| |||||||||||||||||||||||||||
| Moderators: cmoy | Welcome. Please log in. | ||||||||||||||||||||||||||
Registration is required to post a new topic or a reply. 
| |||||||||||||||||||||||||||
tangent  Headphone Council Joined: Aug. 27, 2001 |  Message [#1] posted on: 09-13-2005 11:27 PM CST (US).   |  |  I've been working on a successor to my MINT amp lately. (Schematic, 161 KB)
The major changes are: - No more "folding" the PCB: board is a little over half the size of MINT v1.0, and completely fills out a standard mint tin top to bottom and front to back. (Notice the corner cutouts. It's tight!) - Using high-current op-amp instead of BUF634s. Can you say AD8397? I thought you could... But I'll be testing other chips, including the OPA551, LM6172, AD817, etc. - PPA-like ground scheme: uses one channel of "ground" op-amp plus a divider as replacement for TLE2426, and another fed from the first to drive output ground. - Full ground plane, yet only one more via than in v1.0. - Trickle charge circuit, for charging 9V NiMHs without removing them from the amp. - Only one rail cap, so virtual ground driver isn't swamped by capacitive load. This is partially counteracted, unfortunately, by the next feature: - Added bypass caps. - Rail cap is in front of the power switch, so it stays charged up, to reduce turn-on and -off thumps. - Input caps are much smaller, so you'll have to use polyester caps if anything. (Sorry, something had to give...) - Most resistors surface-mount. (1206 size) - Using ALPS RK097 switched pot instead of Panasonic EVJ. It's not "finished" yet. The grounding scheme is somewhat up in the air. It'd be less current draw to replace one ground channel op-amp and the divider with a TLE2426, but space is tight. I'll try that, and if I can't make that fit, I'll have to consider going with a single-channel op-amp for the ground, and making it drive all grounds. Also, I should probably make RLED an SMT part as well. No particular reason it is the way it is... Comments, questions, complaints? |
| pokeyman Member Joined: Jun. 26, 2005 | Message [#2] posted on: 09-14-2005 12:48 AM CST (US). | | Very nice -- I've been sitting here for hours myself (when I should be sleeping) working on my own layout -- not nearly as sophisticated but a lot smaller... FWIW, one thing I read in the new AD8656 datasheet, in case you want to consider it: "Because load currents flow from the supplies, the ground for the load impedance should be at the same physical location as the bypass capacitor grounds." Cheers! |
| skyskraper HeadWizer Joined: Aug. 7, 2005 | Message [#3] posted on: 09-14-2005 01:10 AM CST (US). | | awesome work tangent, that looks like a sensational layout, and the convenience of that pot too  1206 i think is a very workable size for smd 'phobes (including myself LOL ) finally i can fit a portable amp conveniently, neatly, and easily in to the smallest hammond with a 9v and never have to open the thing |
| cetoole HeadWizer Joined: Sep. 14, 2005 | Message [#4] posted on: 09-14-2005 02:32 AM CST (US). | | Very nice tangent, I actually started chopping up your origional mint design a couple days ago, and seeing this now is kinda wierd, but it looks great. |
| Nisbeth HeadWizer Joined: Oct. 15, 2000 | Message [#5] posted on: 09-14-2005 02:59 AM CST (US). | | Very nice! I'll be following this thread closely :D
|
amb Headphone Council Joined: Apr. 1, 2004 | Message [#6] posted on: 09-14-2005 03:01 AM CST (US). | | Nice work, tangent. Readers of the AD8397 thread here are no doubt aware that morsel and myself have also been experimenting with a little prototype based on this chip. Looks like the pocket amp scene is really heat up around it.  |
Mister X HeadWize Fanatic Joined: Jul. 23, 2003 | Message [#7] posted on: 09-14-2005 04:42 AM CST (US). | | Nice... can we have the LED on the top of the board please? (this creates a very easy way to mount it) Example: [Edited by Mister X on 09-14-2005 at 04:43 AM.] |
| tangent Headphone Council Joined: Aug. 27, 2001 | Message [#8] posted on: 09-14-2005 04:20 PM CST (US). | | I'm counting on the ground plane to make that advice less relevant than it otherwise would be. But I'll see if I can't move the OG point around a bit. This amp continues my philosophy that wire pads go as close as possible to where they need to be, instead of dragging them halfway across the board just so several related pads can be close by. So, if I have to jump through hoops to do that, the idea will be canned. Yes. People get all nervous about SMDs, but the larger types really are within normal people's ability to hand-solder. For 2-lead parts, you have to go below 0805 before I get worried. Yes, that's one of the things that motivated me to get this into a presentable state. I've been planning MINT v2.0 ever since the TPA6120 news broke, and am glad I put it off until the AD8397 became available. The drying up of BUF634s also helped get me moving. As I understand it you've chosen to use a larger board, through-hole parts, etc. It should be easy for your amp to have more features than mine, and to appeal to the boutique parts crowd, to differentiate it. It'd be easier to just swap RLED and the LED than the way you've done it. I'm not sure I can run that power line down at the bottom of the board and still pass DRC. If it doesn't pass DRC, the tolerance factors mean that some boards will end up with this trace exposed, so if it's mounted flush on a metal surface, it'll energize the case. That'll happen a lot with mint tin amps. |
| jamont Headphone Council Joined: N/A | Message [#9] posted on: 09-14-2005 07:24 PM CST (US). | | This looks very nice. Do you think it would fit in a Hammond 1455C case? |
