[digger945]

I wouldn't mind to help out also, if there is anything I can do(stateside).

 

[cobaltmute]

As many of you may have noticed, this thread has sat idle for quite a while. While I have been around the forum, most of the time that I have done this while I have been at work. My personal time outside of work has severely constrained due to a number of personal issues.

A result of this is that I do not think that I would be able to run the group buy on this. I do not want to get into a situation where I have taken money from people for boards and do not have the time to properly package and ship them in a timely manner.

In the past there have been a couple of people who have shown interest in helping out on this. If someone wants to take this on, they are more than welcome.

I've attached the gerbers for the PCBs. As far as I'm concerned they are public domain. Sijosae designed the buffer, Mono modified it, and I simply put it into gerber form.

I apologize if anyone is disappointed by this. Personally, I'd like to see this board out there. Unfortunately, something had to give from my perspective

 

[sachu]

Wondering if these buffer PCBs can be used in a DAC on the output stage.

I have OPA627s doing buffer duty right now per chanel..

Would the Sijosae bufer be an improvement over the 627s?

 

[Nemo de Monet]

Cobaltmute, are you serious about considering them "public domain"? Would you be willing to formally release them under a free license?

If so, I believe Seeed Studio would be willing to make them available for sale quite cheaply; they are big on promoting open source hardware, especially PCBs. I/you/someone would have to contact them to be sure, but given the size of the boards, I'm willing to bet they'd be able to provide pairs of the buffer boards for $10 USD or less, including worldwide shipping.

Not as good a deal as a formal group buy, perhaps, but a way for folks to get the boards with minimal hassle, and without anyone having to facilitate an actual group buy. Could be a win-win for everyone...

I've worked with 'em in the past; the owner is pretty well respected in the *duino microcontroller-hobbyist circles, and is basically an all-round good guy.

(edit: I know some people have had reservations in the past about sourcing PCBs in the far east, but in the absence of someone in North America to facilitate a group buy, something like this is probably the easiest, simplest way for those who'd like to build these buffers to get them.)

 

[cobaltmute]

I just posted the Gerbers in a public forum, how much more public could I make them?

Doing them was a great learning experience for me. It would be great practice for someone just learning to put mono's design on a PCB. It makes you learn the tool since the hard part - optimizing - of the layout is already done.

 

[Nemo de Monet]

Well, there's public, and then there's public, you know?

FWIW, I emailed Eric of Seeed Studios, and already heard back from him; he's evidently willing to produce, stock, and sell the buffers at $2 USD/board; shipping to North America will run ~$2.50 per package. I'm not sure how long before he'll have 'em in stock, but I think we all should be able to order them within a week or so. The Chinese New Year is coming up at the end of the month, which basically shuts the country down for two weeks, so I'm hoping we'll be able to get this rolling before then...

Eh, nothing ventured, nothing gained...

 

[mono]

... but etching it yourself is half the fun!

It's not my design, just rotating some parts around that were a common topology that Sijosae made more popular in certain DIY circles. If someone tried to lay claim to it we could just rotate the parts around again, there's too much prior art for anyone to lay claim to the electrical aspect of it.

 

[Nemo de Monet]

Well, that was fast!

Those who've been waiting for the buffer PCBs can now order them here.

Eric is willing to carry complete kits, if anyone's interested. I know the bc337/bc327 aren't too hard to source, globally, but I believe there are some other combinations - particularly some of the Toshiba 2SCnnnn parts - that can be hard to find. Is there a transistor combination that anyone wants to see made available?

 

[mono]

A complete kit would be really great. I don't claim to know which transistors have the best synergy with any particular combo of amp and cans but something capable of more current than BC327/337 would be nice, to move up to a TO-126 size like BD139/140 for the output pair, and in that case if the amp is for home use to bias it further, replace 1K resistors with 470 Ohm.

What about the size of those PCBs? Pimeta has space for 12mm dia. capacitors, but on mine I only used 10mm dia and even then by etching it myself I was able to use a little less than 0.1" spacing (except for the pins corresponding to the DIP socket pins on the Pimeta PCB so they'd line up) which kept the ground buffer from interfering with the capacitor next to it. It looks like Seed Studio is using 0.1" spacing so it may be too long to fit in the ground position. My use of non-standard spacing was a kludge for sure but it was the quickest path towards making it fit.

Perhaps it'd be easier to see what I mean with pictures, here's the Pimeta PCB, and an early prototype that has one buffer etched from the pattern in the project GIF image, and in the ground I still had one built on a full 0.1" spaced DIP socket. Note how the 10mm capacitor had to sit sideways to make it fit, I had to desolder it and back it out of the hole on longer leads. I suppose people could use 8mm caps instead of 10mm and it "might" fit that way, but the Pimeta PCB hole spacing would require bending the leads of the caps outward and they'd sit off the PCB a bit because of that, and of course in a short case you're even further limiting the capacitance with 8mm caps.

 

[Nemo de Monet]

Seeed are using CobaltMute's gerbers, so, yes, everything's on 0.1 centers. It worked for him, but I'm not sure what he had to do to make it work, if anything.

