor C11, but should go to -24V. C11 should go to power ground. It's possible that your amp is steadily muting and unmuting, chopping the signal.
#2: You connected the noninverting input (-) to the NC pin 11. This is no good practice, as National Semi recommends not to connect the NC pins to a low level signal. It's possible that currents in the die feed back into (-) out of phase and cause oscillation.
Both things you could try to change in a jiffy (with an exacto knife, yuck).
Hope this helps,
Sebastian.
two things I can see:
#1: Your E$10 (10kOhm) mute pin resistor goes from pin 8 to the mute capacitor C11, but should go to -24V. C11 should go to power ground. It's possible that your amp is steadily muting and unmuting, chopping the signal.
#2: You connected the noninverting input (-) to the NC pin 11. This is no good practice, as National Semi recommends not to connect the NC pins to a low level signal. It's possible that currents in the die feed back into (-) out of phase and cause oscillation.
Both things you could try to change in a jiffy (with an exacto knife, yuck).
Hope this helps,
Sebastian.
sek said:
Hi Sek,
I just checked the datasheet again and you're absolutely right. How I managed that one I don't know! I will attempt to correct this one
Thanks very much for pointing that out!
#2: You connected the noninverting input (-) to the NC pin 11. This is no good practice, as National Semi recommends not to connect the NC pins to a low level signal. It's possible that currents in the die feed back into (-) out of phase and cause oscillation.
Both things you could try to change in a jiffy (with an exacto knife, yuck).
Hope this helps,
Sebastian.
I cut off all the NC pins at the chip, so it's not really connected to pin 11
EDIT: Removed the mute cap for now, so it's just 10k to -ve rail. Unfortunately the oscillation is still present.
sek said:Just for the sake of completeness: did you just remove the cap, or did you also bridge it's position?
Sorrry yes.
To be exact I put the resistor where the cap was, and wire bridged where the resistor was. Why so much hassle? It makes it easier for me to add the cap in the proper position later (between pin 8 and ground, right?)
Nordic said:
I don't quite follow. It's definatley something wrong with the amp, it's oscillating with a pure resistive dummy load, not the speaker
Regardless, I already checked it with a multimeter and it measures ~ 4.6 ohms DC resistance. The speaker might even still work, but I dont intend to connect it to the amp again until I have got rid of the oscillations
Hi,
I wonder if the Zobel connection to signal ground is causing the oscillation. A few others with discrete amps report this problem.
The solution is separate the dirty ground from the clean ground.
The best of all is Zobel across the speaker terminals and that returns direct to star ground.
If the Zobel mod does not sort it then run three separate grounds from your PCB to star ground and cut both links. This may require you to disconnect the ground from regulator to star ground.
Can you monitor the regulator output at the reg & then at the amp while the amp is oscillating?
I wonder if the Zobel connection to signal ground is causing the oscillation. A few others with discrete amps report this problem.
The solution is separate the dirty ground from the clean ground.
The best of all is Zobel across the speaker terminals and that returns direct to star ground.
If the Zobel mod does not sort it then run three separate grounds from your PCB to star ground and cut both links. This may require you to disconnect the ground from regulator to star ground.
Can you monitor the regulator output at the reg & then at the amp while the amp is oscillating?
AndrewT said:
Hi Andrew,
Could be. I'm actually using speakon connectors and not binding posts, so it makes it that tiny bit trickier to do at the speaker terminals. I'll probably cut the traces instead...
As above, this is how I was doing things on the old board, so maybe this is actually the cause of it. I wonder if my 'power ground' copper PCB fill might have been a tad excessive.
Both at the same time? Possibly... I have two probes. It'll be kinda tricky though. Are you thinking that maybe it's the regulators oscillating?
I just tried 100nF on the underside of the board right at the pins and it didn't help, so I assume the bypassing is OK.... or I have a batch of dud / inappropriate caps. They're polyester's...
I'm taking a small break from it while I get the motivation to try, so I will try cutting traces a little later and report back.
neutron7 said:
I can't try the 22pF, as I don't have one. I have put 100nF between the -ve and +ve rail and cut the traces as Andrew suggested. The oscillation appears to have gone!
In fact, either I am missing something or it's now very clean indeed. On the old board with 1uS and no music playing I saw some sorta very tiny amplitude wave, but I put that down to just noise from the scope at the time. Now it just looks like a straight line
Now, that's very odd.markiemrboo said:
I can't try the 22pF, as I don't have one. I have put 100nF between the -ve and +ve rail and cut the traces as Andrew suggested. The oscillation appears to have gone!
In fact, either I am missing something or it's now very clean indeed. On the old board with 1uS and no music playing I saw some sorta very tiny amplitude wave, but I put that down to just noise from the scope at the time. Now it just looks like a straight line
The chipamp experts and users have been designing their PCBs with the grounds connected together and telling me I was talking balderdash, although not quite so bluntly.
Maybe running the different types of returns to an off board audio ground is the better way to do it.
I can confirm that off board star (audio) ground will make multi-channel easier.
AndrewT said:
Well, I can definately confirm that the oscillation has now disappeared. The dummy load no longer starts to smoke after being on for all of ~3 seconds with the input shorted either
So thanks for the suggestion! 
I dont actually doubt that having the grounds connected together works fine.... on a well laid out board - something which my board is obviously not
I may well try connecting the Zobel return back to the power ground, but leaving the input return separate and seeing if it comes back.
BWRX said:
Yes. I was using 'thermals' for the polygon fill, but you probably can't see it very well in the screenshot.
While I am here, I dont suppose anyone would be nice enough to explain the rough effect of the big 3.3uF cap between the +ve and -ve rail? I'm thinking of removing it. I can make the PCB's quite a bit smaller without it...
Just thought I would report back. I tried with just the NFB return separate from power and zobel, but reconnected the zobel return to power ground (if you follow) and it's no worse than having them all separate, so I think I am going to keep it like this!
I tested with the 3.3R and even a 0.47R and it was fine. In fact, for some reason a 40Hz square wave seemed to 'perform' better in to the 0.47R, it didn't seem to slope quite as much (it probably shouldn't slope at all? and if it shouldn't I think I am going to blame that on my sound card for now..) ... which doesn't seem right, but oh well
I tested with the 3.3R and even a 0.47R and it was fine. In fact, for some reason a 40Hz square wave seemed to 'perform' better in to the 0.47R, it didn't seem to slope quite as much (it probably shouldn't slope at all? and if it shouldn't I think I am going to blame that on my sound card for now..) ... which doesn't seem right, but oh well
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.