| cetoole HeadWizer Joined: Sep. 14, 2005 | Message [#10] posted on: 09-14-2005 09:57 PM CST (US). | |
I just talked to a rep from TI today, the buf634 isnt going anywhere, and is one of their most common and popular parts, but they just ran out. It isnt anywhere near EOL. However, for the next couple months, it will be a big pain to try to get them. |
| Mister X HeadWize Fanatic Joined: Jul. 23, 2003 | Message [#11] posted on: 09-14-2005 10:00 PM CST (US). | |
Agreed but I figured WTH I was photochopping anyhow so.... I may as well entertain myself. :) |
| pokeyman Member Joined: Jun. 26, 2005 | Message [#12] posted on: 09-15-2005 11:01 AM CST (US). | | Tangent, the other thing you may want to consider are spaces/pads for small caps parallel to the feedback resistors (as you describe in your "cranky opamps" article). I can't remember if the 8397 datasheet recommends them, but my vague recollection is that the 8656 dataset does. I guess you could always piggyback them on top of the resistors, but it could be tricky Cheers, |
Pars HeadWizer Joined: Nov. 12, 2003 | Message [#13] posted on: 09-16-2005 10:34 AM CST (US). | | Nice Tangent... I thought I was done building MINTs.. maybe not now! |
| Fixup HeadWizer Joined: Oct. 31, 2000 | Message [#14] posted on: 09-16-2005 11:47 AM CST (US). | | Tangent, I'm glad you posted it here - this is THE place for headphone amp DIYers. As this design is for 9V batt and uses two ICs, thus every mA counts, AD45084 is better than AD8397. Actually, use ultra low profile IC sockets is a much better design. At least leave the DIP8 pads there in addition to SOIC pads. Try flood both sides of the board, it'll be better than less vias. You know ExpressPCB now has this great feature. As AD8397 etc. does not have high imput impedance like FET OAPMPs, high value R2 does not make much sense and thus C1 should be huge. A 0.1uF should be paralleled to C2. Try delete the U2A (and use a single OPAMP for U2B), change R+ and R- from 100K to 10K or 5K, parallel them with tantanum caps like in CMoy and for power bypassing as well. I'm saying this in good hearted: a good design is not how elegant it looks on paper, but how well it will perform in real life. I can't see why vias are so bad, especialy in small signal area. Of course, try the best to use less vias, but do whenever is necessary, at least you can use larger vias. Shortest signal paths are good thing, but more important is the position of bypassing caps and power routes, as here it is so small, no matter how you route signal paths, they are already very short. Ground route is the most important, but is less so once you have a gorund plane. That's all I can think of for now. Keep up the good work! |
| pokeyman Member Joined: Jun. 26, 2005 | Message [#15] posted on: 09-16-2005 01:53 PM CST (US). | | Xin, I for one really appreciate these kinds of broad comments/opinions (and of course Tangent's and other peoples too!) >>At least leave the DIP8 pads there in addition to SOIC pads. |
| jcx HeadWizer Joined: Jul. 27, 2002 | Message [#16] posted on: 09-16-2005 10:05 PM CST (US). | | R8 inside the feedback loop isn't going to be usable - adds phase shift R8*Cload to the feedback path a added local feedback c from op amp output bypasses the phase shift at high freq but requires unity gain stbility at 66 MHz - demanding more from bypass and ground layout no reason input ground couldn't be center of R divider for battery power or well isolated \"double insulated\" wall warts then output ground using one op amp for each channel can handle full current of ad8397 |
| tangent Headphone Council Joined: Aug. 27, 2001 | Message [#17] posted on: 09-17-2005 09:43 PM CST (US). | | I've renamed the project. It isn't a META42 made tiny any more. Also, there was confusion in the past beteeen "mint tin amps" and "The MINT Amplifier". So, I'm calling it PINT: PIMETA Is Now Tiny... and because it's a pint-sized amp. <...> (Click for hi-res version.) Schematic, 160 KB Changes since the previous version: - Redid the ground scheme as I proposed earlier: single-channel op-amp fed by TLE2426 instead of resistive divider, with TLE driving ground plane. - Added tantalum caps (3528-21 package) on output of TLE to rails. This isn't the same as Fixup's suggestion, because it has a different primary benefit: per figure 17 in the datasheet, it is only conditionally stable with 0.2uF on its output. (That being the op-amp bypass caps.) Adding 0.47uF or more here makes it unconditionally stable. - Enlarged C1s. Probably room for at least 0.47uF of metalized polyester here, and maybe even 1uF in some cap lines. - Rearranged I* pads to make that possible - Moved amp and ground sections closer together to make a few of the below changes possible - Moved pot away from bottom edge of board. I haven't yet calculated where the center of a mint tin's panel is w.r.t. the board, but the pot is at least close to that ideal now. Also, it allows the corners to be rounded if that's necessary to get it to fit into a mint tin. - Fine tuned ground plane restrictions, primarily to get ground plane away from op-amp inputs. There is an unwanted feedback path otherwise. - Rerouted V- line from rail cap and rearranged LED and RLED much as MisterX suggested. - RLED is now an SMT part (1206, like all the rest of the Rs) - Moved RLED and R1 to back side of board. I'm not wild about this, because it will make the board more expensive, but I'm just plain running out of room on the top side. - Renamed R8s to R5 - Moved part labels on silkscreen layer so they're not overlaying any pads. The main consequence is that the amp section resisistors are no longer labelled with their channel letter.
