Cmoy layout.

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Here is my first ever layout of anything. If you think it looks OK, then you can thank Tangent for that. Because on my own, it was a disaster.

I have submitted the gerbers to batchPCB and passed the DRC checks. I am looking for some final critique before I pull the trigger and order some of these. Anyone see any problems that might cause stability issues or oscillation in the opamp? V+ and V- will be decoupled with .1uf ceramic on the bottom of the board.

This pic is from ExpressPCB because I used that first to layout my design and it was later moved to Diptrace to generate the gerbers.
It is easier to show in 1 pic what is going on.
 

Attachments

  • whole.png
    whole.png
    14 KB · Views: 1,220
Hi,

Depending on the op amp you have used you may run into stability problems as;

You have crossed the input and the ouput tracks over, so signals from the output have the potential to cause positive feedback to the input. This will only be a problem if you have chosen a very high speed opamp or one that is not well internally compensated. It looks like you could fix this by turning the opamp round and rerouting.

Since you have a ground plane why not run it under the opamp it should help to provide a good return path for the signals.

If you are concerned about the possibility of instability you could put the option for a capacitor across the feedback resistor. You can solve may instability issues this way.

Just out of interest what is the 3 pin component connected to both power rails and ground?

I could give you better feedback if you posted the circuit diagram with your component choices.

However if you use a internally compensated opamp such as an OPA2134 it has a good chance of working OK, even though the layout is not ideal.

Regards,
Andrew

Regards
 
I'm guessing the TO-92 package across the rails is a TLE2426 "Rail Splitter".

As to the layout - well, it's pretty hard to make OPA2132 unstable. Keeping the ground plane away from the opamps input pins is a good idea. You might consider adding a split to your ground plane, as it is it forms a big loop around the edge of the board, in some cases such a design can cause issues.
 
I was warned to keep the the ground plane away from the opamp by Tangent and that is the reason there is no ground plane under the chip.

Since the output crosses the input perpendicular, I had hoped to minimize any chance of interaction there. If you guys think this will cause a problem then I will try to re-arrange some things.

I have ordered opa2132 and opa2227 for this and they will be here on Friday.

@stream- The schematic is directly from tangentsoft. Since this will be the first time I have ever made this amp I am going with the default for now. I intend to play around some after I get one built.

The TO-92 part is indeed a rail splitter. To me it seamed like the more elegant solution.
 
Hi,

Just did some quick calculations based on the parts data.

OPA2132 max output current is 40mA, using V=IR => max output voltage into 32R headphones is approx 40mA x 32 Ohms = 1.28V peak. This is not great but I am guessing the opamp provides more than this on instantanious spikes so probably sounds ok.

However if you use a TLE2426 rail spliter its max current is listed as 26mA and you are sharing it between two channels. So if they both peak at the same time which is quite common in bass transients(equivalent to halving the resistance) the max voltage is 26mA x 16 ohms 0.416 volts. So you may find that you start getting clipping distortion alot earlier than you would like. This is because the ground splitter saturates and hence starts to move its voltage. Because there are capacitors on the rails some of the charge from quick spikes may be drawn off these so your actual performance may be somewhat better than this.

I have never has any problems with gound plane near the pins of medium speed opamps, even surface mount ones which run much close to the plane. I have only really experienced problems with parasitic input capacitance when the input circuit is capacitive and there is no resistive element to isolate it (such as using CMOS analogue switches on the input to an opamp). But if it is the common wisdon for this circuit; I don't see any problem with not having a plane under the chip, you may just pick up slightly more noise.

The input crossover is minimised as you say by the paths crossing perpendicular to each other. So you will almost certainly have no problem in quiescent operation (i.e not playing). My only concern is as you will be drawing high currents into the heaphones you may cause load induced instability which would be hard to identify without measurement equipment and would just make it sound a bit off. However since you are using unity gain stable opamps you will probably get away with it.

I would also look at the return path from the speaker return (ground) back to the ground splitter at the moment this path is poorly defined as it will flow accros the plane. Again you will probably get away with it, but if it was a "real" power amplifer you would normally use a seperate return track to a star point at your ground referance. In this case that is your ground splitter.

As I said in my previous post the circuit as it is will almost certainly work. It could however be made better with a few modifications the most significant of which I suspect is changing the rail splitter to something with more current capability. However if you are running 600R headphones most of these problems will be much less significant.

