Comparing LME49810, 49830 and 49811

[Sebastiaan]

Originally Posted by kct

Sebastiaan,

you have been recently at a EMI Signal Integrity class haven't you ?

I was, and boy what an eye opener it was. I am looking at a circuit board with a different set of eye' s now. Thanks for highlighting this very important yet often overlooked or better said neglected approach.

Keep the loop area as small as possible and things magically fall into the right place


Cheers

Dear KCT,

Indeed an eye opener. And yes the way you look at circuit boards, traces and inductance change forever! Also it becomes obvious how many commercial designs are just plane wrong, with 100nF caps on the weirdest places with high inductance "ground" traces all over the board.

And you are right also, that like you, I started to look at circuit boards as river and a drain. Once you can visualize where the water flows, what is in the path, and where the high and low currents are located things fall into place. A given with higher frequencies RF returns is that they stay "glued" to the the source traces. And this point also makes you realize that many commercial ground planes are also wrong designed.

For general readers,

Keep dirty current and/or high current circuits as close to the return point as possible, and place sensitive and low current circuits around it. The dirty and high currents will find their way to the drain without affecting the surrounded circuits.

People (and me as well in the past) who get hum, RF or other noise with ground-planses didn't maintained a good component placing and routing discipline. Once again think about your shower sink and imagine how water flows to the drain

If you design as above thump of the rule, you can get away with an unbroken (and it must be unbroken!) one layer ground plane, with high and low current mixed signal, and get an extraordinary good result. A result in my opinion you can't beat with star grounding. The Focal SM9 commercial active loudspeaker is the living proof of it. All boards have unbroken ground-planes.

Everyone who want to get closer to ground-plane designs, should read everything Herry Ott has ever written. Best stuff ever.

Once you start to understand PCB design one can focus less on expensive exotic "audiophile" components. A good PCB design with cheap TIP35 transistors will outperform a bad design with state of the art components.

It turns out for me that a good PCB design gave the most significant improvement in sound. Assumed the design and topology are top-class of course

As addition on above (still reasonable on topic I think). Like others indicated as well. Don't burry low current circuits in a loop of high current traces. What you often see is the VCC and VEE traces all on the outsides of the boards, and the VAS and low current circuits in the middle. Don't ever do that! In class A/B amplifiers the inductance between the VEE and VCC rise with current draw and many odd harmonic spikes occur. Those AC magnetic radiations in between the outer VEE and VCC traces go all over the board and affect everything on it's way in between. If you want low THD. this can't be good... Keep VEE and VCC traces close together as a pair or stacked in multilayer boards. Let them enter the amplifiers from one side, and keep the low current circuits at the other side.

In all matters keep differential lines always together as close pair or stacked layers, and keep the distance as small as possible between the opponent traces. How often we see op-amps based circuits where the VCC and VEE lines are distributed independently. Keep them always close together as pair.

Keep the speaker output traces close to the power amplifiers VEE and VCC traces. Again the Focal SM9 speaker has this all incorporated. VEE, VCC en Output are all stacked in a 4 layer multilayer PCB with layer 2 as unbroken ground-plane. The results are stunning.

Last but not least, If you use an unbroken ground-plane, you can make better advantage of local de-coupling with 100nF caps. Place them direct to the VEE or VCC pins of each transistor. The inductance to ground return is way less then any star-ground topology. Don't put any 100nF caps in the PSU it-selves, only direct at the pins of the last device.

Please don't take my post as a "smart ***" response. I still learn every day and that is the fun part of life and designing. My contribution is meant to share my enthusiasm for the discoveries I did, and I hope to trigger those who didn't saw the light yet, or who are afraid for ground-planes to give it a second thought and learn a bit about it. I really think that PCB design to often is underestimated, and that is kind off odd since we all looking for the highest performance here.

No doubt people will comment my approach is over the top, but the scoreboard (measurements) doesn't lie. I swear by 4 layer board designs for amplifiers. Where layer 1 is all the low current and low signal level traces. Layer 2 an unbroken all cupper ground plane. Layer3 and layer 4 contains the high current traces. Preferably VCC on payer 3 and VEE on layer 4 stacked as semi power-planes.

The layer 2 ground-plane gives a good isolation between low and high current traces since it acts as shield in between. Despite the fact that the overall capacitance will increase, my amplifier designs are more stable with higher bandwidth en less compensation due the excellent inductive and RF qualities of the board.

With kind regards,
Bas

[AndrewT]

Yes.
each "input" and each "output" are circuits. Always run them as Flow and Return pairs. Keep the loop area as small as possible.

Think in "currents" rather than of signals or voltages.
Don't let "currents" share the same length of circuit trace or wire. Different circuits can meet at a point, when necessary.

Do you know if they plan to post an internet video (you tube) of that lecture?

[Sebastiaan]

Originally Posted by AndrewT

Yes.
each "input" and each "output" are circuits. Always run them as Flow and Return pairs. Keep the loop area as small as possible.

Think in "currents" rather than of signals or voltages.
Don't let "currents" share the same length of circuit trace or wire. Different circuits can meet at a point, when necessary.

Do you know if they plan to post an internet video (you tube) of that lecture?

Dear Andrew,

I still have the power-point sheets in hard paper from one of the AES evening given by Bruno Putzeys. I will ask him permission if I may scan it and post it on DIY audio.

With kind regards,
Bas

[polsol]

A little OT but as this thread covers the LME 49XXX chips, has anyone tried running these chips with local (chip only) rather than global (including the output tansistors) feedback?
Also, in lieu of the chip topology info, do these chips operate in Class A mode?

[RocketScientist]

Search on JCX's posts here. He's done a few "multi loop" designs using localized feedback. He's also someone who seems to actually understand all the complex math behind making such designs stable which is often the biggest issue. I believe he does really high-end op amp work (sensors?) for his day job.

[Arkudos]

Veteran,

I have 2 of these boards but never got the PS boards. I would be interested in getthing those from you if you have any left and the STD05N/P transistors if you still have a good source for them.

Thanks again,

Ark