BWRX said:
On the very first board I tried lots of different things. I removed the op-amps from the circuit (one at a time or both) to try to determine if the noise was from the input stage or the filter stage or the amp stage. This gave me inconclusive results - each time was different. I left the op-amps in but cut the +/-15 V power.
I measured the input offset.
I put a 100pF cap across the inverting and non inverting inputs of each op-amp and the LM3886.
I tried wiggling the op-amps in their sockets to check for poor contact.
I changed op-amps for new ones.
I removed the 10uF/50V electros for supply bypass on the +/-15 v rails.
I switched from the the cut down +/-15 v supply to a separate non-regulated +/-15 volt supply.
None of the above had any effect at all.
One thing I did do on the first board was to determine better values for the volume control buffer stage. Using a 110K slide pot, I changed the the values of R14, R15 and R16 from 15K, 10K, and 10K to 110K, 10K and 5K. This gave smoother level adjustment and also reduced gain in this stage down to 1.5 instead of 2. It was probably a coincidence that the noise stopped at that time.
trats said:MJL21193,
Have you checked the dc offsets on each opamp? I'd a similar problem at the beginning, the offset varied in one opamp in the mid xover and went up as high as 4V and the reg PSU voltage dropped like a ton of bricks from +/- 18V to around 10v or so.
I have not checked each individual op-amp output but will do that on the next board (a mid pass).
I have not noticed any supply lag, but the +/-15 V rails are cut down from the +/-35 rails with resistors and zeners so there's a strong current reserve.
Thanks for the help.
AndrewT said:
I have had the Spike protection kick in when I ran the first one without the heatsink. I just cuts off, a few seconds later it comes on again, gets too hot and shuts off again. No noise involved.
As I've said before, I'm stumped as to the reason for this noise. No signs of instability or oscillation and superlative performance after it has settled down.
Before going to bed last night, I took the one out of the freezer and tried it. There was still frost on the heatsink when I hooked it up. No noise. DC offset without a signal ~5mV.
TimA said:
The problem I'm having with the noise (DC spikes as high as 20 volts) and the fact that I can't predict if it will come back makes the cap manditory. First order highpass at 2000Hz for a 6 ohm tweeter is around 11uF. This puts it well below the 3100Hz crossover point for the active amp.
The intended tweeter for these is about $130.00 so I'll err on the side of caution.
I'm waiting to hear how bad the on/off transients are before I implement any muting. If it's not too bad and given the fact that the unit will be left on most of the time, I might not bother.
Working with the one that was in the freezer (it's warmed up nicely), I measured the DC offset at the output of the first stage - the unity gain buffer. It's reading a steady 25mV. At the output from the low pass filter I have 35mV. I have the DC blocking cap (2.2uF) between the volume buffer stage and the chipamp input.
Putting this in the freezer certainly didn't damage it but rather seems to have improved it. The measured DC offset at the speaker leads at full volume is less than 2mV. This is 3 to 4 times better than my Yamaha HTR which has a DC offset of 6 to 8mV a idle with no signal.
I'm impressed.
Putting this in the freezer certainly didn't damage it but rather seems to have improved it. The measured DC offset at the speaker leads at full volume is less than 2mV. This is 3 to 4 times better than my Yamaha HTR which has a DC offset of 6 to 8mV a idle with no signal.
I'm impressed.
OzMikeH said:Cryogenic treatment convert now? Never thought I'd see the day that happened John!
What next from the Machina Dynamica catalog?
Now, now. There's no denying the spectacular results!
The freezer "burn in" is a proven technique for all aspiring high end GCs.
The offset of the high pass board matches this at full output, as I'm sure the others will eventually, or close enough to make no difference.
One of the often mentioned problems with these chipamps is the amount of DC offset and level of hum. Great lengths have been gone to to reduce these, especially in the power supply. You really can't get a rougher power supply than the one I have rigged up, but I'm getting these results.
Kind of proves something.
Latest progress. Completed the mid pass amp for the "right" channel. While testing, this one would crackle so violently, I had to disconnect the power. It would often break into oscillation and have the output go to full rail voltage.
I tried everything, with no improvements. I finally gave up and started to strip the board.
One of the things I tried (and previously too) was to remove the op-amps and disconnect the 15V supply.
I now have the very strong belief that this problem exists in the amp stage and not in the filter/buffer stages.
To this end, I started to populate another board, with just the bare minimum components to make the LM3886 work. I even left off the bypass caps. Also, I was careful to check each component before installing.