Hope this is of some help.
Regards,
Andrew
 
I've built a couple of amps this way, and the current limits of the TLE2426 don't really seem to cause a problem. If you use bulk capacitance before the splitter, it does, but with 2 caps spanning rail to ground, it works fine in every situation I've seen so far...and, of course, even better again if you use both. Buffering it with something like a BUF634 is preferable, but if you're going to do that for the ground, you might as well do it for the channels too, and before you know it, you've got a PIMETA. Which, funnily enough, I'm using now, since I got sick of just edging incrementally towards it.

If you can handle making the board bigger, consider adding an OPA227 as a buffer on the ground, should help slightly with the current handling capacity - I've got one configured that way, with OPA2107 on the L/R channels and OPA227 in the ground buffer, and it's very good for semi-portable duties running from 2 9V batteries in series (or a single 9V if I swap the OPA2107 back to OPA2132).
 
TLE2426 data sheet states max current is 80ma. I will be useing less than half of that. It is generally accepted that the part is good for 40ma. Not sure where you got 26ma.
Going with the formula P=R*I^2 max power into 32ohm phones is 51mw. That is more than enough for 32ohm and well below the max for the chip.

The data sheet for opa2132 states it will swing near 2v at 600ohms.
I also plan to try other opamps as I have few different types arriving Friday.

I will probably not stay with a gain of 11. I plan to set it where I get the most travel of the volume pot.

Thanks for the advice about the ground Andrew. We will see where it goes. This board is 1.4"x1.325". There really is a lot less ground plane there than it appears. Since none of these issues are really show stoppers, I think I will order a few and save any changes for rev2. I will have built this rev and can be more certain about what changes need to be made. I think Rev2 will be buffered and have a ground channel. Then things should get really interesting.:D
 
Ahh what is with this moderation thing? I have been here for a year already guys. Give a guy a break.

Sorry for the double post. But I just remembered that there is the option for R5 to be jumped or for a resister from 10-100ohm. That should make it easy enough to get a little more voltage.

If I use a 50ohm resistor it brings the voltage to 3.24 volt at the cost of some minor attenuation. If I get some new phones later R5 can be removed and the gain adjusted accordingly.
 
Hi,

I have to admit I only skimed the data sheet, this was the point I saw that lead me to belive you would get current limiting at 26mA.

Short-circuit current
Sinking current, VO = 5 V
25°C
26
mA

If you are running ar 12V then the situation is only slightly better.

Short-circuit current
Sinking current, VO = 12 V
25°C
31
mA

this is in the typical operating characteristics.

However having taken a more carefull look you are right the Absolute Max current is indeed listed as 80mA. However operating parts at absolute max they may be operating outside there specified range. i.e the load regulation is undefined all you can say is that it won't be destroyed. This means the output 0v may start to move. Also operating near max power reduces there operating life (and with two channels of 40mA each you could be peaking at 80mA).

If you put a large caps on the output then the caps would be able to supply the transients. But I didn't see any on your design and you only mentioned putting small caps on the back.

In my experience average power has little to do with transient performance, with low crest factor music I am sure you will get plenty of noise. However if you want to capture the dynamics of more complex music you don't want the headphones to start to clip that early. The current limit will cause a transient to be cut off, its not about the average power which is a different issue. I used to use a 5V headphone amp in my designs this would happy drive 2.2V peak into 32 Ohm headphones, but still I got complaints from customers about dynamic perfomance and even volume from some people. I ended up with BUF634 OPA2134 compound design to deliver the current.

Its clear from other peoples comments that this design works. I guess I am just a pedant and don't like to see things that arn't designed to work inside there operating parameters, not right on the edge. Bottom line, if one of my engineers came to me with this design I wouldn't let it out of the building, but that is for comercial use and I guess as a DIY project its just fine.

Regards,
Andrew
 
Ok Andrew, you have compelled me.:D
I don't think it is possible to have too much capacitance. However, it is certainly possible to have too little.

One question though. I don't think that the polarity of the highlighted cap is right. Should it be the other way around?
 

Attachments

  • cmoyrev1.1.png
    cmoyrev1.1.png
    13.7 KB · Views: 705
Your not alone, I grind my engineering team down just as much:) Seriously for the cost of a couple of extra caps I think you will find the improvement well worth it.

You are right the -ve terminal of the cap should be connected to the -ve power supply and the positive should be connected to ground.

It will be very unhappy if you leave it like it is, it may even let the magic smoke out :)

Kuldeepsingh - with this design the limiting factor is the rail splitter so it would be the first place to improve performance. Putting the outputs of opamps in parallel requires that they are isolated with some resistance. But it could be done, a better approach would probably be a high current output opamp or a buffer. A bridged design won't benefit you in this case as the design is current limited not voltage limited and a bridged design will increase the volatage available but not the current.

Regards,
Andrew
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.