Yarg: yet another regulated gainclone

Hi,

while all the threads about the 3886, 4780 and LM338 go in interesting directions, but take their time, I decided to come up with a "sketch" of the good old LM3875T(F) in non-inverting configuration, together with a simple regulated dual power supply.

The intended application is: powering a single speaker in an active 2-way box, yet I'd need two separate circuits with a common transformer but separated rectifier bridges.

Watch this


It features the LT108x/LT108x three-terminal regulators (LDOs), anything between 3A and 7.5A will fit into the board I plan to make a combined footprint for TO220 and TO247 regulator cases.

The LM3875 is implemented in it's standard non-inverted configuration (I'm open to experiment with inverted, btw.). The only difference being the "snubber" instead of the usual zobel network, as I have very short speaker leads but want to be able to experiment with different connections. Btw., the connection between signal and power ground is done on the input side.

--------

I'm very interested in your opinions as this shall become implemented at least four times during this summer. Many of you have different experiences (and knowledge) about the different configurations, please let the comments roll.

As soon as the schematic is discussed and confirmed by - let's say - enough experienced guys for me to feel comfortable, I'll come up with prototypes of the board layout (everything fits on 2" * 3", that's 5cm * 7.5cm).

Cheers,
Sebastian.

Hi Peter,

those 10u capacitors C5 and C10 are in place to improve the regulator's ripple rejection (as explained in the datasheet here). They "buffer" the regulator's reference voltage and stabilize it when an output voltage change appears at the voltage divider (due to current change in the load). C11 and C13 are the caps "behind" the regs.

Thanks for looking at it.

Peter,

Have a look at the circuit on page 8. Page 9 has a paragraph "Ripple Rejection".
This is true for actually any three-pin linear regulator IC. These caps aren't really needed, but they don't hurt anyway (as ripple rejection is what the regulators shall be there for).

It's true that there has to be a cap at the output for stability reasons. Those are provided with the 1mF caps near the LM3875.

BTW, I could imagine substituting the LM3875 with the LM3876 in this schematic, as I've read about the turn on problems over in your thread. Can you tell wether the turn-on noise was related to the input buffers or the supply regulators?

Sebastian,

Okay, maybe the difference between the above schematic (and board) and all the other GCs out there isn't clear enough...

No problem, let me explain:

1. I can live without having the choice between different kinds of rectifier diodes on board. Thus the recifier is kept external, e.g. a metal bridge mounted on the heatsink.

2. I don't feel the need to put many millifarads in front of a voltage regulator, thus, the circuit (and board) suggest only some millifarads

3. I don't want to build a testbed for a multitude of "hyped" components. I want to develop a working circuit. That's why the schematic doesn't feature many different film caps or oversized electrolytics etc.

4. I want the whole circuit on a single board. That's why it's kept to a minimum count of components, yet cared for high sound quality, reliability and usability (one can adapt the regulator to any sensible output voltage).

5. I don't claim it to be HIGH-END in the thread title, nor in the post itself. Because I don't claim to make something better than anyone could.

6. I already made a first attempt on a PCB. It's 5cm by 7.5cm (2" by 3"), double sided, features SMD components (to keep the board layout tight and current paths easily separable) and has a "local" star ground which can be referenced to a "global" star ground and thus the case/heatsinks if more boards shall be combined. I for myself will use it with an active crossover circuit, that's why the above circuit has no buffer.

And finally, I don't plan on doing a group offer, as I don't have the time and resources at the moment. But once the circuit is approved, the board developed and the prototype state behind us, I'll give away the schematics and board layouts for free. At least!

I appreciate your opinions.

Sebastian.

PS: and it should have been "whether" in the above post to Peter, of course.

If it is for stability, shouldn't the output caps be close to the regulator?

Click to expand...

They don't look close in the schematic. But they will be on the board. That's all that counts, isn't it? Do you think I should actually draw them nearby, just to comfort?

Maybe you could ask in the thread this refers to?

Click to expand...

I don't want to hijack it.

I've read Sheldon and Pedja talking about the fact that turn-on noise occurs in conjunction with the buffer (startup). But as I haven't tried it, I would like to hear some comments from someone who built a regulated GC

Sebastian.

Hi Nuuk,

thanks for the reply.
Sounds like this project should go into the direction of number 5.

That means adapting the schematic for the LM3876/86. It's a passive mute delay that needs just a few parts, board space requirements won't increase.

I'll do that asap and post the resulting schematic

Question: What about power-off noise???