Connect power. It's crackling. This dispels any doubt about where it's happening. I try a few different values for bypass, getting to the point where the crackling stops. The bypass caps at this point are 150nF poly-film. To test to see if these caps are the reason for the crackling stopping, I remove them. Connect the power and guess what? No crackling. There is some sort of ghost in the machine here.
So it's not crackling with no bypass caps. I do things to "irritate" it like touching the input pin and injecting an audio signal from the tip of the RCA plug. It doesn't crackle or oscillate.
What can the cause of the crackling be in this simplest implementation? I have one (far-fetched) idea: referring to the original schematic in post #2, would reversing the layout of C11 and R20 make a difference? Consider that there might be a fault with these caps (22uF, 50V) would the 1K resistor isolate the feedback path from C11?
Just me, grasping at straws,
I tried everything, with no improvements. I finally gave up and started to strip the board.
One of the things I tried (and previously too) was to remove the op-amps and disconnect the 15V supply.
I now have the very strong belief that this problem exists in the amp stage and not in the filter/buffer stages.
To this end, I started to populate another board, with just the bare minimum components to make the LM3886 work. I even left off the bypass caps. Also, I was careful to check each component before installing.
Connect power. It's crackling. This dispels any doubt about where it's happening. I try a few different values for bypass, getting to the point where the crackling stops. The bypass caps at this point are 150nF poly-film. To test to see if these caps are the reason for the crackling stopping, I remove them. Connect the power and guess what? No crackling. There is some sort of ghost in the machine here.
So it's not crackling with no bypass caps. I do things to "irritate" it like touching the input pin and injecting an audio signal from the tip of the RCA plug. It doesn't crackle or oscillate.
What can the cause of the crackling be in this simplest implementation? I have one (far-fetched) idea: referring to the original schematic in post #2, would reversing the layout of C11 and R20 make a difference? Consider that there might be a fault with these caps (22uF, 50V) would the 1K resistor isolate the feedback path from C11?
Just me, grasping at straws,
MJL21193 said:
I haven't followed this thread completely and I haven't looked really carefully at your board layout, but it sounds like you have a poor ground design. If you read the National data sheet on the LM3886, in the section on "LAYOUT, GROUND LOOPS AND STABILITY", they recommend using 3 separate ground paths: load ground, the output compensation ground, and the low level (feedback and input) grounds, all tied together at a common point ("star ground"). I've made a number of LM3886 circuits that oscillated when I tried to get by with a single ground as your schematic shows--that may be what you are seeing.
There is a proven layout at the link below that is similar to your circuit, although it uses the dual LM4766 instead of 2 LM3886's, so it has less output. It is circuit #1. It uses SIP resistor packs to set the 4-pole L-R crossover frequency, and there is a tab on the parts list worksheet to compute the capacitor values for various crossover frequencies. I've probably got an extra board laying around if you wanted to try it out. Notice the "star" grounding scheme in the picture, just below the LM4766--that is where all of those separate grounds come together.
An externally hosted image should be here but it was not working when we last tested it.
http://home.comcast.net/~neilrdavis/Plate_amps/
Neil Davis said:
Hi Neil,
My layout actually does use a separate grounding scheme. The fact that there are 4 fully functional boards rules out a fault in the grounding layout.
As much as I appreciate your offer, I have to work through this problem and try to understand the reason for it. I mentioned earlier, this problem goes away on it's own after a while, but what is the source of it? I have never had a problem fix itself before.
This latest mid pass unit would be the sole exception, the problem didn't go away, but actually got worse.
AndrewT said:
Hi Andrew,
Those (C20, C21) would be in the filter/buffer stage on the 15V supply. The bypass caps I'm referring to are in the amp stage (as pictured below), specifically C13, C15. I left the 220uF electros C14, C16 off altogether.
In the picture, arrows point to C11 and R20. Should these be reversed?
Attachments
Hi,
you should not need to reverse the positions of C11 & R21. They will work either way.
Omitting the electrolytics from the local decoupling could give rise to an oscillation problem.
Where does c17 terminate? could this contaminate a clean ground?
Similarly, C14 & C15 should terminate together and on the power ground using very short routes on both the power side and on the ground side.
What is PIN7? Is it allowed to terminate on the signal/clean ground?
you should not need to reverse the positions of C11 & R21. They will work either way.
Omitting the electrolytics from the local decoupling could give rise to an oscillation problem.
Where does c17 terminate? could this contaminate a clean ground?
Similarly, C14 & C15 should terminate together and on the power ground using very short routes on both the power side and on the ground side.
What is PIN7? Is it allowed to terminate on the signal/clean ground?
AndrewT said:
This was for testing only. I'm trying to back the problem into a corner, by using as few components as possible.
