New Posts  All Forums:Forum Nav:

PPA Instability

post #1 of 24
Thread Starter 
A well worn topic here that never seems to go away.
It is the hissing that results from using low impedance high sensitivity phones with the PPAv2.
It is not a gain issue. I tried a gain of 2 which actually increased the hiss over a gain of 11 using grado gs-1000s. To me this indicates that the PPA requires large global feedback to minimize instability.
Ok so what I tried.
All opa627 - hiss definitely present and annoying
All opa827 - hiss is loud enough to drown out quiet passages in music (ie really bad).
Using 100pF silver mica in c6. Tried a ceramic 100pF which was slightly worse.
Using recommended MJE243/253.
3.3 ohm buffer resistors.
Buffer biased to 30mA.
Interesting note is the hiss goes away completely if I connect a 10 foot extension cable between the PPA and grados.
Any help would really be appreciated

What I haven't tried is replacing the pot. I'm using this stepped attenuator. As a last resort I could try to replace that.

EDIT Just want to add that this is a great amp and sounds amazing with k701s and hd650s. Somehow it just doesn't like low impedance phones (at least mine doesn't).
post #2 of 24
Maybe it is better to ask your question in
post #3 of 24
I had a similar problem with the PPAv1 and Glassman buffers (see here ). Pre-loading the output eliminated the noise, but it was pointed out in that thread that it wasn't a proper solution. Your extension comment makes me think that perhaps the extra wire is just enough extra load to bring it back into stability.

I never got around to playing with different values for C6 and I ended up selling the Grados to try something else so I didn't revisit the problem.
post #4 of 24
To me it sounds like it's oscillating and you should try to add a zobel on the output. Try a 47 nF cap in series with a 47R resistor from output to ground.

I also think you should add a small cap in parallel with R6 (opamp's local feedback), something like 10 pF.

I wonder if the recommended value for R11, 1k, really is OK. I think a smaller value would be better.
post #5 of 24
The Jung article buffer has a 10 Ohm series resistor on the output:

ELECTRONIC DESIGN - October 1, 1998 - Walt's Tools And Tips

This really messes up the output impedance of the amplifier and the L/R crosstalk (it is after the feedback point), but it is easy to try as a diagnostic measure?

That said, I am running a 3-channel Jung multi-loop with diamond buffers on a diy pcb and it is very stable and quiet (scoped down to 15 ohm load). R8 = 0 ohm, R5,6 = 3.3 ohm - part names refer to the link, not the PPA. BD139/140s on the output. Modest OPA2132/134s for the VAS, G=4.

Nelson already mentioned the 10pF on the local feedback. I see slight ringing on a square wave input without this cap. This is not audible though, so I run without it. The M^3 does have it, so worth a try though.
post #6 of 24
I found that where the OPA627 would have obvious instability, another opamp, such as the OPA134 or the AD8610 would work just fine. FYI.
post #7 of 24
Thread Starter 
Ok, so I've tried swapping opa134 in right and ground (I only have two) and that had no noticeable effect.
I tried putting 100pF (what I had on hand) in the feedback loop (of the right channel op-amp) and that increased the hiss significantly.
So in summary lowing the gain increases hiss. Adding a feedback cap increases hiss. Swapping op amps does nothing noticable. Adding a headphone extension cable practically eliminates all hiss.
The cable itself when plugged in measures 0.4 ohm from the buffer output (OG for example) to the female end of the cable. I'm thinking it's the complex impedance that results in stability.
Thanks for the comments so far. Any further thoughts?
post #8 of 24
Thread Starter 
Potential Solution Found.

Adding a 100pF cap in parallel with r11 kills the hiss completely. How does this sound as a solution?
It would be nice to hear from Tangent, PPL, or Morsel on this one as they were the design team afaik.


