Head-Fi.org › Forums › Misc.-Category Forums › DIY (Do-It-Yourself) Discussions › Troubleshooting my Pimeta2
New Posts  All Forums:Forum Nav:

Troubleshooting my Pimeta2

post #1 of 11
4/9/10 at 9:25am
Thread Starter 
Finally got around to stuffing the Pimeta2 PCB I ordered a few weeks ago...

<> Some preliminaries:

Built pretty much per Tangent's schematic...
- Op-amps are AD8620 for L/R and OPA132U for G
- Buffers are LMH16321MR.
- Using both LED1 and LED2, HLMP1790s
- RLED=2.7K
- RLIM=1k-ohm (per schematic)
- RBIAS=5k pot (per schematic)
- class-A engaged

- Battery-power, "9V" cell: 7.2V 150mAh or 9.6V 230mAh. I have not tried any other pwr option.
- Current draw at battery: ~60.0mA (measured with Fluke DMM)
- V drop across RLIMB=1.064V, so Ibias=1mA (I can't squeeze more than about 1.25mA per Tangent's schematic values?? Not sure if more Ibias will make a diff. IAC??)

<> The amp was "successful" upon first test. Sounds great, no distortion. But extended testing/experimenting revealed two "issues"...

(1) After a few minutes of playing (using either Senn IE-8 IEMs or Beyer DT880), using either of the batteries noted above, the sound "cuts off" suddenly (battery drain I assume). Randomly, there may be some classical characteristic "fuzzy" distortion of drained batt. before the cut-off -- but not always. I tested the batts *after* suspected drainage using a gauge I trust (the gauge successfully predicts batt life for other 9V-powered headphone I amps I use, such as the CMoy or Go-Vibe). Weirdly, *both* batts test "good" *after* suspected Pimeta2 drainage. If I recharge the batts, I can repeat the above...i.e., cuts off after a few mins. If I don't recharge, and simply cycle power, I'll get sound for a few millisecs before the sound cuts off as noted above; the LEDs stay lit with LED1 being *slightly* dim. Smells like batt drainage...so...Why are these batts draining so quickly, even when low-power IEMs are used?

(2) Unreferenced (TTBOMK) 0.1uF capacitor and switch around R4 L/R in Tangent's *schematic* only. There is, TTBOMK, no reference to either the cap or switch in Tangent's write-up -- please correct me if I'm wrong
What are these for?
Experimentally, I added the 0.1uF caps and got a slightly richer, more fwd *but* treble-muffled sound. I might like this sound if the treble wasn't so muted.

Any help is appreciated!
Reply
post #2 of 11
4/9/10 at 10:26am
The cap and the switch by R4 are part of optional bass boost circuit, see the tweaks sections of his site.

Keep you meter on between the battery and the board and measure the current draw and see if it slowly increases until the sound dies. I had a run-away condition somewhere in the biasing if I went too deep on my board. Condition was the same as yours in that it run and then shut-off. This seems due to current limiting in the op-amps.
Reply
post #3 of 11
4/9/10 at 6:20pm
Quote:
Originally Posted by headfone View Post
V drop across RLIMB=1.064V
No, closer to 1.25 V. That's 1.9 V for the LED minus approximately 0.65 V for the Vbe drop.

Quote:
I can't squeeze more than about 1.25mA per Tangent's schematic values??
You're right on the current -- 1.25 V across 1K = 1.25 mA -- but keep in mind that you changed the LED1 voltage drop from my recommended value. You should therefore decrease RBLIM to compensate if you want as wide an adjustment range as in my schematic.

That said, 1.25 mA is probably enough.

Quote:
battery drain I assume
Don't guess, find out. With NiMH, it won't be subtle: the voltage at the terminals will drop sharply under load when this happens. You might keep the DMM connected to it while this happens to watch it in real time. If you don't do that, don't turn the amp off, then test, as the battery voltage will climb back up when you take the load off. You have to do the terminal voltage test under load.

Quote:
Randomly, there may be some classical characteristic "fuzzy" distortion of drained batt. before the cut-off -- but not always.
Could be clipping. See Op-Amp Working Voltage Considerations

Quote:
Why are these batts draining so quickly, even when low-power IEMs are used?
The IEMs have nothing to do with it. 60 mA is a fairly hefty load for such small cells. You couldn't do it at all with nonrechargeables.

