[linuxworks]

if you only want to switch BB in or out or even have a pot like PPAv2 does, then why even waste scratchpad for that? the pot has enough room to mount an R and C in parallel and then run THOSE 2 wires back to the r4 l/r holes.

r/c at the pot or at the board - its all the same in the end

 

[rds]

Looks like the other forum is still down. Is there any word on when pcbs will be available?

 

[tangent]

The latest thing holding the production run up is that the AD8620/10 combo was oscillating. I think I found the fix last night, but need to verify it before proceeding.

For anyone who has a prototype and wants to test it, here's the scenario: With a BrownDogged AD8620 in OPALR and a BrownDogged AD8610 in OPAG, and everything else set to stock values, you should see oscillation of 0.5 or 1 MHz -- these seem to be the two stable states of the system, with the choice seeming to depend on the C3-5 values. Sometimes the oscillation only appears at some volume settings, or at some supply voltages.

Known and suspected fixes, some of which are impractical but interesting, are:

- Change OPALR to something else that's always stable. Changing OPAG does not help.

- When scoping OR/L, connect the scope probe's ground clip to IG instead of OG.

- Solder the chips to the board; no BrownDogs, no sockets. I haven't tested this, because I built my prototype mostly thru-hole. I'd have to cut the sockets off the board to properly test this, as some of the stray capacitance would remain otherwise.

- Lower R7. I've tried 1K and 100R, both of which seem to work just as well, despite the instructions in the LMH6321 datasheet to use 10K. Now that my test amp is stable, I need to return to 10K to see if this is a real fix, or if something else I tried fixed it.

- Add a few puff of capacitance across R6L/R. This damps out some severe ringing when testing with square waves. 10 pF overdamps it, but probably doesn't hurt the AF performance audibly. Still, something smaller would probably be better.

Until I've nailed this down more precisely, it will continue to block the production run. Can't allow such a popular chip to be excluded.

 

[TzeYang]

I'm pretty sure regardless, the compensation capacitor is needed. Unless you go super tight SMD but even so I doubt it's still going to output perfect square waves.

 

[NelsonVandal]

Tangent, since you're not using the 1 nF cap bypassing the 10 k input resistor suggested by National I don't think 10 k is an appropriate value. Besides, the LMH6321 is inside the loop of an opamp in PIMETA and is run open loop in the data sheet. I think PIMETA is deemed to be unstable with a 10 k resistor in this position. I've found all my similar amps to be stable without any isolating resistor, but I've read about instablity issues that's been cured by a 100R resistor. I don't think you'll need anything larger than that.

 

[grenert]

These SOIC-DIP adapters
SOCKET ADPTR SOIC/8PIN .300 DIP - 08-350000-10
have traces that take a far less circuitous route than Browndogs. Maybe they'll fix the oscillations. Just don't look at the price

 

[NelsonVandal]

Just to illustrate what happens in this circuit (I have no generic buffer symbol, so I use a discrete diamond buffer instead, but it should make no difference):

with no PIMETA resistor R7, R14 in the schematic above (it looks exactly the same with a 100R resistor)

with a 10k resistor

with an added 1nF cap in parallel, but I can't see why one would want capacitor coupling

with no or a 100R resistor and a 3pF cap in parallel with R6

So you see, even in a perfect simulator-world the amp oscillates with a 10k resistor. Just lower it to 100R. I think it would be wise to have pads for compensating capacitors. You could have SMD 0805 pads on the bottom, under the R6 resistors.

 

[tangent]

Quote:

Originally Posted by Horse /img/forum/go_quote.gif How can there be such a problem with such a stable opamp as the AD8620?

Even slow, docile chips can be made to oscillate, and 25 MHz isn't slow.

It appears from further study of the datasheet that the bandwidth of this particular op-amp is probably what's to blame here, not some special thing about the chip itself. Another 25 MHz chip might behave the same way.

Quote:

what if one wants to use a better (and bipolar) opamp such as the LT1028? Or one of the LME duals? Or an LT1358? This could be easily done if that jung multiloop thingy was left out, one and for all

It's easy to convert the PIMETA v2 to the standard feedback configuration: leave R6 out, and short R5. Balance the input offset currents, and there you are, bipolar-input based amp.

Quote:

Originally Posted by NelsonVandal /img/forum/go_quote.gif Tangent, since you're not using the 1 nF cap bypassing the 10 k input resistor suggested by National I don't think 10 k is an appropriate value.

Ah, I clearly misunderstood what they were doing there. And page 15 is suddenly so clear now. It all but told me this would happen.

Luckily the current board design doesn't have to change to fix it. Not that that would be a blocker. There are already changes coming in the next version. I just don't want to try to make room for more parts if I don't have to.