Reducing r11 to 100 ohm has the same effect of killing the hiss. This makes me feel like I've found the culprit. Maybe having r11 at 1k is just causing too much reflection between the op-amp and buffer? 200 ohm does the job too... though their might still be a tiny hiss, but definitely barely perceptible.
post #9 of 24
Do you have a resistor from the output of the attenuator to the pcb inputs?
I know absolutely nothing about the PPA so...
What size resistor do you have from pcb input(signal) to ground?
post #10 of 24
Thread Starter 
I've replaced r11 with 220 ohm resistors and the hiss is completely gone... at least with the grado gs-1000i - which is very sensitive. Haven't tried iems yet.
I'm interested to hear some thoughts on this.

Originally Posted by digger945 View Post
Do you have a resistor from the output of the attenuator to the pcb inputs?
I know absolutely nothing about the PPA so...
What size resistor do you have from pcb input(signal) to ground?
4.32k from the pot/attenuator to the op-amp non-inverting input and 1M from the non-inverting input to ground.
There schematic is here.
post #11 of 24

I built several PPA style amplifiers on perfboards before even with extra upgrades and modifications but never ran into issues with hiss when it comes to low impedance headphones.

Until I had to fix my friend's PPA (official PCB) which exhibits the same issues as OP, trying to crack my head day and night yet I could not find the problem.

thanks rds, i'll give your method a try. Though it does sound a bit "off" as in why the interstage resistor would cause an issue at the first place. The resistor is there for stability at the first place.
post #12 of 24
Originally Posted by rds View Post
I've replaced r11 with 220 ohm resistors and the hiss is completely gone...
I would call the cap across R11 a hack, but since it was a successful hack, it means lowering the value of R11 is a legitimate fix. Something at HF isn't liking all that resistance between the op-amp output and the buffer input, so the fix is to lower the resistance, not eliminate it.

I recall discussions during the design phase where we decided going down to 100 ohms would probably be fine, but that we should make it as high as possible to *reduce* the possibility of problems. You'd have to ask someone else why lowering it seems to reduce problems rather than increase them. It doesn't make sense to me.
post #13 of 24
I'm glad your amp is quiet now.

I think what happens when you use an extension cable is you get some inductance between transducers and amp. If you look at speaker amps, there's often an inductor on the output.

In speaker amps the general recommendation is to use a zobel on the output. I use it in my discrete headphone amps, and it's often needed to kill oscillation. I think more should use this trick, because a lot of PIMETAs, PPAs, CMOYs etc are borderline unstable with heavy peaking when simulated. I don't have an oscilloscope so I have to trust sims.
post #14 of 24
^Agreed. Make a small extension with a male on one end an female on the other(obviously). Wrap it a few turns around a bar ferrite and tape it or heat shrink it. Plug it in between the amp in the phones. See if your problem persists. If that fixes it look into fitting proper inductors to the output.

The only other explanation would be the increased resistance of the extension, which I doubt because the resistance of 10f of cable is less than minimal.
post #15 of 24
Thread Starter 
I appreciate the comments guys, but a properly implemented amp shouldn't need those kind of hacks. They come at a cost.

The right interstage impedance is always the solution to coupling a low output impedance stage to a high input impedance stage.
I can't say why 1k is too high, but a complete analysis of the small signal model would probably reveal the culprit somewhere in the complex impedance of the buffer stage.

I agree with the cable comments. My guess is that it's the capacitance to ground that kills the high frequency noise.

Thanks to Tangent for weighing in on this. It's always nice to hear from the actual designer(s) as their hands on experience in testing is better than our (un)educated guesses. Also, I'm very cautious about even slight modifications to other people's designs.

EDIT I wanted to add that I don't think the actual design is in any way flawed. The issue is might be a function of different current gains for different transistors. It is my understanding that in a configuration like the diamond buffer the input resistance is determined almost entirely by the current gain of the transistors. As we know current gain has a huge range in any given transistor. If the stability is related to the current gain then it may be an issue that different r11s are needed for different builds.
I may be way off track here too
New Posts  All Forums:Forum Nav:
  Return Home