the Pimeta-V2 thread - Page 29

Quote:

Originally Posted by firefish15 

At the bottom of the board there are four holes. Are these for mounting a daughter board?

Yes.  They're spaced for the now-discontinued Modified Linkwitz Crossfeed board.

Yes, as long as you're careful.

The easiest guide for a safe cut is to just cut off the "tabs" containing the mounting holes.

You can actually cut a little deeper than that, right along the line of the rear-most diode inline with the trace from WALL +, but you risk nicking traces.

Either of these cuts will expose copper, so be careful not to touch anything metal to that raw board edge. You might want to re-insulate it.

Oh, and don't breathe the dust. :)

Thanks Tangent.

Any suggestions on what I can use to insulate the board edge?

 

Quote:

Originally Posted by tangent 

Either of these cuts will expose copper, so be careful not to touch anything metal to that raw board edge. You might want to re-insulate it.

 

Oh, and don't breathe the dust. :)

http://en.wikipedia.org/wiki/Conformal_coating

Thanks MisterX

Quote:

Originally Posted by MisterX 

 

 

 

 

 

http://en.wikipedia.org/wiki/Conformal_coating

Hot glue would work, too.

Quote:

Originally Posted by alphaman 

I'm away from my testbench so it'll be a few hrs before I can re-test.

 

I did change the opamp. Several X's using several diff models: 2132, AD825, and I now have 5532 plugged in. Same confusing result.

 

BUT ... (and not sure I made this clear in the orig. post) ... with the opa (or even buffer) REMOVED I continue to get that 2.6V wrt gnd (so, that's what buffer sees, and sends to the jack as a pwr-on pop). IOW, it's coming from an in situ Q3L. Any clues as to how much voltage SHOULD be present (i.e., w/o opa)?

 

Also this may not have been clear in OP above:

When I changed that Q1 with a broken leg, I also CONCURRENTLY changed Q3L because it looked bent and roughed up (not sure it was damaged but since i was in the area anyway, and had a new 2N3904 on-hand, I thought I'd play it safe). But the orig. Q3L was almost surely a 5088, so me now thinks diff model/manuf = diff bias metrics??

I don't have any 2N5088s in my bin ... just about every other small-signal NPNs, like 3904, etc.

 

EDIT:

Issue Resolved (I think...)!!

 

It was the small-signal NPN  "species" that I'd speculated was suspect, above.

 

I found a BC639 in my bin , and threw it in [CAUTION: despite the European "BC" prefix, this trans. uses Japanese pin config. Mine is a Fairchild.]. No need elaborate: L-ch is behaving normally again -- as is the whole amp AFAICT. Just to re-confirm, I did throw the 3904 and 2SC2240 back in -- in case something else, known to me or not (like cold solder joint, etc), "magically" fixed the issue. But those produced same problem as before -- returning the BC639 into position restored orig. functionality.

Sorry, Tangent, it was the small signal trans. so nahhh ... nahhh ... na... nahh nahh ;)

Seriously, tho ...WOW ... what an epic. I may ditch DIY for a while and give Fiio and others some of my upcoming head $$'s. As  I said above, time=$$ and I'm not sure about the DIY hassle any more. Great sound, good initial learning experience ... but as an ongoing "hobby" ... methinking paradigm shift ;)

Okay, you Pimetans figure this out ... 

As noted a few weeks ago, Q3L was replaced with Fairchild BC639 (the only one I could get to work as I did not have a Fairchild 2N5088 in my bin; remember, a Fairchild 2N3904 did not work -- see above); Q3R was/is Fairchild 2N5088.

One thing I've noticed with the BC639 is that the L-ch turns on with a relatively loud (but not dangerous) transient  from cold start [I never had this transient -- much less ANY transient -- with the orig. 2N5088]. If the unit has been on for a while and/or I cycle power (on/off) quickly ("warm pwr cycle"), no transients audible. I thought this new phenomenon may have been due to different Beta (or other) parameter diff. between Q3L BC639 and Q3R 2N5088. So I ordered new Fairchild 2N5088s. When I installed a new Fairchild 2N5088 Q3L, the L-ch -- as with the 2N3904 -- gave a loud thump (and I once again had that prev.-noted out-jack 2.6V L-ch wrt gnd ... like with 2N3904. And a dead L-ch, of course). I put the BC639 back in and ... voila! ... things are back to normal. WTF!!!???

The ONLY conclusions I can come to are: Fairchild has recently changed the parameters of (at least) some of their transistors; and/or the Pimeta's topology, WRT the class-A biasing ckt, is too damned sensitive to sample-to-sample (batch) diffs in small-signal NPN parameters (Beta, etc.)
 