If those batteries are weak, this load may be pushing them past what they're capable of delivering.
Reply
post #4 of 11
4/10/10 at 6:35am
Thread Starter 
Quote:
The cap and the switch by R4 are part of optional bass boost circuit, see the tweaks sections of his site.
Thx! I have no interest in BB so I ignored that section. IAC, it should be labeled on the schematic.

On to more experiments...

Quote:
Originally Posted by tangent View Post
Don't guess, find out. With NiMH, it won't be subtle: the voltage at the terminals will drop sharply under load when this happens. You might keep the DMM connected to it while this happens to watch it in real time. If you don't do that, don't turn the amp off, then test, as the battery voltage will climb back up when you take the load off. You have to do the terminal voltage test under load.
I ran two experiments: with class-A on and with it off. No real diff. in results except that with Class-A off, the batt lasted a 4-5 minutes more (not surprising since class-A bias only adds 2-4mA addit. current draw).

The input and output jacks were not "loaded". I.e., no (input) source nor (output) 'phones were plugged in.

Results for Class-A on experiment:

Time: 00:00
V(no load)=10.36VDC (Maha PowerX 9V cell)
I(nl)=0.0mA
Note: To test for shorts between the batt and main pwr switch, the batt was plugged into Pimeta2 during the "no load" ("nl") test; the main pwr switch (Alps) was OFF.
V(load)=10.33VDC and falling 0.01V every ~3sec.
I(l)=60.2mA

Time: 00:03
V(l)=9.9VDC and falling
I(l)=60.2mA

Time: 00:16
V(l)=3.3VDC and falling (but more slowly)
I(l)=24.5mA
LED1 is no longer lit (AFAICS) and LED2 is dimmer.

Since I was not monitoring via phones, I presume the "cut-off" happend between Time: 00:03 and Time: 00:16. The slower drain, henceforth, may be due to opamps and/or transistors no longer ON (conducting).

Quote:
60 mA [current draw at 9V batt.] is a fairly hefty load for such small cells.
Is 60mA current draw "normal" for Pimeta2, given the opamps/config I'm using? (See above for opamps/config ). I seems quite high.

For comparisson, I have two Go-Vibe that also run off 9V. Both draw 7.7mA and the V(load) is only about 0.8VDC below V(no load).

Quote:
If those batteries are weak, this load may be pushing them past what they're capable of delivering.
I tried several different NiMH "9V" batts. -- Maha Powerx 9.6 230mAH (I own and tested with several of these) as well an Eveready 7.2V 150mAh. All fully charged. Tested them all with a 9V tester as well as in the two Go-Vibes. Bottom line: the batts are fine.

Puzzling...
Reply
post #5 of 11
4/10/10 at 8:54am
Quote:
Originally Posted by headfone View Post
Time: 00:16
V(l)=3.3VDC and falling
Okay, so the question is did that happen because the battery is failing, or because the current load increased?

Quote:
Is 60mA current draw "normal" for Pimeta2
Yes. Each LMH6321 draws ~15 to 20 mA, varying depending on a few different conditions.

Quote:
Tested them all with a 9V tester
I expect they're fine, but realize that this test is far from conclusive. Few battery testers put the battery under any real load, and those that do only do that for a short time. You still need to test the current draw under load to exonerate the batteries.

If, as I expect, you find that current draw increases sharply at some point, causing the precipitous drop in battery voltage, you then have to find out why.

Pics?
Reply
post #6 of 11
4/10/10 at 6:03pm
Thread Starter 
Quote:
Originally Posted by tangent View Post
I expect they're fine, but realize that this test is far from conclusive. Few battery testers put the battery under any real load, and those that do only do that for a short time. You still need to test the current draw under load to exonerate the batteries.

If, as I expect, you find that current draw increases sharply at some point, causing the precipitous drop in battery voltage, you then have to find out why.
I agree about the battery tester but as far as testing under load, I thought that's what I did with the ~20 min test noted above?? If you built the schematic Pimeta2 on your site, and powered it with an NiMH 9V batt., how long did it last under "normal" use (i.e., with source music; what cans did you use?)?

