Group Build: Dynafet

[luvdunhill]

Looser101: not that it matters all that much, but is your order for 3 stereo amps, or one balanced and one stereo? Just trying group some of the harder to group transistors tonite and this would be marginally helpful

[Looser101]

Looser101: not that it matters all that much, but is your order for 3 stereo amps, or one balanced and one stereo? Just trying group some of the harder to group transistors tonite and this would be marginally helpful

One balanced, one stereo.

Thanks

[luvdunhill]

excellent!

So, on that note, despite the fact I'm not finished with the n-ch MOSFETs, I feel that I will in fact have a few matched devices left over from my contribution to the MOSFET order after I take the devices out for another project. After a lot of thinking, I will only be offering them as matched pairs, not quads per the rest of the group buy. Also, I'd like to sell the rest of the devices with them, excluding the JFETs (unless they don't sell and I get really, really bored). I'll make sure to order enough boards to cover these possible builds as well. I'm tempted to ask the same price as the group buy ($60 a board), despite the reduction in parts, just due to the unforeseen additional effort with the MOSFETs, and the fact that it's pretty hard to get decently matched pairs without buying two times the needed devices, so I feel that the market will bare this price. I plan on posting a graph for the p-ch parts sometime today and you'll see what I mean I want people to realize that these aren't just sub-par matches, but in fact are even better complimentary matches than the other sets, just not available in quads, which really makes this possible as they are drawn from the "tail" of the distribution. Please PM me if interested.

[luvdunhill]

This graph is worthless without the compliment devices, but ...

[luvdunhill]

well, I have the boards and have started working with one and getting the other ready to send to Kevin. Not an easy board to stuff so far... A few initial suggestions, in random order:

1.) The silk is sort of messed up in places. Kinda like printing on a bad ribbon, for those that remember those days, it's offset at points. I'm assuming this is an issue with the fab, not the files?

2.) The output transistors I think could be better labeled using "2SJ" and "2SK" rather than Q27-Q34. I'd remove that A17,5mm text as well (whatever that means).

3.) The top ground plan extends too far and will be under my aluminum angle, which I'm uncomfortable about. I'm not sure if this will be the case for everyone, but I'd like to move back the ground plane to the most northern edge of the pads. This will cause some of the ground plane in the middle of the output devices to be unconnected and thus removed, but this is okay.

4.) The "1145 / 2705" legend will be partially hidden by the bracket. I'd move it down a bit. Also, perhaps another legend would be more appropriate, as there are four transistors in the middle that are at a 90 degree angle and have no reference designators that I need to track down on the board files to see what polarity they are... maybe add a dot to all the 2SA devices right above the center pin? That way you could accommodate the mystery devices in the middle of the board that are currently at a 90 degree angle from your legend.

5.) If at all possible, there needs to be more room around the output resistors. Especially between R53,R54,R49,R50 and the row below it (R44,R46,..). The bodies on these three rows of resistors practically touch. Any more room would be greatly appreciated. Looking at the layout, you can create some more room at the bottom quite easily. Move the JFETs toward the power connector and move the regulators around a bit and move up R57,R58, etc. You get the idea. More room between the first row and second row of the output resistors, as well as more room between the second row of the output resistors and the R44,R46... row. I think we definitely need to do this.

6.) R33 and R32 actually touch the caps. These should be moved out a bit, you have plenty of room.

7.) On most the caps, I'd remove the "A" and "B" nomenclatures. It gets confusing. Less text is better. Also, I'd remove the smaller silk screen and just keep the larger outline on C11B, C12B, C13A, (and the unnamed capacitor to the right of the power connector), C32B, etc. It's clear enough with the single large outline with the four pads inside and helps the silk look less busy.

8.) The resistor between R28 and R32 has no printed reference designator, that I could find.

9.) On C1B, the south hole could be enlarged a tiny bit, as I cannot quite fit my capacitor in there. I can give the component dimensions, but my guess is as large as the IN connector holes would be great. While everything fits, it is tight in there. I'd move C11B to the left a bit and then C1B to the left as well, since currently everything touches.

*10.) In my opinion, the resistor foot print is just too small. I know it will be hard to enlarge it, but I think a lot of people aren't going to like the fact that you have to slam the leads against the body and then the ring gauge on the board is so darn small. I realize it's work to fix this, but I think it will be worth it, as it's very, very tight. I'm just preempting Pars complain on this matter I dunno what to say... it's very, very tight.

11.) I'd add a silk layer to the bottom to explain the rather clever jumpering system and a note to the top of the board explaining that the bottom jumper must be populated. I didn't even notice it myself until I looked at the EAGLE files.

12.) For the four 7zip patterns, I think this will be confusing to stuff, since we're using the small device heat sinks. I know mistakes will be made and people will have to consult the data sheets as is. See attached image for what I'd recommend to do here for the silk.

