[cobaltmute]

I've got an old signal generator (not sure how well it works), but no scope.

I really need one, am cheap, but something that could monitor the signal lines in a DAC.

 

[cobaltmute]

Figured out that it is something in the Class A bias circuit that is running away when the supply voltage passes 18V. I rotated op-amps (in both OPALR and OPAG) to see if it was the opamp and experienced the issue in all combo's I could do. On further testing as I increased the supply voltage, the current draw wouldn't change and then it would start slowly increasing. Snipping the RLIM resistor (removing the biasing) results in stability at 20V supply.

I'm going to replace the BC337s with 2n3904s and see what happens.

The LME49720/LME49710 combo does sound very good. I'd recommend giving it a try - especially for the price of the chips.

I'm going to look around at a couple of the local stores and see if I can rent a scope for a day or two. Can't justify buying one yet.

 

[tangent]

That's odd. I would expect that if the biasing circuit were the problem, the symptom wouldn't go away until you got rid of the Q3s, as they're the only parts to "touch" the feedback circuits. The CCS side of things should be quite decoupled from the op-amps.

Before you swap the transistors, try putting a switch in for RBLIM and see if you can correlate class A bias current with stability. Without RBLIM, you'll be able to dial up bias current way beyond practical levels, which may be educational. It would be useful to know if you can get it stable somewhere in the RBIAS range without having to break the CCS path with the switch.

 

[cobaltmute]

As a note, you may have missed - I don't have RBIAS in line right now. I put RLIM across RBIAS + RLIM. I was set for doing fairly deep biasing (10ma), but well within spec for the chips (the LME49720 can output 25ma). Cutting RLIM took out all the Class A bias.

Part of the reason for this was that I was going to do something different and not use the bias at all, but at build time I changed my mind and threw in what was in my parts bin.

I'll check to see if I can get a pot at the local shop and get some adjustability.

 

[cobaltmute]

Tests made wiith a 340Ω RLIM resistor and a 5KΩ RBIAS pot

12.5V supply, Highest setting on pot - stable (50mA draw for amp)
15V supply - stable (52mA)
17.5V supply - stable (53mA)
20V supply - stable (54.5mA)

Starting at 12.5V supply with a 60mA draw:
15V - stable (61.5mA)
17.5V - stable (64mA)
20V - stable (67mA)

Starting at 12.5V supply with a 65mA draw:
15V - stable (69mA)
17.5V - possible very slow runaway
20V - slow runaway

Starting at 12.5v supply with a 76mA draw (maximum)
15V - stable (83mA)
17.5V - runaway

What I don't understand right now is why the CCS is not stable for a particular setting - by looking at the schematics, I would figure that it would be independent of the rails for the large part. I can see a small current increase with the increase in supply voltage, but it seems to be a bit dramatic. Playing with this in simulation shows small increase in current for change in rails, but nothing like what I see in real life.

Any thoughts as to what might be messed up are appreciated.

 

[tangent]

Many chips vary in current draw as the supply voltage varies. I, too, have observed that the LMH6321s are in this class. Perhaps your op-amps are, too.

What value do you have in R7? Do you have any small caps across R6, as discussed above? Both changes (relative to the original docs) aren't necessary in my experience, but then, I have never used the chips you're playing with.

 

[nullstring]

I am thinking about making a pimeta sometime next year as my first PCB amp

I am curious, should the pimeta v2 be better, worse, or about the same performance wise vs the pimeta v1.

the pimeta v1 certainly looks more impressive than the pimeta v2 from an outsider and amateur perspective. more large capacitors, etc.


I want to make one like tomb's
http://www.head-fi.org/forums/f6/diy...7/#post3991061

I think those look really really nice. I plan on using 4 to 6 CR123A's instead of the 9V batteries and adding li-ion charger inside as well if possible.
I'd also like to change it so that the LED is multicolor and will change color and maybe flash based on charging status.
Of course, this is gonna require space.. and so anything the saves space is nice, but I just hope that the original performance of the pimeta is unhampered =)

This is all completely in the planning stage, however. I don't plan on building it till maybe next summer.

