Chipamp working OK, then bad distortion

[Gecko123]

Well I constructed my lm3875 chip amp the other day, had some slight hum issues which I still need to figure out, but when you play music you can barely hear the hum. So I suppose I was driving it fairly hard, but I wanted to see how hot the heatsinks would get, and the sound was not distorting, then about an hour into playing tunes, the speakers just belched out really loud and bad distortion, I thought my speakers were toast, but they were fine after I checked em. So what could have caused the issue? The heatsinks were barely warmer than my hand at the time, I measured the secondary voltages and DC offset and all was good, exactly as specced in the chipamp building tutorial, very curious what the issue was, I will post pics soon so you can see my setup.

 

[Atilla]

If DC is fine then it could be problems with the PSU - bad connections, bad caps, or it could be thermal protection kicking in.

Have you tried touching the chips, are they mounted on those heatsinks nice and tight, using paste?

Can you tell us more about your setup - transformer rating, capacitance, rail voltage, etc.. What are the voltage limits of your PSU caps?

Check for bad solder connections, lose wires, or anything that might cause a component to disconnect or and/or make a short circuit.

Finally, pics will help people identify potential problems as well.

 

[Gecko123]

Here's some pics

http://i524.photobucket.com/albums/cc323/Jordo_01_2008/IMG_0361.jpg
http://i524.photobucket.com/albums/cc323/Jordo_01_2008/IMG_0362.jpg
http://i524.photobucket.com/albums/cc323/Jordo_01_2008/IMG_0363.jpg
http://i524.photobucket.com/albums/cc323/Jordo_01_2008/IMG_0364.jpg
http://i524.photobucket.com/albums/cc323/Jordo_01_2008/IMG_0365.jpg

So the transformer is Plitron 300VA, 2x25V, the kit is the classic kit from audiosector.com
http://www.audiosector.com/lm3875.shtml
So i guess the capacitance would be 3000uf per channel. Secondary voltages came out at right around 30V or just over, cant remember exactly, since I have the 10uf smoothing caps installed on rectifier board, and I dont have a volume pot installed so the voltages at the speaker terminals were around 69-70mV.

 

[Gecko123]

Ok heres some more recent data.
Right channel at speaker terminals - 72.6mV
Left channel at speaker terminals - 86.9 mV

V- to PG- 32.2mV
V+ to PG+ 32.2mV

And the solid copper wire you see there is connected from one channel board OG to the other channel board OG, and I have left speaker(-), Right speaker(-), PG-, PG+ all connected to that, and then that is connected to the chassis grd. Why do I still have hum as well? and do those speaker terminal voltages look right? seem high. thanks for anyones help.

 

[Redshift187]

You mean 32.2 V, not mV right?

 

[pacificblue]

Sounds like the protection system kicking in. Either too hot or clipping. Probably too hot. A 2x25 V transformer gives quite high rails for the isolated package even with the enormous heatink you have. The IC simply cannot transfer the heat fast enough to that heatsink.
You have three options.
- Replace the LM3875TF with the LM3875T.
- Add a fan.
- Reduce the volume.

 

[Atilla]

What impedance are your speakers? Are you trying to drive 4ohm speakers on 32V rails? That can get the chip very hot, very fast.

I've a LM3886 setup with much smaller heatsinks though, and with those and slightly voltages I don't seem to have any problems. I'm using the TF packages as well. Is lm3875 worse at handling heat?

 

[martinbls]

Quote: "And the solid copper wire you see there is connected from one channel board OG to the other channel board OG, and I have left speaker(-), Right speaker(-), PG-, PG+ all connected to that, and then that is connected to the chassis grd."

Are you sure that this is the right way to do it? From what I read I guess only CHG of both amp boards have to be connected to chassis ground (earth), perhaps with a 10R resistor. OG is for speaker minus rail only, and PG- and PG+ have to be connected only to PG- and PG+ at rectifier board. Perhaps it's worth a try?

I know that there is already a connection between OG of speaker and CHG on amp boards, but the exact way you connect all this might be importand, especially when it comes to hum problems.

Regards - martin

 

[martinbls]

Hi,

I'm sorry, but perhaps I'm wrong! Just found this post

http://www.diyaudio.com/forums/show...7773#post787773

and I think this is exactly the way you did it and it's supposed to work well. But this concept of grounding is new to me, I thought I read somewhere that only CHG of both boards have to be connected to chassis ground...

Well, now I'm asking myself how I am going to connect all this on MY amp project! Funny...

Regards - martin

 

[martinbls]

Sorry again, link doesn't work.

http://www.diyaudio.com/forums/showthread.php?postid=787773#post787773

 

[Yoshy]

Sounds like the protection system kicking in. Either too hot or clipping. Probably too hot. A 2x25 V transformer gives quite high rails for the isolated package even with the enormous heatink you have. The IC simply cannot transfer the heat fast enough to that heatsink.
You have three options.
- Replace the LM3875TF with the LM3875T.
- Add a fan.
- Reduce the volume.