13.) Depending on the connector you want to use for the 4 pin SIP to the left of the power connector, I'd recommend putting a bit more space between here and the cap. For example, you might want to use a Molex KK connector that has the retention clip, and currently it wouldn't fit.

14.) For the two regulators, I'd use the TO-92 silk pattern. Since the higher voltage parts aren't available in TO-220, I think we should just use this pattern and emphasize this on the silk. The polarity needs to be marked somehow as well.

Ok, more tonite after I finish stuffing the boards. Simply a gorgeous design, by the way... hats off!

Edited June 17, 2009 by luvdunhill

[Smeggy]

You write all this and don't add sexy pics for reference?

Pics or you made the whole thing up!!

[Pars]

<snip>

*10.) In my opinion, the resistor foot print is just too small. I know it will be hard to enlarge it, but I think a lot of people aren't going to like the fact that you have to slam the leads against the body and then the ring gauge on the board is so darn small. I realize it's work to fix this, but I think it will be worth it, as it's very, very tight. I'm just preempting Pars complain on this matter I dunno what to say... it's very, very tight.

Preemptive bitching... what a fun concept! Yes, too small of resistor and pad footprint = a no go for me. I'm only in for two boards, so certainly don't do anything solely on my behalf however.

11.) I'd add a silk layer to the bottom to explain the rather clever jumpering system and a note to the top of the board explaining that the bottom jumper must be populated. I didn't even notice it myself until I looked at the EAGLE files.

Good idea. I need to look at the Eagle files to see what these are, but documenting the jumpers right on the board is pure win.

12.) For the four 7zip patterns, I think this will be confusing to stuff, since we're using the small device heat sinks. I know mistakes will be made and people will have to consult the data sheets as is. See attached image for what I'd recommend to do here for the silk.

Good as well if you do a custom part in Eagle. I may use duals, and will almost certainly socket whatever I use. I would like to use these, which require a decent sized pad (the Dynalo boards will take these; the Dynahi boards will not).

13.) Depending on the connector you want to use for the 4 pin SIP to the left of the power connector, I'd recommend putting a bit more space between here and the cap. For example, you might want to use a Molex KK connector that has the retention clip, and currently it wouldn't fit.

Absolutely.

[luvdunhill]

I would like to use these

I'll check tonite, but these should fit fine for the 7zip packages. Would that suffice?

Mouser 575-410322010

[Pars]

I have some of those (28 pin) and yes they work. The flush pin sockets are much slicker, however They do require drilling (a slight bit) and are dependent on the pads/holes being of sufficient size that you don't go through the thru plating. Not a biggie, but if they will work, I would use them, even for singles. I used these for my work dynalo for all transistors, and it worked out really nice.

Edited June 17, 2009 by Pars

[luvdunhill]

I have some of those (28 pin) and yes they work. The flush pin sockets are much slicker, however They do require drilling (a slight bit) and are dependent on the pads/holes being of sufficient sizer that you dont; go through the thru plating. Not a biggie, but if they will work, I would use them, even for singles. I used these for my work dynalo for all transistors, and it worked out really nice.

eh, you lost me after the drill part

[Pars]

The normal pad/hole size is not big enough for the flush sockets, so I had to drill them out slightly (on the Dynalo board). I have a pretty full set of carbide PCB drills

[digger945]

I've never tried some of the Mill-Max products you gave a link to, tried some Mill-Mas inline(8 pin, 32 pin, etc. sip package) type ones and the holes in the pin sockets were never big enough to put the leads into, or at least anything I was using at the time.

I finally settled on the Molex 64 pin ones, cutting the plastic away leaving just the pin to solder into the board. They do stand a bit proud of the pcb, but only the largest leads are too big to fit into the receptacle. They have a nice positive lock when the lead seats into the socket, making it hard to mistake if you have the lead in far enough.

I think the Mouser part is 315-93-164-41-003000

Easy to use, just give one side of the plastic a nip with your cutter and the hatch right out.

Thanks for the update Marc. Having poured over the pbc images for weeks myself I know it can demand a fair amount of attention and get tiring quickly.

Look forward to some pics.

EDIT:The link is for a Mill-Max part, I can't find the Molex and Tyco parts I used right now, maybe tomorrow.

[luvdunhill]

A few pics that illustrate some of my points above:

[j4cbo]

1.) The silk is sort of messed up in places. Kinda like printing on a bad ribbon, for those that remember those days, it's offset at points. I'm assuming this is an issue with the fab, not the files?

That does look like a fab problem. I'm surprised; PCBFabExpress has done better with the other boards I've ordered from them. Might be worth contacting them to see what's up.

2.) The output transistors I think could be better labeled using "2SJ" and "2SK" rather than Q27-Q34. I'd remove that A17,5mm text as well (whatever that means).

Will do.

