6 channel LME49830 power amp

[cuibono]

Progress update - I changed the resistor to 2.2k, 1.2k and 100 ohms, and used a 1k pot. Got a working voltage of 5.5 to 6.6V - for some reason, the I couldn't get up to the predicted voltage. There was no voltage across the ouput source resistors at 5.5V, but at 6.6V, I measured .08mV. Not much.

So, I changed Rb3 from 2.2k to 3.3k, and got a working voltage of 7.5 to 8.6V. Now, I was able to get some voltage across the source resistors! Actually, almost a perfect range, I think - at the bottom of the pot, I got 5mV (across .1ohm), and at the top, somewhere around 140mV! That's plenty, and I was having some problems - the amp is still oscillating, and past a low level, things got a little to hot. I think the subber resistor let out some more magic smoke, but once again, it didn't feel hot, even less than 10sec after shutting it all off. The amp didn't seem to be passing a signal either.

So I've got my bias - now, why the heck do I have such bad oscillations? Is it picking up noise from something? I'm using a 20pf comp cap, and that part of the circuit is connected correctly...

[SpittinLLama]

In my post above I was talking about my work on the LME49830 reference note (AN-1850). Because of time limitations there is only info on the toshiba FETs but I tested the other two, as mentioned. The work I did with the LM4702 is in AN-1645.

Check to be sure you have a capacitor on the supply line for each FET. The grounding might be causing you some trouble (I haven't looked at your plots). I don't know why you're not getting a signal but when I have some time I will trace out your PCB plots to see if I can find a problem. Try to connect a source to the input or put a pull down resistor so the input isn't floating, then increase the comp cap up. You can also try adjusting the gate resistor values. You're using the LME49810, not the 830? Is shouldn't matter but just so I know for sure. I'll also look over the schematic closer. Might take me a bit but I'll try to get back to you. Normally using these parts is pretty easy, depending on past experience.

Thanks for keeping us posted.
-TH

[cuibono]

Thank you! I am using the LME49830. I'll try what you suggested, and post as soon as I can.

[cuibono]

Well, things are looking good. I added 1k across the input to ground, and increased the comp cap to 30pF, and I'm not getting oscillations, it seems. I haven't tried loading it yet. I held a slowly climbing bias of about 125mA per transistor. The board trace connecting to Bm (the negative bias adjustment and comp cap) is rather long and convoluted - could this be giving me problems?

Now, another smaller problem: I'm including a picture of my scope - I'm getting large narrow spikes on the output, at 120Hz. I'm not sure why I'm getting line switching noise, I followed Nationals recommend power supply, which has more caps than I'm used to. I guess I've got to put everything into an enclosure now, so I can really look at noise and stability issues.

So, I'll be looking to get the board more stable, increasing the gate resistors, adding a chassis. Looks good, but I wonder if I will ever really trust it...

PS - the scope is set to 2mS and 20mV a division

[SpittinLLama]

Sorry for the long delay since I last posted. I am no longer working for National Semiconductor (part of the most recent layoff and either had to relocate back to the headquarters or take the layoff) and have been spending most of my time looking for a new job. I think your layout might have some issues which will limit the performance you can get. First off is that your bias transistor (BD139) needs to be very slow to the output transistors for the best thermal tracking. If too far away then there will be more of a temperature gradient and more time delay making for a bias current that has lots of variation. If you don't mind then don't worry about it. Your large local supply caps are a bit out of the loop on the supply lines. The grounding seems to be using a ground plane, which for these types of designs might not be the best. For example, the local bypass on each FET is pretty critical for stability and yet the GND return paths for these capacitors is relatively long. To get the absolute best THD performance from these chips you have to be careful with GND'ing. I could move wires and watch the harmonics change so it just depends on how far you are trying to get on the low THD levels. The comp capacitor is pretty far away from the LME chip. It is hard to tell exact sizes but I think your power traces to the FETs might be a bit small too, as in narrow. Remember that the board views show in the AN-1850 note are pretty thick but then 2 oz. copper was used for the board instead of the usual 1 oz. This was done to be sure the traces would be able to handle the current plus keep them as low impedance as possible.

