Offset problems, again

[sangram]

Hi

So I'm sold that I do need input caps to avoid the risk of blowing my speakers.

Short story, my 4780 amp is showing about 30mV on one channel and 8 mV on the other channel. What am I missing?

I have a list of possible things to do, which one/s of these are wasted time?

1. Shift speaker ground returns to power supply instead of amp board.

2. The chip is bad, it's not an issue normally so when I fit in the rest of the chips it'll all be OK.

3. Assemble in 3D and use a feedback decoupling cap. The only problem is finding a good cap - the ones I used killed the music completely.

4. Push the input impedance of the amp down to 4.7 K with a resistor. That should bring the offset down to about 20 mV on the bad channel, I think.

Long Story:

I put together a 4780 stereo amp (using Brian/Peter's PCB) using a 220K/10K combo, with a 33K resistor from the + input of the chip to ground (call this Rdc). A 2.2 uF film cap on the input. This thing plays music fine, but I'm getting 80mV on the output of one amp and 30 mV on the other.

I think OK, let's try getting closer to the feedback resistor value so I put a 220K at Rdc, and I get 500mV on the worse channel. The better one is at 200 mV.

Then it's time to change all the resistors. Man, what a pain to resolder stuff into a PTH board. By now the board looks like a butcher's block and the resistor legs do not fit the holes anymore. So I'm holding the resistors in with tweezers and using blobs to hold them... Thank God this is a spare test board that will not be in the main system...

So I try 47K/1.5K feedback and 47K Rdc, and I get about 120 mV/60 mV on the outputs. In desperation, I fit a 100uF feedback decoupling cap the way it is in the National PDF, and the offset is now 1mV on one channel and 2 mV on the other channel.

But now the music has disappeared, it becomes flat, lifeless and dull. This is not acceptable.

In a mad rage I remove all the resistors, and refit the 220/10K feedback and now use a 10K resistor in the position Rdc. my offset is now 30mV on the bad channel and 8 mV on the better one. No cap is used except the input cap.

Edit: Typos

[sagarverma]

hi Sangram

is ofset dc or ac?

check p.s.use three pin plug with earth pin to gnd.connect speaker gnd to p.s gnd,also the i/p gnd.i.e all gnds common and connected in star.

[peranders]

Originally posted by sangram
Short story, my 4780 amp is showing about 30mV on one channel and 8 mV on the other channel. What am I missing?

30 mV is nothing to worry about.

The input offset voltage comes from two components, the voltage offset (1-10 mV) plus the input bias offset currents. If you have AC coupled input, you should match the (DC) resistance from both inputs. This is not so easy if you have DC gain of 30 and let's say 22 kohms at the input.

If you have 10k/220k as feedback and 10k as pull down at the input you MUST also have your amp AC coupled at the input. If not you will get lower DC resistance than 10k and this will create offset. The input resistor should be 220//10 = 9.56 k.

One way to get rid of all problems is to use a DC servo like I have done in my Gainclone designs. Then you can have the amp totally DC connected with less than 1 mV offset in all times.

[sangram]

Thanks Peranders.

Do you mean by 'matching the resistance' connecting an equal resistance to Rf//Ri at the + input to ground? For the datasheet circuits that would mean about 953 ohms (22//1K), which seems a little low as that also determines the input impedance of the amp... Unless I misunderstand you.

What I find is that the value of the offset is directly proportional, almost mathematically, to the value of the input resistor. It has finally stabilised at about 26 mV on the bad channel and just over 6 mV on the better one, with the same 10K input resistor.

30mV seemed high to me as my earlier implementation with Ci cap resulted in <3 mV offset so I was a little alarmed. The problem is that this chip stays a little toasty (on a section of extruded aluminium heatsink, quite beefy) but gets better/cooler as the volume is turned up. I don't think it is oscillating (yet), can't hear the telltale hiss but don't have access to a scope so I can't be sure.

The amp input is/has to be AC coupled, as I modded my CDP and removed the output caps and muting transistors - so there's a permanent ~2mV on the CDP output. with my Ci implementation I could DC couple the amp but since the Peter/Brian board doesn't have space for it I couldn't put it in and I have to keep a capacitor in one position, either at the + input or the Ci...

I was keenly studying servo and other 'stable' offset correction systems, but IMO injecting essentially a part of the DC supply voltage into the input of the amp is really not something I'd like to be interested in for this particular design.

Would you think that unloading the ground line would make sense, if I moved the speaker returns back to the supply grounds... Though, in final implementation this is going to be a 4-chip bridge/parallel setup so it won't matter then as I won't be using supply ground.

One last question, OT but while we're on it. For a bridge parallel setup will I need both chips of one channel to tie to a common power ground reference or can they float with their individual supplies? I am looking at running them off a fully balanced input off a DRV 134 or from a Delta 66 card, so the inputs will be referenced to the cable shield or chassis graound, will I also need to tie their power grounds together?

[peranders]

Originally posted by sangram
Thanks Peranders.

Do you mean by 'matching the resistance' connecting an equal resistance to Rf//Ri at the + input to ground? For the datasheet circuits that would mean about 953 ohms (22//1K), which seems a little low as that also determines the input impedance of the amp... Unless I misunderstand you.

That is correct! If you have DC gain set to 1, 22 kohms is the right value. You see those input bias currents are causing you some trouble.

[sangram]

Got it, thanks.

We'll see the heat situation and then take a call.

Any advice on my Bridge/parallel thing?

One last question, OT but while we're on it. For a bridge parallel setup will I need both chips of one channel to tie to a common power ground reference or can they float with their individual supplies? I am looking at running them off a fully balanced input off a DRV 134 or from a Delta 66 card, so the inputs will be referenced to the cable shield or chassis graound, will I also need to tie their power grounds together?

[peranders]

I'd say that you should study the AN-1192.

[sangram]

Thanks, I already went through that but it doesn't answer my question. The AN-1192 primarily talks about a high-power amp running off a single-ended input by running one pair as inverting and one pair as non-inverting, and adding each pair in parallel.

My question was on balanced inputs and power grounding, whether it was OK to leave the power floating on both the amps (assuming independent transformers/rectifiers) or would I need to reference them to a common ground at some point?

I know that the signal inputs need to be referenced to something, and my second question was, whether to refernce them to the signal ground (shield wire) or to the other side of the balanced connection. Obviously in the second case we cannot tie the grounds of the amps together at any point or we will short the inputs to common ground...

This is complex...

[peranders]

If you make two identical amps I think it's possiible to drive them from a floating source but if the source not is floating it's possible that it will work anyway but the truth is that I have no experience of this and possible hum problems with ground loops etc. Sorry can't help you here.

[sangram]

Thanks P-A. No worries, I've already floated questions over to Peter and some other members who've built similar setups, hopefully I'll get a reply.

FYI I have two sources, a computer soundcard which can run in fully floated mode, and a CDP which I hope to be able to convert to fully floated by using a DRV134.

Thanks again. I am assured by Peter that my current offset voltages are nothing to worry about, except that it does get very warm (almost worryingly so) at idle or very low power levels.