Notes: No, I did not remove/replace Q3R 2N5088 to test its parameters. I'm not superstitious but I don't want to jinx this touchy piece of kit any more than I have to! And its a royal pain to remove (de-solder) parts from that crowded-footprint space.

No.

The whole point of assigning part numbers to transistors is that they are labels for a specific bag of parameters. When manufacturers change the contents of the bag, they change the label, too. This is why there are so many types of bipolar transistors on the market despite the fact that they're all fundamentally the same device.

It is common for parts from different manufacturers that have the same base part number to have slightly different parameters. But, these differences will remain over the long term, because each manufacturer isn't going to change the parameters without at least adding a new variant code to the part number scheme.

Also no. I told you that two weeks ago, and it's still true.

Are you ready to consider the possibility that I might just have some idea of how this circuit works, and that when I tell you that there is no magic transistor type, it means the problem lies elsewhere?

It's probably time for you to post high-res pictures of the top and bottom sides of the board.

Measurements of the voltages at the op-amp pins relative to IG would be good, too. Maybe you did that already in the past 2 pages and I don't see it in skimming, but you should re-measure anyway, since you've gone and changed the circuit.

Any chance of scoping the op-amp outputs to check for oscillation?

Quote:

Originally Posted by alphaman 

Tangent wrote: "Are you ready to consider the possibility that I might just have some idea of how this circuit works, and that when I tell you that there is no magic transistor type, it means the problem lies elsewhere?" 

 

Not sure what type of omnipotence you are claiming here???

I think you mean "omniscience," but regardless, the answer to your question is "none." I am neither all-powerful nor all-knowing. I am merely claiming to have some confidence that I have applied some well-known principles of electronics correctly.

I think I have several reasons to be confident:

1. The part of the circuit you're blaming is well-known outside this little world of headphone amps. More below.
2. The PIMETA v2 is about 3 years old now, and none of the others built so far has shown this same problem. That suggests that the problem is in your particular amp, not in the amp design. If the problem were in the circuit design, why are you the first in three years to have this problem?
3. It's been two weeks since your original claim to have diagnosed the problem. Where are the confirmatory reports from those who have also been affected? Absence of evidence isn't evidence of absence, but all my training in how Internet geek communities work tells me to expect some piling on by now if you were right. Silence, on the other hand, means either no one else has a better idea, or that everyone else has gotten bored of the discussion and moved on.

I am well aware that there's a fine line between confidence and arrogance. That's one of the reasons I remained silent for these past two weeks. Instead of piling on, we got crickets.

So you want an argument from authority? Okay, I refer you to pages 73 and 90 in The Art of Electronics.

I just re-read these sections, and I see only two quibbles you can bring against the PIMETA v2 class A biasing circuit from what's given:

1. I don't specify matched transistors in the mirror. This means the per-channel currents aren't necessarily absolutely identical. Since we can be within a factor of 2 or so and still get the behavior we want here, I don't see that this matters. But if you think I'm wrong, go and match your transistors, then see if that fixes things.
2. Instead of pinning the voltage drop across the current-setting resistor with an LED, Horowitz and Hill use either a zener or a diode string. I talked about the diode string alternative in the docs and explained why I thought the LED is better. The zener option has the same squishiness problem, and is nosier besides. Besides, I got the LED+bipolar current source idea from Jung, which should satisfy your need for authoritative reference. But if you feel Jung and I are wrong and Horowitz and Hill have the right of it, go ahead and swap in a zener or diode string for LED1. I predict it will perform about the same as it does now. You'll lose one LED, and you'd have to make some careful measurements to show the flaws in the modified circuit, but the main point is that I don't believe it will change your symptoms.

Apparently you are assuming that when I called this circuit "somewhat novel" that I think it is some grand invention, created from whole cloth, and that I am therefore some kind of electronics super-genius.

All I meant is that I had never seen anyone use this particular arrangement for op-amp class A biasing.

You should be able to show this circuit to any EE graduate and maybe get, at most a "huh, that's interesting" reaction. I imagine many would be somewhat bored by it; it is after all, AoE Chapter 2 material. Not exactly high tech stuff.

I already told you why that happened. You will get massive DC output when removing the op-amp from any buffered op-amp headphone amplifier design. When you break the feedback loop, there is nothing to prevent the biased buffer input from being pushed to one rail or the other.

This is not a flaw of the PIMETA v2 or an indicator of a fault in your particular amp. It is simply the way the circuit works when you do that.

You are asking me to add a cautionary note that if you remove essential parts of the circuit that it will not work correctly? That's not obvious?