Almost. The current board height is 0.8", which is just 13 mils too high to fit into the Hammond. You could easily sand that much off the board edges to make it fit, but I'll see if I can't reduce the board height by 25 mils to make that unnecessary. If I do that, it'll have to be on the back side, with vias punched through the pads to bond the two together. The part numbers on the schematic are just examples at this point, not statements of intent. I won't recommend anything formally until I do listening tests, and I see no point in doing that until I get board prototypes. I can easily see situations where test results on a protoboard lash-up wouldn't match what I get on this board, so... Also, as with all my amps, not all DIYers will build this the same way. I am desigining this for 2x9V in a mint tin as a primary goal, but it won't be limited to that. Sorry, but no. I'd have to do something DIY-unfriendly like change to 0603 parts to make room for that. The board can't get any bigger. Indeed, it'll probably shrink a tiny bit. As currently routed, I'd end up with 3 separate islands on the bottom side. Yes. But this isn't ExpressPCB. I'll see about dropping some SMT pads on the bottom side of this cap, now called C3. Vias aren't inherently evil, but I've found via count to be a good metric of layout tightness. The fewer vias you need -- especially in the middle of a long trace -- the tighter the layout usually is. I'm much less worried about vias right at SMT part pins. The only thing you can do to avoid those is what I've already done: find out which signal will have the most connections, and route that on the top. In this case, the most connections are to ground, so the ground plane is on the top side. That's not been my experience. I've seen adding 10-47 ohms here quell oscillations and quiet low-level hiss. If you want resistors outside the loop, you can tombstone them in the O* pads, or tack them onto the output jack's solder lugs. That said, I might remove this feature simply to make room for feedback caps, so I can avoid the via-thru-pad thing I mentinoed above. So far, I've managed to avoid any vias thru pads, which is necessary if anyone is to assemble this with a reflow method. With a via thru the pad, pressure can build up during reflow as the flux boils and blow the part off the board. Where did you get that number? |
| tangent Headphone Council Joined: Aug. 27, 2001 | Message [#18] posted on: 09-17-2005 10:39 PM CST (US). | | There's a forum bug preventing my board image from being seen. Here's a new link, coded differently to get around it:
(Click for hi-res version.) |
silvervarg HeadWize Fanatic Joined: Oct. 12, 2004 | Message [#19] posted on: 09-18-2005 11:48 AM CST (US). | | One of the main problems with the MINT name was that it since it is a normal work it makes it very hard for people to search for it (e.g. with google). PINT is not much better in this regard. Very short names like M3 or A47 is not too great to search for either, so you are in good company... |
| tangent Headphone Council Joined: Aug. 27, 2001 | Message [#20] posted on: 09-19-2005 09:36 AM CST (US). | | Really? "mint amplifier" and "mint headphone" both turn up my MINT page as the top hit on Google. |
| tangent Headphone Council Joined: Aug. 27, 2001 | Message [#21] posted on: 09-19-2005 08:25 PM CST (US). | | v0.92: (Schematic, 184 KB)
(Click for hi-res version.) - Added 0.1uF cap in parallel with big rail cap - Added 100pF in parallel with R4s - Renamed most caps as a result of previous two changes - Added example values to all Rs on schematic (they assume C1 is in place) - Moved D1 to back side of board, and changed its position slightly. This makes other things possible. - Fiddled with the tants on TLE's output, which allowed moving the two traces near the bottom edge up, which allowed the board height to be reduced by 25 mils. Now it will fit in Hammond 1455Cs without a problem. - Enlarged C1s some more; now they will accommodate 1uF caps in the Vishay MKT1826 line, and probably in other PCM-5 metallized polyesters as well. - Verified pot position: it is indeed centered within mint tin panel
|
| Page 1 2 3 4 5 6 7 8 9 10 11 12 13 14 | Next Page | Prev Topic | Next Topic | ||||
 (remove _nospam_ ) |
© Chu Moy, 2001.