If you have the time/interest, I'd suggest looking over the layout in AN-1850 closer and ask how you can improve that design. It depends on what you enjoy/want to do so if just having fun trying to do your own design, then keep at it. If you want a design that is fairly proven then just copy the board design in the note and see about improving it (surely it can be improved more).

-TH

[SpittinLLama]

I traced your board out and it appears that the circuit is all correct except for how analog GND is done. If you look at the board layout in AN-1850 and trace out the GND you will find that all the GNDs associated with the LME chip are routed back to the analog GND connector (labeled J10). The input signal GND, the GNDs from the LME supply bypass caps, the chip GND (pin 3), feedback GND, mute GND, etc. There is no on board connection between the LME chip GND and the output stage power GND. The note probably doesn't explain this very well (sorry about that) but it was done on purpose. An external connector was routed to the off board single point GND where the load GND also came back (also no on the board on purpose). Anyway, your board has the LME portion of GNDs running all back to pin 3 except the signal GND which goes to your analog GND. So how do these two GNDs connect together then connect back to the power supply GND? This might be your problem.

Don't take the power supply note (AN-1849) as gospel. I didn't have the time to really dig in and provide info on all the different caps, optimal value, and how to get there. Instead, as the note says, many caps were added based on common practices and may not even be helpful. The general info is helpful to some and the layout is a good one for routing all the currents well but it is more of a basic guide.

-TH

[cuibono]

SpittinLlama,

Please, don't apologize about the delay- I actually haven't looked at this tread in a couple of weeks now - I've been building the chassis up, getting ready for more testing. But thank you so much for all your suggestions! They are very useful! I'll try to post some pictures in the next couple of days.

And thank you for reading the layout so carefully!!! You did indeed catch an error - I forgot to connect the signal ground to the LME ground. Definitely a problem. The app sheet was clear about the separation of grounds though, I just didn't get it right.

The bias transistor BD139 isn't mounted where its footprint is - that's where I take three wires and mount the transistor right between two output transistors. Sorry if I wasn't clear about that

Comp cap closer. Local supply caps closer. Shorter ground on the FET bypass cap. Wider power traces. Star grounding recommended? All good points. Thanks.

I do enjoy routing, and will probably give it a second go through. Grounding is definitely something that takes a lot of revision. I was going to use the layout from AN 1850, but had problems with it - I'm mounting three amplifier boards on one large heatsink (with fan), and the placement of the power transistors made that difficult - also, since I make my boards at home, two sided boards (designed with vias) add a lot of difficulty to the build. Part of my problem now is that I have a limited area I can use in Eagle. But I'll see what I can do.

Anyway, best of luck with your new job! We never know what the future holds...


PS - I was at BestBuy yesterday, thinking how it would have been easier and cheaper to buy one of those "700" Watt 7.1 amplifers... but I don't have any regrets.

[SpittinLLama]

Well I hope some of my comments help you in your design. It is pretty much impossible to build amps cheaper than buy because of cost the big manufacturers get on items such as cases, transformer, etc. It is just for fun to build.

I have recently accepted a new job so things look good for me. I don't want to say who my new employer is because they have some stuff in audio areas. But if people have questions about National's audio I can give answers but still won't give out confidential information, not that I really know any anyway. I made the Overture and Class D Design Guides, for example. Funny how a little tool for one's self can be made pretty useful.

Keep us posted on your progress.
-TH

[cuibono]

YES! I got it working, at least on a scope. I haven't tried a load yet. I finished the enclosures, and put everything in them. I am going to be gone for the next month, so I'm happy knowing things are good to go when I get back. I still see some very tall sharp spikes on the power supply, so I guess those will need some sorting. The heatsink is barely warm with a highish bias, which is nice.

More testing when I get back, and perhaps build the rest of the modules! Then on to the speakers...

[cuibono]

I hope anyone who uses the eagle files posted above reads post #76 and #77 -- there is an ERROR IN THE SCHEMATIC.