LM4780 heavy distortion at low levels

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.

ra7

Member
Joined 2009
Paid Member
Hey guys,

I'm building Peter's LM4780 kit. I've had quite a few problems finishing this amp. Right now, I have it on a mdf board, with temporary connections. I have a fairly large heatsink (I think it is from conrad): 12x5. And the heatsink gets pretty hot from just one of the LM4780 chips. Currently, only one channel is running.

The toroid is about 160VA. The supply is basic with the nominal 10uF caps. I do have Rz and Cz installed.

I've tried to twist the wires as much as possible. When the amp starts up, there is a high freq. hiss, and then it becomes silent after a couple of seconds. However, when I play some music, it starts distorting very badly, at very low levels. The behavior is curious, because just yesterday it was playing fine, but at that time it was not mounted on the heatsink. The distortion problem has started after I mounted it on the heatsink. I've tried installing a decoupling cap at the RCA jack, in series with the input, but that didn't change anything.

Any suggestions will be greatly appreciated! Thanks!

The temporary setup is a bit messy, I know, and please forgive that.
 

Attachments

  • photo.jpg
    photo.jpg
    147.6 KB · Views: 244
Disabled Account
Joined 2002
My guess is: a short between some pins. The picture is not clear enough to check that though. Please check all the joints around LM4780.

Some other points I can see from the picture:

- The heatsink is unnecessarily large. Although it can never be too big really I would reuse this one for a more powerful amp.
- Wiring is way too long. With these lengths I would advise to use shielded cable to the inputs tot avoid reception of unwanted signals.
- Why are the lead wires sticking out of the PCB (underside) ? Please cut them carefully to a normal length.

I know I am alone with this (on this website) but it is good practice to bend the lead wires then cut them and finally solder them. That way you won't put mechanical stress on the solder joints. It is quite popular to cut after the soldering but that gives less reliable joints in the long run.
 
Last edited:

ra7

Member
Joined 2009
Paid Member
Thanks Jean-Paul. I've checked for shorts on the pcb. I couldn't find any. The wires sticking out not connected to anything were previously connected to some big cans. But, on the other channel, I blew up some diodes, and thinking that this was because of the low esr presented by the big cans I disconnected them and now am using just 10uF in the supply.

The heatsink does get pretty hot leading me to believe that there is some oscillation going on. What is curious though is just yesterday, when the chip was not on the heatsink, everything was working fine. It is baffling to me how it has gone from that to sounding so distorted.

I read up about signals coupling through wires in the LM4780 app note, and i'll probably try that, although its quite a pain to do that in my current setup.

Thanks!
 
From the LM4780 datasheet:

Note 14: The TA27A is a non-isolated package. The package’s metal back, and any heat sink to which it is mounted are connected to the V- potential when using
only thermal compound. If a mica washer is used in addition to thermal compound, θCS (case to sink) is increased, but the heat sink will be electrically isolated from
V-.

Did you insulate the LM4780 from the heatsink?
 

ra7

Member
Joined 2009
Paid Member
Yes, I did. There was a little plastic tape like thing with the kit. Used that. Also checked to see if the heatsink is carrying a voltage - it isn't. It is sitting safely at ground.

Although, the muting resistor is very close to the heatsink. Wonder if that is causing some kind of problem.
 

ra7

Member
Joined 2009
Paid Member
Here's a picture showing the muting resistor. There is a smd like spot provided for it (wonder why a proper through hole isn't provided) and it is on the underside of the pcb.

The solder work isn't the best :)
 

Attachments

  • photo1.jpg
    photo1.jpg
    164.3 KB · Views: 220
Disabled Account
Joined 2002
Please practice soldering skills on an old device. Just desolder all parts of a scrap amp/radio/receiver and resolder them all in. You will learn to work OK then. I think you call it "practice makes perfect" ?!

So cut the excess wire and clean the PCB afterwards with isopropyl alcohol. Flux should be removed, especially between IC pins.

BTW I am still thinking why your diodes blew. They should be able to cope with large cans. Did you check them all again ?
 
Last edited:

ra7

Member
Joined 2009
Paid Member
Yeah, I don't know why they blew either. Their polarities were correct. I think I blew 5 out of eight in one channel. The other channel is fine. There could be one reason though: the toroid that I was using was 25+25V and 400VA. However, on no load it is showing 10V in both channels. The toroid could be the problem as well (although, I would imagine it is pretty hard to screw up a transformer during manufacturing).

I have since replaced it with a 160VA one, 22+22V.
 
Oscillation?

One possibility with mysterious stuff like this is oscillation, somewhere above the audio band...that usually encourages simultaneous conduction of the output devices, which makes the amp really hot, and can kill the power supply.

Do you have a scope to check the output?

Might also be good to post a schematic...perhaps someone could spot a missing component...e.g. part of the zobel network...
 
Disabled Account
Joined 2002
10V offset ? I would advise to recheck all wiring, components values etc. and measure offset again. If it still so high please replace the LM4780 chips. There went something wrong obviously ... I can not think what went wrong except that you soldered the diodes in wrong !? But you didn't as it worked after replacing the diodes and you soldered them in the same way the old ones were. Or not ?

Oscillation/RF pickup could be possible too with the excess wiring working as antennas. I suppose you used the values of parts as given by the manufacturer ?

Without wanting to be a wise guy nor do I want to offend you: if you learn to work tidy straight away you won't have to loose time afterwards with repairing errors etc. Make a choice in which cabinet it will be built, drill the mounting holes and wire it up with some slack so it can be taken apart again. Maybe you want to change thing afterwards so the use of Phoenix connectors might be wise during trial/testing. You can remove those when the amp is finished. Just a thought...
 
Last edited:

ra7

Member
Joined 2009
Paid Member
Jean-Paul,

10V was not the dc offset, it was what the secondaries of the rogue transformer were showing on no load. Also, the channel in which the diodes blew has not been touched. The troubles I'm having are in the other channel, which survived the initial transformer problems.

I agree on the tidiness part, but I'm learning and hopefully, the next one will be better.
 

ra7

Member
Joined 2009
Paid Member
Jean-Paul,

I'm certain it isn't a measurement error. I do get 10V AC on both secondaries and I get the correct voltage with the new transformer.

Back to the distortion, is there anything else I can do, apart from cleaning up the wires and pcb to alleviate the problem? Does installing R0 have this kind of effect?
 

Its very brave to build a paralleled amp with no scope to check for oscillations. I reckon here you're getting the amps fighting each other, perhaps because their gains aren't matched - what tolerance resistors are you using in the feedback networks?

My suggestion would be to get the individual amps working separately before installing the 0.1R sharing resistors to common the outputs. In particular use the AC volts function of a multimeter to check how close the gains of the two amps are for a 1kHz input signal applied to both.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.