New amp: The Mjolnir
Jan 6, 2010 at 11:10 PM Thread Starter Post #1 of 18

Spasticteapot

100+ Head-Fier
Joined
Oct 16, 2006
Posts
381
Likes
11
After a marathon session in EaglePCB, I finally finished a PCB design for an amplifier, which I've chosen to call the Mjolnir. This is my first time designing a PCB both this small and this complicated, and I'd appreciate any thoughts you might have on the PCB or design.



What's in it:
It's an implementation of the headphone amplifier from the LME49600 (National's equivalent to the BUF634) datasheet, though using a single quad instead of four dual op-amps and a much smaller PCB. The amplifier contains an op-amp gain stage with a buffer in the feedback loop for more current, and a DC servo so you don't need to use an input capacitor (which suck.)

http://www.national.com/an/AN/AN-1768.pdf

What parts does it use:
I based it around the LME49600 buffer and LME49740 op-amp. Both of these are excellent components and not too expensive. However, just about any quad op-amp should work.

The resistors are a fairly standard 2.8 x 6.4mm size - I'm using 1% metal film resistors. I chose to use multilayer ceramic capacitors in the power supply due to extremely low ESR, but there's enough room for WIMA polypropylene capacitors to be used instead.

All the parts used are through hole for easy soldering, with one exception - the zener-based static discharge protection diodes at the input and output. Fitting eight more diodes on the board simply wasn't practical, and the IC is of a fairly large and easy-to-solder size - and it's optional anyway.

How well does it work:
As you can see from the datasheet, distortion is very, very low. Despite the amplifier's tiny size, it's good for up to 250mA at +/-15V at the outputs - enough to drive just about anything.

Why is the PCB so ludicrously small:
Seeed Studio does 2" x 2" PCBs at the low price of 5 for $16 if you give them all the documentation for your project.

Why the name:
I could lie and tell you it's due to the impressive power output - not many headphone amps can do more than 3WPC. However, I'd be lying. It actually refers to my design philosophy - "All problems can be solved with a hammer - so long as you use a sufficiently big one." If you'll look at the PCB, there's a very literal interpretation of a star grounding scheme; after spending rather a lot of time mucking about with ground planes, I found it was almost impossible to use big, through-hole components without mincing the ground plane into tiny little pieces.
 
Jan 6, 2010 at 11:59 PM Post #2 of 18
Not a bad start IMO
smily_headphones1.gif
A few obvious ones:

1) The resistors should use the 0207/10 footprint instead of the 0207/7, otherwise they are difficult to mount. This shouldn't be a problem for the layout as far as I can see.

2) I would widen the thin top layer traces.

3) The LME49600 only comes in a surface mounted package (TO263) so the BUF634T footprint wouldn't work (unless I am missing something)

/U.
 
Jan 7, 2010 at 12:55 AM Post #3 of 18
the starground looks a bit odd though.

Are you sure you've tapped the correct point?

You should tap a line out from the mid point of two big rail capacitors and use the end of the line as the star point.
 
Jan 7, 2010 at 1:00 AM Post #4 of 18
Quote:

Originally Posted by Nisbeth /img/forum/go_quote.gif
Not a bad start IMO
smily_headphones1.gif
A few obvious ones:

1) The resistors should use the 0207/10 footprint instead of the 0207/7, otherwise they are difficult to mount. This shouldn't be a problem for the layout as far as I can see.

2) I would widen the thin top layer traces.

3) The LME49600 only comes in a surface mounted package (TO263) so the BUF634T footprint wouldn't work (unless I am missing something)

/U.



1. A good idea, but I'm not sure everything will fit if I do. I'll give it a try.

2. Once again, I'll give it a try, but I have a feeling everything will short.

3. Straighten the pins, and it should fit. Alternately, you can just solder the pins to the pads.
 
Jan 7, 2010 at 5:31 AM Post #6 of 18
Quote:

Originally Posted by dsavitsk /img/forum/go_quote.gif
Not to ask a dumb question, but are all the pads on the buf634 shorted together?


Yes... yes they are...
 