3.) The top ground plan extends too far and will be under my aluminum angle, which I'm uncomfortable about. I'm not sure if this will be the case for everyone, but I'd like to move back the ground plane to the most northern edge of the pads. This will cause some of the ground plane in the middle of the output devices to be unconnected and thus removed, but this is okay.

Looking at it, yeah, that should be no problem to change.

4.) The "1145 / 2705" legend will be partially hidden by the bracket. I'd move it down a bit. Also, perhaps another legend would be more appropriate, as there are four transistors in the middle that are at a 90 degree angle and have no reference designators that I need to track down on the board files to see what polarity they are... maybe add a dot to all the 2SA devices right above the center pin? That way you could accommodate the mystery devices in the middle of the board that are currently at a 90 degree angle from your legend.

I've actually got a more compact TO-92 layout that I think I'll try to switch to (it's used in my custom layout for my own builds, here). Hopefully that will allow making all the devices face the same two directions.

5.) If at all possible, there needs to be more room around the output resistors. Especially between R53,R54,R49,R50 and the row below it (R44,R46,..). The bodies on these three rows of resistors practically touch. Any more room would be greatly appreciated. Looking at the layout, you can create some more room at the bottom quite easily. Move the JFETs toward the power connector and move the regulators around a bit and move up R57,R58, etc. You get the idea. More room between the first row and second row of the output resistors, as well as more room between the second row of the output resistors and the R44,R46... row. I think we definitely need to do this.

I'll see what I can do, but I'm not too worried about the positioning now... they're staggered, so heat shouldn't be too much of an issue, and the actual terminations are well-spaced.

6.) R33 and R32 actually touch the caps. These should be moved out a bit, you have plenty of room.

7.) On most the caps, I'd remove the "A" and "B" nomenclatures. It gets confusing. Less text is better. Also, I'd remove the smaller silk screen and just keep the larger outline on C11B, C12B, C13A, (and the unnamed capacitor to the right of the power connector), C32B, etc. It's clear enough with the single large outline with the four pads inside and helps the silk look less busy.

Will do.

8.) The resistor between R28 and R32 has no printed reference designator, that I could find.

Whoops, that's R31. Not sure how it got unlabeled.

9.) On C1B, the south hole could be enlarged a tiny bit, as I cannot quite fit my capacitor in there. I can give the component dimensions, but my guess is as large as the IN connector holes would be great. While everything fits, it is tight in there. I'd move C11B to the left a bit and then C1B to the left as well, since currently everything touches.

Will do.

*10.) In my opinion, the resistor foot print is just too small. I know it will be hard to enlarge it, but I think a lot of people aren't going to like the fact that you have to slam the leads against the body and then the ring gauge on the board is so darn small. I realize it's work to fix this, but I think it will be worth it, as it's very, very tight. I'm just preempting Pars complain on this matter I dunno what to say... it's very, very tight.

I don't think there are any spots that immediately preclude enlarging the outer pads on the resistor footprint, actually. I'll see what I can do there.

11.) I'd add a silk layer to the bottom to explain the rather clever jumpering system and a note to the top of the board explaining that the bottom jumper must be populated. I didn't even notice it myself until I looked at the EAGLE files.

Adding silkscreen on the bottom will increase the cost a bit. Maybe just put a label in copper and then a little more explanation on top?

12.) For the four 7zip patterns, I think this will be confusing to stuff, since we're using the small device heat sinks. I know mistakes will be made and people will have to consult the data sheets as is. See attached image for what I'd recommend to do here for the silk.

That looks great.

13.) Depending on the connector you want to use for the 4 pin SIP to the left of the power connector, I'd recommend putting a bit more space between here and the cap. For example, you might want to use a Molex KK connector that has the retention clip, and currently it wouldn't fit.

14.) For the two regulators, I'd use the TO-92 silk pattern. Since the higher voltage parts aren't available in TO-220, I think we should just use this pattern and emphasize this on the silk. The polarity needs to be marked somehow as well.

Will do. Let me know what else you find.

Ok, more tonite after I finish stuffing the boards. Simply a gorgeous design, by the way... hats off!

Thanks! I want more photos once it's all stuffed

[luvdunhill]

I've actually got a more compact TO-92 layout that I think I'll try to switch to (it's used in my custom layout for my own builds, here). Hopefully that will allow making all the devices face the same two directions.

one other option.. you could easily make the two paralleled transistors before the buffer face each other (flat side would mate on the silk), as well as the two transistors in the CCS so that their faces could be bound together with a zip tie and some thermal grease. (so, all the non buffer transistors could be bound together in pairs, three pairs total) I've done this and it works very well. You can move a single trace to the top and do this quite easily. I think I'd slightly prefer this. What do you think? Then you'd need a way to mark the polarity, maybe a dot or shaded version of the outline? What other TO-92 silk patterns does EAGLE have?