Sebastian.

PS: A speaker protection circuit shouldn't be introduced, if possible. After all, it's a GC, the protection circuit would have more parts than the whole amp board itself...

Hi Circlotron,

thanks for the pointer, I'll correct the network with the next schematic revision (the capacitor value is a little off, too).

Sebastian.

Hi Nuuk,

yes, that's right they way you put it.

My thought went more into the direction of implementing a power-up and power-down control with a mutable LM3876/86. The mute function could be utilized to prevent a power-off noise too (using the same principle like the classical protection units), while keeping the overall complexity low.

Sebastian.

Okay, here is a quick update on the current schematic.

I've settled for the LM3876/86. The LM3875 might make turn-on/-off problems with the regulators (it does even without).

The large reservoir capacitors have gone where they belong: on the board with the rectifier diodes. They are replaced with small but sufficient bypass caps near the regulator IC pins.

The local supply capacitors are still indicated to be big, because I currently want to keep the mounting space available (can't let them go for now). Mounting holes for small capacitors will be available.

As I changed the chip, I changed the accompanying circuit, too: It now features an input filter network (optional, for EMI suppression due to long connection leads from the input buffers), a feedback HF bypass for the same reason (optional), and mounting space for ac coupling the input ground connection (optional), as this may be needed by different implementations/uses of the circuit. And finally there is a mute pin connection as to the usual recommendations. Thanks to DD, ESP and SL for inspiration.

Note that there's no Zobel output filter (RL series or RC shunt). I'm currently evaluating the influence of such networks on the given schematic, depending on the use of none, some or all optional components. I'd expect the Zobel network to be recommendable with the minimal circuit and to be "optional" with every component populated.

Please remember, this circuit is not intended as a general purpose amplifier, but as a bi- or multi-amping module system for active speakers. However, with all optional components and an additional Zobel network included, this one should make a pretty good standalone amplifier, though!

Here you are:



I'm thinking of including (smd) trimpots for fine tuning:
1. the regulator output voltage according to low voltage drop and/or power needs,
2. the negative input ground resistor in order to adjust dc offset and noise.

And a couple of questions are still unanswered (or have contradicting answers on the forums):

1. Should the input GND connection really go to the board (that's what R* is for)?
2. Should the speaker (-) connection really return to the board (that's what the GND terminal is for)?
3. Do I really have to avoid ground planes at all? I'm thinking of wide traces for current (return) paths where needed and ground planes everywhere else.

What do You think so far?

I'll go buy some components to start prototyping the circuit today. I'll appreciate any answers. I really need some comfort for my decisions.

Sebastian.

PS: Board space should increase a little, in order to feature bold connection holes for the power supply. The board is currently 75mm (3") by 30mm (1.2") wide.

Thanks for the replies.

c3 and c7 could be 470uF and should be as close to the reg out as posible.

Click to expand...

Yes. As I said, I thought about this, too. The problem is this: a 50V low ESR type of 470uF can be nearly as big as a 50V 1mF standard type. That's why it makes no difference in occupied board space, only in the schematic value. But I changed it, because it always leads to misunderstandings about my intention.

c4 and c8 should be as close to the lm3886 as posible.

Click to expand...

Okay, I introduced two new components C13 and C14, film caps close to the regs. Because of this, I increased the values of C4 and C8 to 1uF (they are available in the same footprint).

a 0,1uF cap as close to the reg input as posible.....+ 10.000uF also.......

Click to expand...

Let me comment this together with the next recommendation.

I believe it is best to keep your main power supply caps close to the regulator circuit.

Click to expand...

Well, I wasn't clear about that. The rectifiers are sitting on a board directly under the amplifier board, distance is lower than 5cm (2"). Because of this, I can move the main supply bypass caps there without losing benefit for the regulator. For still compensating this little distance, C1 and C5 are in place.

Now, C1 and C5 can probably use some 0.1uF bypassing, I'll include that, if it's neccessary. But I don't think so, as the main supply caps will have appropriate bypassing, already.

Also I would recommend using the second protection diodes on the regulators.

Click to expand...

Hmm, I've wondered about it, too. Okay, if You think so, I'll include them.

That's the updated version 0.2a now:



May I point to the questions in my above posts, regarding the current version 0.2a in this post?

Sebastian.

Great!

And how about the trimpots, would that make use for someone besides me? I don't think so, as a better solution would be to experiment with the values and solder in a reliable, stable, simple resistor...