 

[JamesL]

A quick look at the schematic looks like there's no considerable changes in the design.
The most evident changes are the addition of some very nice features. It looks like a trickle charger and bass boost was introduced in v2, as well as a scratchpad for additional features or experimentation. some changes regarding the class-a bias.
SMD pads to allow for almost a all-smd build, layout changed to accommodate the lmh6321 without modification, a mounting hole in the front to replace the evj-c20 support.
That's all I can see so far... Performance-wise, I'd expect it to be similar, but I'd like to hear it from tangent as well.

 

[JamesL]

Quote:

Originally Posted by nullstring /img/forum/go_quote.gif the pimeta v1 certainly looks more impressive than the pimeta v2 from an outsider and amateur perspective. more large capacitors, etc.

What does "etc" pertain to? The 2 rail capacitors lost in v2 can easily be soldered onto the scratchpad. I wasn't particularly interested in building a v2 before but after looking into it, I'm definitely looking forward to it's release =)

 

[cobaltmute]

Quote:

Originally Posted by tangent /img/forum/go_quote.gif Many chips vary in current draw as the supply voltage varies. I, too, have observed that the LMH6321s are in this class. Perhaps your op-amps are, too. What value do you have in R7? Do you have any small caps across R6, as discussed above? Both changes (relative to the original docs) aren't necessary in my experience, but then, I have never used the chips you're playing with.

R7=221Ω R6=499kΩ. I don't have any caps across R6.

IIRC, the OPA2107/OPA602 were in the amp for testing.

 

[cobaltmute]

Quote:

Originally Posted by nullstring /img/forum/go_quote.gif the pimeta v1 certainly looks more impressive than the pimeta v2 from an outsider and amateur perspective. more large capacitors, etc. I want to make one like tomb's http://www.head-fi.org/forums/f6/diy...7/#post3991061

My Pimeta 1 is like tomb's using 12xAAA's instead of 2x9V's. This allows for lots of run time (800mAh vs 250-300mAh). My Pimeta 2 will be cased the same way.

As for the number of caps etc between Pimeta1 and Pimeta2, there are a lot of things to take into consideration. IIRC, my Pimeta1 uses 4x330uF caps for a total of 1320uF. For my Pimeta2 which is the same kind of tomb like build, I'm using 2x560uF caps for a total of 1120uF. The area of those two caps is a little larger in the Pimeta 2 so you can get away with a larger (not taller) cap. Yes, I did lose a little capacitance, but you can always add one in the scratchpad area if you really feel like more.

 

[cobaltmute]

As another note, after some longer sessions, I'm starting to not like the LME49720/LME49710. It may be "fuller" and "warmer", but they are not without their flaws that I find annoying.

 

[hopeless]

Quote:

Originally Posted by cobaltmute /img/forum/go_quote.gif As another note, after some longer sessions, I'm starting to not like the LME49720/LME49710. It may be "fuller" and "warmer", but they are not without their flaws that I find annoying.

What exactly? I'm curious... To me they're overly smooth in the upper mids, making them appear very detailed (and a bit spiky) in the treble while they aren't necessarily.

Plus they're blue-sounding like the rest of the series.


Try the LT1355 and LT1354? Or LT1358 and LT1357?

Or LT1469 and LT1468?

 

[hopeless]

Ouch, wait, I guess the LT's are worth nothing to you.

 

[tangent]

Quote:

Originally Posted by JamesL /img/forum/go_quote.gif A quick look at the schematic looks like there's no considerable changes in the design.

- The new class A biasing scheme removes previous requirement for hand-matched transistors
- The new biasing scheme is adjustable, with a single pot adjusting both channels' bias
- 4-layer board, allowing full ground plane and heat-sinking for the buffers
- better bypassing

Quote:

Originally Posted by cobaltmute /img/forum/go_quote.gif R7=221Ω R6=499kΩ. I don't have any caps across R6.

Try dropping R7 to 100R.

Quote:

Originally Posted by cobaltmute /img/forum/go_quote.gif my Pimeta1 uses 4x330uF caps for a total of 1320uF

The rail caps in the v1 are from the rails to virtual ground, so you can only count half the capacitance.

Actually, I can make an argument that you only get a single cap's worth of overall capacitance in PIMETA v1: two equal caps in series have half the capacitance of the single unit (but twice the voltage tolerance) and two in parallel add: (330 / 2) + (330 / 2) = 330.

In PIMETA v2, the caps span the rails, so total capacitance is either equal to that of a PIMETA v1 using the same caps, or twice as much if you buy my shaky analysis.