Yes, the ground placement on a Pimeta could be tight, but IMO there are workarounds. For one, you have a little wiggle room towards the middle of the board - around 4mm before you start running into the output film cap. (And even more room towards the back of the board, if you don't use Molex connectors for the power wires.) For another, I think the determined builder can get by with one cap per rail - Nichicon's HN or HZ series have 16v, 1000uF caps in a 10x16mm size, and there's even a 16v, 470uF cap in an 8x11.5 size. (On both of mine, I used a 10x20mm, 220uf Nichicon HE in each of the four positions, and am fairly confident I'm not lacking capacitance... )

Also, I suspect there is no real penalty for having the ground buffer here - any of the buffers, really - "floating" above the board, like a lot of the JISBOS installations in SOHA I amps have been done. People with really cramped cases might not have a lot of options, but folks building a "desktop" Pimeta should have enough room to figure out something that will work for them.

Also, let's not forget - there are uses for these buffers other than in a Pimeta.

How much more power do people really need? Aren't the BC devices good for something like 800mA of current and 500mW or so dissipation?

 

[mono]

Higher current transistors also tend to sound a little smoother, not as bright. Sound quality improves with more class A biasing but if the transistors aren't larger they get fairly warm and may drift in specs. By simply moving to a larger die, casing, and often heat spreader on the back of larger case size, the thermal density is reduced.

Since this is a portable amp there seems a higher potential to accidentally short the headphone socket while it's turned on which would fry the smaller transistors more quickly. You don't really want to think of transistor specs as staying within the maximum instead of having some margin either. Regarding frying them, I have with BC327/337, and later put (forget the exact value at the moment) ~ 6 Ohm resistors on Pimeta's R8, L & R PCB positions.

I don't see what you mean by wiggle room since the PCB pins have to align with the socket pins under them, unless the components are shuffled around again but if the pins to the socket are just moved upwards without again decreasing the PCB size, they'll not fit in the L & R channel buffer positions so there's need be a separate PCB layout for the ground channel instead of keeping them all the same.

I agree one cap per rail could work, but many people won't want a 16V limit and IMO 470uF per rail is not enough for some higher output current scenarios. It just seems a lot of concessions would be made to stay with 0.1" spacing, maybe it is better to use a CRD or diodes instead of one of the pairs of transistors like a Lisa II buffer uses to shrink the board size if it needs to be 0.1" spaced.

Maybe the best option is to just use 12mm caps for the leftmost two and 8mm for the right. Since Tangent designed the PCB with a staggered layout of alternating voltage rails per each cap in the row, that would keep capacitance per rail equal.

For home use, I'd definitely go with more than 1000uF capacitance, the larger transistors, and more biasing... or just build a PPA2 instead.

Floating above the board seems a pretty ugly solution, aesthetically worse than just tilting one capacitor sideways and we really shouldn't make this signal path any longer than absolutely necessary because it is in the feedback loop. Next time I build a Pimeta I'll use BD139/140 and skip having a DIP socket on the board at all.

 

[cobaltmute]

One of the things that I did to get the board into the ground channel was that I left the leads slightly longer so I can move the buffer in slightly towards the middle of the boards. Bigger caps in all the rail positions could be a problems so you have to be sure what you place where.



You probably should do Nemo's idea and run with just two larger rail caps in the Pimeta. If it is not for portable use, you could leave the leads a little longer and angle the cap away from the buffer.

 

[cobaltmute]

Oh, and I was going to try MJE243/MJE253 on the output side, but they were too tall to fit into my case.

 

[Nemo de Monet]

Quote:

Originally Posted by mono /img/forum/go_quote.gif I don't see what you mean by wiggle room since the PCB pins have to align with the socket pins under them, unless the components are shuffled around again but if the pins to the socket are just moved upwards without again decreasing the PCB size, they'll not fit in the L & R channel buffer positions so there's need be a separate PCB layout for the ground channel instead of keeping them all the same.

What I meant was what I think CobaltMute did - since you've got, what, 0.5in leads out the bottom of the buffer to serve as "legs", fitting into the socket, you just - carefully! - put doglegs in the leads, as necessary. Insulation might be a good idea, obviously...

Quote:

I agree one cap per rail could work, but many people won't want a 16V limit and IMO 470uF per rail is not enough for some higher output current scenarios. It just seems a lot of concessions would be made to stay with 0.1" spacing, maybe it is better to use a CRD or diodes instead of one of the pairs of transistors like a Lisa II buffer uses to shrink the board size if it needs to be 0.1" spaced.

You could have chosen to raise these concerns before boards were actually fabbed, you know. I mean, this thread has been around for, what, three months?

My point was really that there are a lot of options - just because there are four, 12mm spots for capacitors on the Pimeta board, doesn't mean that people have to populate all four of them, and certainly not with 12mm caps. To judge from Cobaltmute's picture above, the whole thing of fitting the board in the ground buffer position is really a non-issue for everyone who doesn't have enormous connectors in the way.

Quote:

Floating above the board seems a pretty ugly solution, aesthetically worse than just tilting one capacitor sideways and we really shouldn't make this signal path any longer than absolutely necessary because it is in the feedback loop.

There's no doubt it's ugly, yes, but - just like with the JISBOS - you have to make compromises when you want to do things the original board wasn't designed for.

 

[NelsonVandal]

BC639/640 are the same as BD139/140 in a TO92 package, at least Philips/NXP are. BC639/640 from ONSemi sound worse than BD137/138 from ONSemi, so they probably use different dies.

If you run these buffers class AB, heat will never be an issue. Even biased to something like 20 - 30 mA, heat is no problem. The transistors are just lukewarm. Compare this to monolithic opamps/buffers like AD8397, LMH6321 or BUF634. AD8397 is specified to deliver 300 mA - per channel, SOIC!