Diagnosis needed!

I have a pair of LM3875 based amps used as a stereo pair, with a single transformer/supply powering the pair. I am very happy with the sound...but there is a problem.

About 10 minutes into a listening session, I begin to hear faint distortion, mostly in (but definitely not limited to) the vocal range. The distortion gradually increases until it is audible even to my girlfirend ;) (she's just not into good sound...weird).

Here is the weird part. I can increase the volume to near max for a few seconds, return it to normal levels, and the distortion "disappears" for several minutes. Also, it seems that the lower the listening volume, the more likely I am to experience the problem (it's not just the volume covering it up at loud volumes). BTW, the LM3875's are barely warm at all times. I am using an 18-0-18v 6A toroid, so it is not getting too much juice.

Have you ever heard of a problem like this? Could it be dirty AC in my apartment? I'm new in this apartment, so if this amp is telling me something, I would like to look further into the AC thing. Any ideas?

Thanks for your help and patience!

f4
 
Marginal stability

I would rather suspect that you have an instability problem that induces a slight amount of high frequency oscillation as the chips warm up. This instability is likely on the verge of not been there when cold, but crosses the point were it is triggered on when warm. Changing the volume may cause this oscillation to stop, but since it so close to being there all the time it will come back when the trigger point is reached again. It may be there at lower lower all the time but well above the range of hearing.

I would suggest looking at the output with a scope to see if you can spot it. This is the only likely cause in my opinion.

John Fassotte
Alaskan Audio
 
Thanks for the quick replies!

John,
You might be right...your description is accurate. It is very possible that it is there all along and a threshold is crossed. Unfortunately, I don't have access to a 'scope, but I will look into adding some stability.

Duo,
The design is basically an application-sheet design (see below), but it is in inverted configuration.

UPDATE! I just listened to the amp for 20 minutes until the "enemy" attacked Steely Dan - The Royal Scam. This time, I turned the power off for 15 seconds, and then turned it back on. The distortion was not audible for another 10 minutes. Also, if I turned the source off (silence) for 30 minutes and then played the source, it was VERY distorted. But when I gradually increased the volume to a low listening level, the distortion decreased (or it was masked??? I can't figure that out).

Everything is point-to-point with a very tight layout and a star ground. The grounds for the power caps, signal, output, and pot all connect at star ground........could that be part of the problem??

I am hearing the same problem from both channels simultaneously. I took great care in making the joints nice and clean. I measured 0mv of dc offset.

Any tips? It is frustrating, because it sounds GREAT until the problems start. 20 minutes of listening won't cut it. :eek:

Thanks again for your help, and sorry about the long posts.

-f4
 
Pic didn't work...

Here is the non-inverting version from the spec-sheet:
 

Attachments

  • lm3875.jpg
    lm3875.jpg
    23.9 KB · Views: 276
seems really weird if both chips are doing this...
Are you sure all you capacitors and resistors are the right values and int the right places? Are all the electrolytic caps in the right way around?

I don't think a star topology for the grounding could cause this...
Did you check with your meter to see that the power supply voltage stayed the same for the time it took to start distorting??
Or put a meter in series with the amp and PSU and leave the amp idle to see if the current draw changes after a while??


EDIT: i just checked the prev messages in the thread...
Does the spec sheet for that chip suggest any kind of inductor
or zobel network of some sort on the output? It could be very likely that, due to a reactive speaker cable, your amp is oscillating at high frequencies, generating heat in the OT devices in the chip since they cant turn off fast enough at that frequency. In other words, you might want to try putting zobel network on the output of the amp before the speaker, it consists of a resistor of about 10 ohms and a capacitor of about 100nF in series from the OT to common ground. If that fixes the problem, then your amp almost certainly was oscillating:)
 
Here's a pic of your amp with a zobel net attached, it's really easy to add on, heh, you could even add the network right at the speaker jack if you didn't want to open the box if it has one...


EDIT: never mind the caps labeled "1n" on the power supply rails, my diagram editor just picks that as a starter value and I forgot to change the them;)
 

Attachments

  • zobel.jpg
    zobel.jpg
    16 KB · Views: 264
Hi, f4
How old is your amp ?
Is your volume potentiometer in good condition ?
I met similar effects coming from dirty pots, that were fixed by using the "magic" contact cleaning spray.... or a new pot !
Also check connections to the speakers, from the source, and the input switch... Urban atmospheres are quite corrosives and dirty contacts not good for sound :(
 
Thanks, everyone!

This place is great. Thanks for all the good advice.

Duo,
I will give the zobel a try. I was being stubborn with my minimalist approach, so that simple solution might do the trick. I am only using about 3 feet of speaker wire right now, but it is junk wire. :rolleyes:

vbd,
The amp and all its parts are about 6 months old, so it should be relatively clean. I have, however, dealt with the urban corrosion problem in the past with some older equipment...it makes you wonder if breathing this stuff is a good idea! :eek:

Thanks again, and I will post with any good news.

-f4
 
Your welcome:)

Also, it doesn't really matter what quality your speaker wire is...
I've heard of many oscillation problems even with the most expensive speaker cables. Either way, cable is cable, metal and plastic and rubber. It will always support some kind of oscillation under the right circumstances. And the length would determine the resonant frequency, along with the impedance of the cable itself. So anyway, good luck, I hope that's all the problem is;)
 
Hi Bill,
I am not using a separate pre-amp...just a signal fed through a pot to the amp. I have tried the amp with a Toshiba SD-3750 DVD player (my main source), a Sony Discman, and an old Fisher 25 disc changer. The problem presentation was identical with all three sources, so I ruled that out.

I just did a refresher course on Zobel networks, and I think it might solve my problems. If I am understanding this correctly, high frequency oscillations can occur when there is an absence of load at high frequencies. This is consistent with my "purge the system" volume burst that eliminates the symptoms temporarily. The distortion seems to appear the most when the signal level is at its lowest. An RC Zobel network can reduce these oscillations. So does adding an inductor in series with a resistor at the output just cut off the high frequencies at a level lower than the oscillation? Which one of these fixes would most adequately address my problem? (I would ask which solution would degrade the sound quality the least, but at this point I'll take any clean sound I can get! :cool: )

Sorry to bother everyone with elementary electronics, and thanks again for the help. I should have the parts in hand tommorrow to try the Zobel.

-f4
 
Hmm, just read up on this thread again. You can use an inductor and resistor as well, some people even put it in with a zobel at the same time. The zobel however, will deliver higher sound quality and is used by many of the better amp designers. Though, you can design an inductance network that won't really reduce quality, but it's a bit harder and I wouldnt consider it worth it.
 
Some progress!

I haven't put in the Zobel yet, but I did a little snooping with a jumper wire. Here's what I found:

Each channel has its own metal chassis, but both are fed by the same power supply. Also, both metal chassis's are housed in a single wooden enclosure that contains a dual-ganged pot (I think that is what it's called) and the PS. When I ran a length of wire between the star ground of each amp, I was able to listen distortion free...1 hour and counting!

I'm not sure what was going on. I assumed that the grounds on the dual pot were isolated from each other. If they are not, that may have been a source for a ground loop, I think. Other than the PS, that was the only possible common connection.

I will still try the Zobel network for stability's sake. There is a slight hum at full volume/shorted-input, but it is 1000x better than before. And it is consistent...no changes during a listening session. It sounds great! :cool: Thanks again for everyone's help.

-f4