BTW: I did do the finger heat test several times, and none of the PCB components felt even warm. I'm not sure what would cause a catastrophic (= all-of-sudden), perhaps random-occuring, and perhaps random-duration current spike, which you seem to suggest. What could do that?
Reply
post #7 of 11
4/10/10 at 6:42pm
You did a voltage test. Now you need to do a current test. Yes, I know, you say 60 mA, but is that continuous clear through the failure, or does it change? It must if the battery is not at fault. So, what does it change to?
Reply
post #8 of 11
4/10/10 at 9:02pm
Thread Starter 
Quote:
Originally Posted by tangent View Post
You did a voltage test. Now you need to do a current test. Yes, I know, you say 60 mA, but is that continuous clear through the failure, or does it change? It must if the battery is not at fault. So, what does it change to?
Watchdogging current for spikes/ anamolies requires infinite attn span and/or special test tools -- I have neither That said, and in a just-completed test, I did have my DMM actively monitoring current for over 4 hours. Its display was at eye level, next to my PC monitor, while I worked on myriad computer assignments. I used the 7.2V batt and switched class-A OFF. W/o input signal or cans, I ran the Pimeta2 in "idile" mode. Weirdly, the batt. didn't drain for over 4 hrs. Current draw dipped over the hours by only a few milliamps (60 - 57). Not sure what's going on, but this test's results are mythologically cliche: when you're watching it, it behaves!! But I'm not religious: don't believe in angels or gremlins.
You're right that the batt. tests are inconclusive! I'm conducting a few more empirical tests that may, indeed, point the finger on the (my) batts. Stay tuned on that as I run them and crunch numbers.
Meanwhile...
Can you give a ball-park est. of how much additional current a Pimeta2 -- use your schematic component choices (which roughly parallel my build also) -- may eat when, e.g., its is in active use powering IEMs while a line-level signal (music) is being inputted?
Reply
post #9 of 11
4/10/10 at 9:51pm
Quote:
Originally Posted by headfone View Post
Watchdogging current for spikes/ anamolies requires infinite attn span and/or special test tools -- I have neither
I wouldn't have asked if I thought it required special tools or infinite patience. Above, you said it last happened after 16 minutes. What changed?

Quote:
don't believe in angels or gremlins.
Heisenbugs are real.

Quote:
Can you give a ball-park est. of how much additional current a Pimeta2 -- use your schematic component choices (which roughly parallel my build also) -- may eat when, e.g., its is in active use powering IEMs while a line-level signal (music) is being inputted?
Short answer: negligible compared to the idle current. Single digit mA, in all likelihood.

Long answer: There's no easy way to calculate it. It depends on how deeply the amp's output stage is pushed into class A (which we can't know without detailed design specs from Nat Semi), the RMS value of your music (which we could theoretically calculate but aren't going to bother to), and the impedance of your headphones (which is complex and depends on frequency).

When anyone attempts a calculation like this, they make several simplifying assumptions to make it feasible. You often see people talking about a plain sine wave into an ideal resistive load, for instance.

If your cans require 0.5 Vrms to sound as loud as you'd ever want to listen and they're 32 ohms nominal, it takes 16 mA to drive each one, 32 mA for stereo. But, the real number is much lower because some of the current comes from the output stage biasing with buffers like these, and even modern pop has more than 0 dB dynamic range. It could easily be an order of magnitude lower, all considered. Thus my ballpark value above.
Reply
post #10 of 11
4/12/10 at 5:47am
Thread Starter 
Still having issues with my Pimeta2...
While listening, the audio in one of the channels became "fuzzy" then cut off. I started fingering the components to isolate, say, a cold solder joint. Indeed, pressing down on one of the film caps restored sound. I re-soldered the joint but the problem happened again, and pressing down on the same cap area restored the sound. I wonder if the PCB is defective? I know that it's multi-layer with a ground and power plane (TTBOMK). May be out on a limb here but ...Could there be some "leakage" or "shorting" between layers due to sloppy PCB manuf.? This may also explain the batt. drain issue.
Reply
post #11 of 11
4/15/10 at 11:08am
Quote:
Originally Posted by headfone View Post
I wonder if the PCB is defective?
Because of the 4-layer design, I had the PIMETA v2 boards electrically tested at the factory. If the board no longer electrically corresponds to the schematic, it would have happened after that test.
Reply
New Posts  All Forums:Forum Nav:
  Return Home
Head-Fi.org › Forums › Misc.-Category Forums › DIY (Do-It-Yourself) Discussions › Troubleshooting my Pimeta2