LM4780 - problems with high idle current and no oscillations?

[hellthom]

Hi all!

I'm currently in the process of building a parallell design LM4780 gainclone. It's a kit from Peter Daniel at audiosector. Im having some hard-to-solve issues regarding high idle current draw from one of the channels.

So to the story:

I initially had some slightly audible white noise after completing the boards. This was cured by adding a snubber network in parallell with the feedback resistors, as described in the datasheet of the LM4780. The chips now sounded good, and also had close to no DC offset. But, while one channel was working perfectly, the other was drawing 2 amps of current per rail!

A 220pF capacitor was added between each input pin and input ground pin of the two sections of the chip, as per the datasheet.

Also, a Zobel network was added to the outputs, as described in the audiosector circuit diagrams.

The powersupply has a total of 30'000yF per rail, with decoupling caps at each of the larger caps, pluss decoupling caps soldered directly at the amp boards smaller caps. It's also "snubberized".

I have also implemented star grounding in my design, and as for now, power supply ground and signal ground is separated. All power wires are twisted in pairs to reduce potential noise.

I have also tried to lift the legs of the output resistors to figure out if it's a problem with mismatched resistors, so that the two sections are fighting each other. That didn't help any.

I am now unable to find any reason as to why one channel would work perfectly, while the other is still acting up, drawing about 2 amps from each rail. Does anyone have more suggestions for me to try? Could there be oscillations that i have not filtered out by now? Have anyone experienced any of these chips to be "dead on arrival"? Could the chip be toast?

Any help would be greatly appreciated!

[dannyjmf]

since u mentioned one channel is working perfectly.... and another one is not .... why not try isolate the two .... i mean temporary wire them normally instead parrallel n try do voltage measurement on both channel and compare readings .... i hope this might help u ....

good luck ...

[abraxalito]

I'm curious as to why you didn't start out with the snubbers fitted since they're shown on the provided schematic? Check the tolerance of R1,R2,R3 & R5. Also take out the 220pF cap as that's not required and might actually be a problem in a paralleled application. Also take off the 'snubbing' networks in parallel with the feedback resistors, they also might be a cause of difficulty.

What actually happened when you lifted the leg of the sharing resistors?

[AndrewT]

Have you posted on the Peter Daniel (PD) area?
Have you built it in compliance with the PD build manual?

[hellthom]

abraxality, im not sure of wich snubber network you are refering to, but i did start out with snubbers on the psu, but not on the feedback resistors (it was not a part of the design of the boards from audiosector).

I have also checked the resistors, and they all seem to have correct values (Also, Peter sorts parts so that they're mostly accurate).

Since the snubber over the feedback removed noise, and not the 220pF one, i will try to remove that one as you suggest. Thanks for the tips

When i lifted the leg of the sharing resistors (which as far as i know, would cause the same effect as running it in stereo as dannyjmf suggests) nothing happened! Absolutely nothing! It still got as hot as it used to, alltough i did not try to measure current, since the heatup was as it used to. That also rules out differences in the sharing resistors as the culprit, right?

AndrewT, i have posted on the audiosector area, and that didn't really help any. Basically, all the help that i got was "I have never heard of anything like this". I also have followed the guidelines in the manual. I built a LM3886 based amp that works perfectly a few years ago, so im quite sure that i know what I'm doing.

What appears wierd to me is that there is no audible distortions, and no visible or measurable differences between the working and the not-working channel.

[abraxalito]

Originally Posted by hellthom

abraxality, im not sure of wich snubber network you are refering to, but i did start out with snubbers on the psu, but not on the feedback resistors (it was not a part of the design of the boards from audiosector).

Snubbers on PSU can't be involved in this problem, it was those on the amps' feedback resistors I was concerned about.

I have also checked the resistors, and they all seem to have correct values (Also, Peter sorts parts so that they're mostly accurate).

You'd typically need a 4 digit DMM to verify that these parts match within 0.1%.

Since the snubber over the feedback removed noise, and not the 220pF one, i will try to remove that one as you suggest.

A snubber network across the feedback resistor I'd say was a bad thing - it increases the loop gain at higher frequencies, exactly where you don't want more loop gain because it tends towards instability.