Also, verify that the IC can transfer heat to the heatsink, make sure the thermal geese is present and didn't melt or burn away (or some new audiophile patented method that we don't know about )

 

[pacificblue]

I've a LM3886 setup with much smaller heatsinks though, and with those and slightly voltages I don't seem to have any problems. I'm using the TF packages as well. Is lm3875 worse at handling heat?

LM3875 and LM3886 can handle the same amount of heat. The TF-package is worse at distributing it. The T-package has a thermal resistance of 1 K/W and a Pdmax of 125 W at 25 °C, the TF-package 2 K/W and 62,5 W at 25 °C. In other words, the TF package can only dissipate half as much power at any given temperature and is only half as efficient at dissipating a given amount of heat.

You can also have less problems, because your supply voltage is lower or your speaker impedance higher or you listen at lower volume or your speakers are more efficient or because they pose a less complex load. The nominal designation 4 or 8 Ohm does not tell the whole story. Inductive components, like voice-coils have a complex behaviour. Effects like self-induction and reactance can load an amplifier significantly more than the nominal impedance implies. Phase shift that is inherent in all non-Ohmian loads does not make things better. So maybe you are lucky with your speakers, while Gecko123 has more demanding ones.

 

[Atilla]

Yes, fair enough, every load is different indeed. I've yet to try to run them *real* hard though, I don't expect them to handle anything crazy, that's for sure. Resistive load was a joke to handle, small test speakers went also alright, they were going to melt long before the amp heated up at all. But I haven't run them for too long and too loud in a big 3-way speaker with crossover yet, mostly because I risk to deafen myself or get evicted

 

[Gecko123]

yeah, my bad.

 

[Redshift187]

If the heatsink is not even warm, I would suggest you have a poor connection between the heatsink and the chip.

 

[Peter Daniel]

And the solid copper wire you see there is connected from one channel board OG to the other channel board OG, and I have left speaker(-), Right speaker(-), PG-, PG+ all connected to that, and then that is connected to the chassis grd. Why do I still have hum as well? and do those speaker terminal voltages look right? seem high. thanks for anyones help.

Connect all grounds in exactly one point on the copper wire, as you have it now PGs are spaced into 4 separate points and ground connections to the speakers are at opposite ends: connect all 6 wires in one central point.

Since you are already connecting copper wire to the chassis, there is no reason to runs separate wires from CHG to the chassis: remove them.

Make sure that all your connections are fine and chips are properly mounted on a heatsink: that means smooth contact surface and very little paste, tighten the screw reasonably hard.

Sounds like the protection system kicking in.

There is no reason to replace isolated chip package with non isolated. With properly built amp and even 40V rails, 4 ohm speakers should not be a problem. I'm talking from experience using such system for a year.

 

[Gecko123]

I will try some of these things and see what happens, but i'm gone to mexico for a week now, so probably in a week I can play with it more. I think I was just running too much juice to the little amp, I mean my speakers were playing LOUD, I didnt even want to be in the same room at the time, but it wasnt distorting till after the hour or so. The speakers do dip into 4 ohm territory, so I was definitely driving the amp hard. There is still paste there so it didn't melt off or anything, and it is screwed tight to the heatsink. I think it was just an overdriving issue, but now I just gotta rework the grounding and I should be good to go Thanks for all the help! Btw the speakers are 88db efficient
http://i524.photobucket.com/albums/cc323/Jordo_01_2008/IMG_0270.jpg
the look like that, and I had my Onkyo 805 running as a preamp, with volume up to -10db(it hurts the ears) So im not sure how much wattage I was actually sending the amp.

 

[AndrewT]

With properly built amp and even 40V rails, 4 ohm speakers should not be a problem. I'm talking from experience using such system for a year.

Hi,
could you confirm the speaker sensitivity and the smoothing capacitance that are fitted to that working amplifier?

What voltage do the supply rails fall to when full power testing is carried out?
How much ripple is on the supply rails when doing full power testing?

Do you ever ask your amp to deliver a full voltage transient?

 

[Gecko123]

http://www.geocities.com/cc00541/statementresponsesplots.JPG

that is the speaker response with the speakers I was driving.

 

[Peter Daniel]

Hi,
could you confirm the speaker sensitivity and the smoothing capacitance that are fitted to that working amplifier?

What voltage do the supply rails fall to when full power testing is carried out?
How much ripple is on the supply rails when doing full power testing?

Do you ever ask your amp to deliver a full voltage transient?

I was using 3 x 12V, 7.2Ah batteries per rail, which produced approx 39V DC, there was not much voltage drop at full power.

Smoothing caps were BG N 1000/50. Speakers: FAL, 4ohm, 96dB efficient.

If that doesn't sound comfortable enough for you, I later switched to bridged configuration, using 300VA transformer per channels and 1000/50 smoothing caps. That means 2 ohm load per isolated cheap, still no problem, even with very limited heatsinking. And that was for almost half a year, do you think that during that time I could resist not to ask for a full voltage transient?

I got reports from Srajan Ebaen that his was using his Patek amps, also in bridged configuration, with those speakers: http://6moons.com/audioreviews/markdaniel3/ruby_2.html which are pretty hard load, also without a problem.

We still don't know what Gecko's asking from his amps, so why you bother what I'm asking?