Jan 7, 2010 at 7:05 AM Post #8 of 18
Besides what everyone else said, it looks good to me. It would be even better if you could fit a couple 3.5mm jacks for input/output (the vertical-style ones would be great for saving space) and possibly even a pot. The two jacks would fit nicely in the bottom left and the pot could go on the bottom right, just to have an all-in-one implementation that wouldn't require wiring anything. Anyone who didn't want to use 3.5mm jacks or that specific pot could still wire them up using the holes of those components.
 
Jan 7, 2010 at 10:58 PM Post #9 of 18
Quote:

Originally Posted by dsavitsk /img/forum/go_quote.gif
Not to ask a dumb question, but are all the pads on the buf634 shorted together?


That's an artifact caused by a funky DRC setting. I've got it fixed now.

Incidentally, after posting my design on the Parts Express "Tech Talk" board, a professional PCB designer came up with a better version in about twenty minutes. The above took me about five hours - and it's my third try at designing a PCB for this schematic.

Quote:

Originally Posted by Juaquin /img/forum/go_quote.gif
Besides what everyone else said, it looks good to me. It would be even better if you could fit a couple 3.5mm jacks for input/output (the vertical-style ones would be great for saving space) and possibly even a pot. The two jacks would fit nicely in the bottom left and the pot could go on the bottom right, just to have an all-in-one implementation that wouldn't require wiring anything. Anyone who didn't want to use 3.5mm jacks or that specific pot could still wire them up using the holes of those components.


I think there's just barely enough room to squeeze one of those on. That said, this would make a poor portable amp - those buffers hoover up a lot of juice.
 
Jan 7, 2010 at 11:03 PM Post #10 of 18
Quote:

Originally Posted by Spasticteapot /img/forum/go_quote.gif
That's an artifact caused by a funky DRC setting. I've got it fixed now.

Incidentally, after posting my design on the Parts Express "Tech Talk" board, a professional PCB designer came up with a better version in about twenty minutes.



Could you post the new version?
tongue_smile.gif
 
Jan 8, 2010 at 3:54 AM Post #11 of 18
Quote:

Originally Posted by Spasticteapot /img/forum/go_quote.gif
I think there's just barely enough room to squeeze one of those on.


Ah, the scale of the schematic deceived me. Look forward to seeing some board get made.
 
Jan 8, 2010 at 5:27 PM Post #12 of 18
Quote:

Originally Posted by Juaquin /img/forum/go_quote.gif
Ah, the scale of the schematic deceived me. Look forward to seeing some board get made.


Will post new schematic this afternoon.

Despite the rather impressive power output (over 3 watts per channel!), it's a really tiny amplifier - the PCB is only 2" x 2". Parts cost is relatively low - you can stuff a board for about $35, and build the whole thing for about $50 if you're thrifty. Or you could buy one from me - I'm going to have some extras.
atsmile.gif
 
Oct 14, 2012 at 12:11 AM Post #14 of 18
Too bad you didn't get a patent on your amp name, Schiit has a amp called Mjolnir too. :blink:
 
Oct 14, 2012 at 1:04 AM Post #15 of 18
do you think you could have used a more accessible buffer? IMO there is no advantage to the buf634T over the LME and the LME is both more widely available and cheaper now that Ti acquired National. the current handling of the LME49600/10 is vastly reduced without a heatsink and it has no way to easily screw to a heatsink if you can manage to mount it vertically. Basically, just use the PCB for heatsink as intended, I really dont understand why people persist in trying to make this excellent and troublefree SMD part into a throughhole part, its dead easy to solder. you could also save more space by using SMD thin film parts, or MELF if you want something better.

why the servo? there are no parts on the board that will drift or need matching, so a proper design does not need a servo, or a cap.

it would be great if you had the soic16 pattern in parallel for optional soic parts, like opa1644

also do you think you could host your pics somewhere else? when I clicked the pic it loaded imageshack without asking, which spawned no less than 3 popup spam advertisements, one of which was fullscreen and it played a bit of havoc with my i5 for about 5seconds after I closed it all, as if it was trying to do something else.
 

Users who are viewing this thread

Back
Top