Quote:

I've read about instablity issues that's been cured by a 100R resistor.

Yes, page 15 of the datasheet covers this, in the Bandwidth and Stability section.

Some buffers, like the BUF634, have this input resistor already. If you go back to the META42, it had this resistor, because the sadly departed EL2001 and 2002 don't have internal input resistors.

Quote:

Originally Posted by grenert /img/forum/go_quote.gif These SOIC-DIP adapters SOCKET ADPTR SOIC/8PIN .300 DIP - 08-350000-10 have traces that take a far less circuitous route than Browndogs.

Aries' real strength is in the breadth of their offerings. BrownDog doesn't have nearly as many adapter types.

Quote:

Originally Posted by Horse /img/forum/go_quote.gif BTW maybe one could have chosen the LME49600 for the buffers.

Or maybe one could not tolerate a footprint 50% bigger than a DIP-8 on a board that wasn't allowed to get bigger just to make room for it.

If they made it in a DIP-8 or SO-8, I'd definitely have tried it. Then again, if it were in that package with the same pinout as the BUF634, maybe PIMETA v2 wouldn't exist. Constructive destruction.

Quote:

Originally Posted by NelsonVandal /img/forum/go_quote.gif Just to illustrate what happens in this circuit

Awesome, thanks for that. The 10K goes.

The schematic won't change for a while yet, as my web host seems to be having network problems. (Ain't my server, really!)

Quote:

with an added 1nF cap in parallel, but I can't see why one would want capacitor coupling

You might be forced into it with some chip combinations. Page 15 again.

Quote:

with no or a 100R resistor and a 3pF cap in parallel with R6

Beautiful, that's almost exactly what I see in the real world with 10pF across R6, only with somewhat rounder corners, as you'd expect.

I have some 1s, 2.2s and 4.7s coming, so we'll see how little capacitance we can get away with here.

So the next question is, do I shove parts out of the way to make room for the new C6L/R, or do I just have people use SMT caps with leaded R6es, and leaded caps with SMT R6es? It's fine with me, but maybe there are good reasons not to do it?

Quote:

You could have SMD 0805 pads on the bottom, under the R6 resistors.

Almost every thru-hole part on the PIMETA v2 board has SMT pads on the bottom side of the board already. They're mostly 1206es, though. Nothing smaller than that.

The only place this isn't done, where it might have made some sense, are the transistors, which could probably accept SOT-23s on the bottom side. Not that explicit pads are really necessary.

 

[MisterX]

Actual 100kHz square wave responce with a 10K R7 and a *slow* OPA2107 op-amp.

Change R7 to 100 and things look a lot better.

(the strange twisty shape is because of the camera)

 

[Voodoochile]

Quote:

Originally Posted by tangent /img/forum/go_quote.gif I think I found the fix last night, but need to verify it before proceeding.

It's good... it's good!
Wish I'd seen this thread 15 hours ago.

I Browndogged an old AMD Duron processor and stuffed it in there; even that doesn't oscillate.

 

[odigg]

Any ideas when the PIMETA board will be for sale again?

 

[tangent]

From the FAQ: "summer".

 

[odigg]

Quote:

Originally Posted by tangent /img/forum/go_quote.gif From the FAQ: "summer".

Ahhh..sorry for not checking there!

 

[cobaltmute]

After too long of having the parts on my bench, another Pimeta 2 is born:



Would take more pictures but the battery in the camera is dead so all I got was one.

Running it currently at ~14V as that is what I plan to put together for a battery pack. Biased by a single resistor to 10ma Class A if I figured the math right.

First pass test was running OPA2107/OPA602 which sounded almost the same as my Pimeta 1 running OPA2107/OPA227 with discrete buffer. Maybe a bit fuller in the sound and a touch more bass.

Currently it is running LME49720/LME49710. Seems a bit more of a natural and full sound than the OPA combo. Have to listen a bit more but this may become my preferred combo.

My main deviations from Tangent's parts list:
- set for gain of 3 (iirc correctly)
- no current limiting on the buffers
- bypass caps on the opamps are 1uF
- currently no C4 caps
- R7 is 221 Ohms - missed the discussion here and had 10K resistors to put in. Then saw the discussion and pulled the 221 out of a box on the shelf.
- My C2's are 560uF Panasonic FCs
- Biasing NPN transistors are BC337s

Have one little issue that if I try to increase the input voltage to 20v, I lose the left channel (just goes to quiet noise) but it comes back if I reduce the voltage. But that will be tomorrow's problem.

Other than that, thanks to Tangent for letting me have one of the first boards. It was a good build. Now I've just got to build my battery pack and case the thing.

 

[tangent]

Thanks for the report!

Do you have the ability to scope square waves thru it?