Quote:

Originally Posted by alphaman 

 Only that 400mV L-ch transient, which is not too bad. And I can get around it by plugging in cans AFTER pwr-on.

One other mod that I did concurrently with the initial opa roll was putting two 9.6v NiMH in SERIES (to give a total of ~20v) as the new opa that I rolled in was a bit more voltage hungry. With my Pimeta2, I normally have the two 9.6's in PARALLEL for longer run time .

Anyway, to make a long story short, I went back to the two-batts-in-parallel (or even a single 9v) arrangement and the L-ch transient is GONE. All totally normal again.

Batteries are strange creatures. NiMH 9v especially so IME. So another species (alk, LiPo, etc)  may work better IN-SERIES.

I've also got some wall-based reg'd DC adapters, which output higher-than-10vdc, so I may experiment sometime with them.

Say tuned... 

You risk battery damage by putting rechargeables in parallel. You don't have to believe me on this, either. Check any NiMH manufacturer's application manual.

Hi,

As mentioned a couple of weeks ago, find herewith the Desktop version of my Pimeta v2 amp with the following features:

- Integrated power supply with the r1 regulator - 25V DC

- Selectable 3 line Input and 3 lines Output

- Integrated modified Linkwitz Crossfeed circuit

- Integrated Bass Boost (pot) and Volume Boost (switch) circuit

Here a few pics:

The main boards (Pimeta v2, Linkwitz Crossfeed and r1 regulator) assembled and ready for installation.

As a support, I've chosen an acrylic plate that I can slide into the Hammond enclosure.  Bottom-left see a little board for the power cabling, while on the top-right side the micro-switches have been mounted on another little board.

From the bottom, with the I/O connectors and the back panel.

These little boards are used as indicators on the front panel for the selected I/O lines - the selectors are on the back panel.
 

Here the Bass and Volume Boost board - it has taken me a while to get the components for the best audio experience - that can be mounted as piggy-back directly onto the Pimeta v2 board.

Lastly, another little board that mounts the Bass Boost pot and an LED supposed to indicate the operation mode (Headphone = green, Preamp = red).
 

Here from the bottom, pretty much ready to be put into the enclosure.
 

Same from the top, all modules are plugged-in and the amp has undergone all tests.
 

Sliding it into the enclosure, there is not much room left for other stuff.  I've tried to isolate the toroid transformer (german manufacturer, high quality) as good as possible, and luckily there is no electric or magnetic interference.
 

Here the finished product - front and back view.  Note the holes that I had to drill in the cover to allow some airflow - a bit of an artistic touch with the 4 inside LEDs spreading rays of blue light.

Here another view of the two units I've built.

My objective was to have a modular design with all the features I needed to have.  In order to eliminate the hum, I had however to ground the IG to the case.  Else everything worked as per plan, exception made for the 'Preamp' mode of operation that is miserably poor...

Cheers!

Originally posted by firefish15:

The second one will do and will fit into the Serpac case, no worries.  You will have to pack it along one of the long sides of the battery pack.

I had to go for a 453048 spec, I don't believe the 603040 will fit into the case if you pack 4 cells together with the PCB, the cabling, the insulation, etc.

You can trim a bit the board as explained by tangent.  Note however that 600-700 mAh is plenty of capacity, you don't need 900 mAh.  In my case I can operate the amp for about 10 hours, even though I've biased the amp to class A with 1 mA on RBLIM.

That's an exceptionally beautiful build, mfuerst!

As for the "miserably poor" preamp mode, I suspect you have some sort of ground loop problem. Since the PIMETA wasn't designed to do that, I haven't collected much info on how to hack it to behave as you want. My vague recollection is that people have managed to make it work by disabling [part of?] the ground channel, so that IG comes from the outside. This means IG goes floating when you unplug the source cables, but again, the PIMETA (v1 or v2) wasn't designed to do this, so...

Hi Tangent,

Thanks for commenting on the preamp mode quality issues and potential solutions.

I appreciate that the Pimeta v2 wasn't built as a preamp, but the motivation to build this feature came from the Project FAQ page on your website:

What I did is to add a switch on the scratchpad area that would toggle between "headphone" and "preamp".  A picture may be helpful:

The molex connector right beneath is meant to power the dual-color LED in the middle of the front panel and switch from green to red respectively.  When in "preamp" mode, I do short the Ground Channel as per schema below:

With the gound channel shortened for "preamp" mode, the effect is that the sound is deamplified to about half the volume (I believe because of the Linkwitz filter) and the sound characteristics is also very poor, like if the low and high frequencies have been cut.

To be honest I didn't have time to debug and find out the root cause, but I would appreciate any comments from fellows that are using the Pimeta v2 as a preamp.

Thanks!