AndrewT said:
C17 grounds on the main power ground, as do C13 to C16 (very short leads)
AndrewT said:
Pin 7 is also attached to the main power ground. I re-routed this on my test board to the signal ground and it made no difference. Likewise, the boards that first exhibited the crackling sound, pin 7 goes to main power ground on these too. These have not shown any bad behavior since the crackling has stopped.
Here's what I just did: I re-built the power supply with a new bridge rectifier and new 6800uF caps. This made no difference.
I am beginning to think that there is some problem with the circuit board itself. Really, it's the only thing left. I have built other LM3886 amps without boards and didn't have this problem.
To test this I will go ahead and build on on the heatsink, wired point to point.
Ok, it's the boards. I did a rough point to point without any decoupling caps, connect the power - silence. No crackling, no hum.
Is there a possibility that the board substrate (fiberglass?) has impurities that are causing this problem?
I noticed arcing in a couple of the other boards before, but assumed it was some solder specks or other conductive debris, although cleaning didn't immediately fix the problem, it did eventually go away.
This means that each of the boards that I finished has some fault within the board itself. It cannot be a layout problem or the noise would persist.
Here's a photo of my quick test. Rough enough?
Is there a possibility that the board substrate (fiberglass?) has impurities that are causing this problem?
I noticed arcing in a couple of the other boards before, but assumed it was some solder specks or other conductive debris, although cleaning didn't immediately fix the problem, it did eventually go away.
This means that each of the boards that I finished has some fault within the board itself. It cannot be a layout problem or the noise would persist.
Here's a photo of my quick test. Rough enough?
Attachments
MJL21193 said:
??Doesn't sound like you have really ruled out the board layout.
I looked more carefully at your board layout and it is not "optimal". I don't know whether it is the board layout that is actually causing your problem, but the layout could certainly be improved. You should at least use two separate ground lines for the LM3886 circuitry, connected together at one point. If you follow the following 5 rules the LM3886 will not oscillate--guaranteed.
Look at the circuit board layout for the LM4780 to see how they followed these rules:
LM4870 app note
1. Use a separate power supply and signal ground for the LM3886
2. Connect the speaker return to the power supply ground, not the signal ground. The power supply ground between the main filter capacitors is the ideal connection point for the speaker return
3. Return the snubber circuit to the power supply ground, not the signal ground
4. Connect the power supply decoupling caps to the power supply ground, not the signal ground
5. Connect the signal ground and power supply ground with a small resistor to ensure that heavy current doesn't flow in the signal ground. 1 to 5 ohms will work. However, if you have a perfect layout this resistor isn't necessary.
Again, this is general advice and it may or may not account for the problem you are seeing. However, the fact that removing the decoupling caps on the supply changes your symtoms makes me wonder whether your grounding scheme is the problem. If that is in fact your problem, then welcome to the club--I have done this many times, as have many others.
Good luck.
PS: get rid of R24
Neil Davis said:
Hi Neil, I appreciate your attention here.
First, oscillation in not really the problem I'm having. It's this staticy cracking noise when I first finish a board. With my point to point construction above, I established that it is a fault in the board (but not the layout as the other completed boards would have the noise problem continue)
I had some good advice during the design process, and although the layout might not be optimal, it's far and away better than I would have done on my own. The proof is in the pudding though and as detailed above, I have 4 units which are performing exceedingly well - low DC offset (<20mV) and no hum or evidence of ground loops.
The signal supply ground is isolated from the power ground on the board. I changed this back in post# 54. These grounds are separated by a low value resistor (50 ohms).
Speaker grounds in the chassis connect direct to the star ground.
R24 has an inductor wrapped around it and it's supposed to eliminate HF oscillation. It's actual resistance is not 10 ohms but closer to 0 ohms. It's there to keep the purists happy.
I am at a loss as far as the boards are concerned. How do I determine the problem? I've measured the capacitance between traces and resistance. This is on a clean board though, it might not have the same fault as the last one.
It's discouraging, to say the least. 2/3 of the project is done and I'm afraid to complete another board and deal with the same problem.
MJL21193 said:
Thanks for understanding that I'm not trying to find fault--just helping you get avoid some of the painful learning experiences that slowed me down in the past.
I had looked at the layout in #54. This board layout might work with some chips, but it violates the rules that I outlined in the previous post because the grounds for the LM3886 are all tied together. This can create large currents through these traces that can feed back into the input and cause oscillation and odd behavior, such as hum or even crackling noises. Again, I'm not saying that this is your problem, but it is something that you should address in a future board design and it MIGHT be your problem.
Again, good luck and hang in. I'm sure this is frustrating to you.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.