When i lifted the leg of the sharing resistors (which as far as i know, would cause the same effect as running it in stereo as dannyjmf suggests) nothing happened! Absolutely nothing! It still got as hot as it used to, alltough i did not try to measure current, since the heatup was as it used to. That also rules out differences in the sharing resistors as the culprit, right?

So let's see if I've understood this right. You took one or more of the sharing Rs out of circuit and the rogue amp module stayed much hotter than the good one. Right? Differences in the values of the sharing resistors are immaterial here - even 10% types would be close enough. Though if it were me, I'd be using 0R22s rather than 0R1s for sharing unless I had very precise gain resistor values.

Have you checked the voltage drops across those sharing resistors?

[hellthom]

I have a 4 digit DMM, so i guess I'm ok at that one.

A snubber network across the feedback resistor I'd say was a bad thing - it increases the loop gain at higher frequencies, exactly where you don't want more loop gain because it tends towards instability.

Initially i thought this would filter high frequency, but obviously, after a little bit of thinking, you're absolutely right. Weird how this did not worsen the problem, but instead removed hiss. I guess adding a cap before the feedback-resistor to signal ground would be more of a low-pass-filter?

So let's see if I've understood this right. You took one or more of the sharing Rs out of circuit and the rogue amp module stayed much hotter than the good one. Right?

You're spot on. I just don't see why the chip then would run hot, with no output connected on any of the two outputs, no sharing resistors, and inputs shorted to ground.

Have you checked the voltage drops across those sharing resistors?

Measuring voltage drops over the sharing resistors is rather difficult, since I can only measure DC or 60/50Hz AC with my DMM. the DC offset though, is way lower than maximum. I think i measured 18mV and 23mV

So to sum things up, until now i've tried:

• No snubber over feedback, or cap at input.
• Same as above, but "unparalelled" (No effect)
• No snubber over feedback, cap at input (No effect)
• Both feedback snubber and input cap. (Removed hiss)

I'm really unsure about wich step to take next. I don't wanna order a new kit until I've tried all possible options. Ordering a new kit is cheaper than buying two chips from my supplier, and I can allways use the other one for a smaller amp of some sort

[gootee]

The reason that dannyjmf astutely suggested to unparallel them was so you would have two identical circuits that were performing differently and you would then be able to measure at each corresponding point, i.e. in both circuits, looking for a difference, which should lead you to the problem component or interconnect.

You could measure:

a) Resistance from each circuit node to ground (with power off). Do for every point in both circuits and compare the two for each point.

b) DC voltage from each point to ground, with power on but inputs shorted. Do for every point in both circuits and compare for each point.

c) (Optional: ) Same as (b) but use a 50 OR 60 Hz input signal and measure AC voltage between each point and ground. Do for every point in both circuits and compare for each point.

[hellthom]

Originally Posted by gootee

The reason that dannyjmf astutely suggested to unparallel them was so you would have two identical circuits that were performing differently and you would then be able to measure at each corresponding point, i.e. in both circuits, looking for a difference, which should lead you to the problem component or interconnect.

You could measure:

a) Resistance from each circuit node to ground (with power off). Do for every point in both circuits and compare the two for each point.

b) DC voltage from each point to ground, with power on but inputs shorted. Do for every point in both circuits and compare for each point.

c) (Optional: ) Same as (b) but use a 50 OR 60 Hz input signal and measure AC voltage between each point and ground. Do for every point in both circuits and compare for each point.

Ah! How come i didnt think of that! Thanks a lot gootee. I will get right down to measuring, as sonn as I'm done with that $€"#¤ homework

[hellthom]

Finally got my a** around to try to solve this thing.

After doing all the aforementioned solutions i ended up buying a new kit, and assemble one new channel with the same modifications as to the two others. This worked perfectly. Looks like the chip itself was damaged.

Might be worth knowing for other people out there, since I find it really annoying myself to read through threads that end in nothing

Just some tuning to adjust the gain a little higher, and some noise reduction, and this project is finally done!

Thanks for the help to those who contributed !