I'll see what I can do, but I'm not too worried about the positioning now... they're staggered, so heat shouldn't be too much of an issue, and the actual terminations are well-spaced.

Look at the photo. See how the second row output resistors are almost right up against the Dale resistors? The PRP are a bit fatter in the worse case (their body seems to have a looser tolerance) I think as well. Just a bit more room here would be nice, if you cannot also separate the two output resistor rows. It's really tight in there...

I don't think there are any spots that immediately preclude enlarging the outer pads on the resistor footprint, actually. I'll see what I can do there.

what about a hair more pitch, say 0.5mm? This is probably more important, as the PRP resistors are kinda fragile if you slam the leads against the body like we need to here. The larger pads seem very doable, as I'm looking at the silk now and would be greatly appreciated.

Adding silkscreen on the bottom will increase the cost a bit. Maybe just put a label in copper and then a little more explanation on top?

perfect!

I think we can safely remove C32B. The other input capacitor is basically equivalent, and this way is better given the feedback capacitor.

The mica option for the feedback cap doesn't quite fit. I dunno why CDE has odd pitches, but I'd look at the datasheet for the one in the BOM and make the distance from the second hole from the bottom to the top hole this exact pitch, as I had to really shove it in to get a good fit...

If you look at my pics, I'd move in the 2SA1139/2SC3381 pairs towards the middle of the board about 1mm or so (I'll measure to be exact, but look at the one on the right in the pic, you'll see how much room you have). The fins of the heat sinks are very close to the resistor leads.

We will almost assuredly have to move the two JFET pairs closer. I didn't have time to build and document how to put these together tonite, so perhaps tomorrow I'll get to it.

edit: so you see why the silk is confusing for the regulators by looking at the pics.. alternatively / in addition, we could put surface mount parts on the bottom of the board, but I think just using the TO-92 parts and scrapping the TO-220 silk is fine...

Anyways, things are looking like I'll power it up on Friday late or Saturday evening.

Edited June 18, 2009 by luvdunhill

[luvdunhill]

I'll see what I can do, but I'm not too worried about the positioning now... they're staggered, so heat shouldn't be too much of an issue, and the actual terminations are well-spaced.

one thing to do at a minimum is work a way to add a test point in somewhere, so we can easily measure across the output resistors to set the bias. This has PPAv2 type slip ups written all over it. I started installing the transistors, but thought better of it after 6 or so. I want to verify the schematic to make sure I get the polarities right.

[j4cbo]

OK, I've got most of the changes done on the board. I'll hold off on finishing it up until the final JFET layout and test results are ready.

What do you mean by "PPAv2 type slip ups"?

[n_maher]

What do you mean by "PPAv2 type slip ups"?

It's pretty easy to short the PPAv2 while trying to test and adjust the bias which results in the near total failure of that channel. Not fun, trust me.

[DigiPete]

awesome work by all. We should have a great product when this is all done.

Now I hope it sounds as good as we all hope!!

[luvdunhill]

Gotta head out, and will be busy until tomorrow night Maybe I can get another set of eyes to double check the transistors? I'm most worried about the orientation of the 2SC3381/2SA1349 and JFET "singles". I double checked it, but you never know.

From top to bottom, left to right I have:

JJJJ KKKK (not populated yet)

ACC CAC

ACA AAC

AC

CC AA

AC

AA JJ KK CC

... and pics to show orientation of the "duals"... zoom all the way when you view the pic, as the light made things hard to capture, and seems to make it look like there are some flaws in the dual assembly, which are pretty much smooth when looked at straight on... oh well, just a bit self conscious I guess

Tomorrow I need to jumper the regs, jumper the feedback and install the MOSFETs. Then have to bolt my sigma22 to the heat sink and get things ready to flip the switch. I'll probably load test the sigma22 first, assuming I can find enough power resistors around here.

[digger945]

Shweet.

Got wood.

[DigiPete]

Wow! that is one jam packed board

The Dynahi looks simple in comparison

[luvdunhill]

ok Jacob, everything fits. No need to move the JFETs, they are perfect. Still need to double check the orientation of the TO-92 parts, but it will have to wait until tonite, unless someone else can look.

The jumpering for the regs is a bit annoying (pin 1 to 3 on one and pin 1 and 2 on the other), and given how slick the feedback jumper is, could a similar SMD jumper be added to the bottom of the board for the regulators?

[digger945]

The inputs and CCS's are good, checking the buffers and others now.

[digger945]

You have a 20k pot parallel to the 681 ohm for the negative rail bias adjust(the trimmer on the left in your pic).

What value is the one in the middle that we can't see. It should be the 20k for Vbe multiplier adjustment.

Looks like the one on the right is 10k, as it should be.

I'm still looking, slow as it is.

I like your solder, care to share what brand it is.