[zxc]

Quote:

Let's say zxc is trying to make himself deaf rightquick

No, I hate to listen to music at high volume levels but actually this was another question that I wanted to ask. The pimeta worked great with the Sony CD changer and all the portable players I had; but for some reason I had the following problem with my desktop computer with the M-Audio 2496 Audiophile soundcard:

At 1/4 volume there's no sound. At 1/2 volume, suddenly the sound kicks in; but it's loud and very very distorted (cracky). As I turned the volume up to 3/4, the distortion fades away but the volume is so high it's unbearable. So what I had to do was to lower the volume on the computer as much as I can; and make the volume high on the pimeta to avoid distortion. Since the amp was working fine on everything else; I blamed the sound card for having a weak signal. But then I didn't have this problem when I connected the soundcard to my home stereo. Any ideas why this might be happening? And is there a way to fix this within the amp? (The problem became worse when I switched from 637/627 to 823/843.)

I will build a new Pimeta next weekend and socket all the buffers (and fix this one too) but I think I will have the same problem again...

 

[tangent]

Quote:

Originally Posted by amb Given your line of reasoning, then there shouldn't be any benefit to stacking the buffers on the L and R channels either.

Not at all. I can hear the benefit from stacking the L&R channels. I cannot hear a benefit from stacking the ground channel. Before I tested it, I believed as you seem to. (Assuming you're not just being the devil's advocate.)

EDIT: One improvement we get from stacking buffers in improved slew rate, but on the ground channel, the buffers are not slewing. The other is lower output impedance, but on the ground channel this shows up as lower crosstalk, which is very hard to hear past a certain point with most music. So, it's not surprising that improvements are hard to discern past a certain point.

Quote:

At 1/4 volume there's no sound. At 1/2 volume, suddenly the sound kicks in; but it's loud and very very distorted (cracky).

Sounds like it's oscillating. Is C4 installed?

If the amp was this way from the start, your problem probably always existed. It just stayed off the radar until now.

Quote:

I blamed the sound card for having a weak signal

No, if you adjust it correctly, it should be capable of putting out no less than 1V, and probably 2V. Plenty, given that the minimum gain on a PIMETA is 2 (without removing resistors).

 

[PeterR]

Quote:

Originally Posted by tangent Not at all. I can hear the benefit from stacking the L&R channels. I cannot hear a benefit from stacking the ground channel. Before I tested it, I believed as you seem to. (Assuming you're not just being the devil's advocate.)

But that means that if there's an audible improvement from increasing the number of buffers in the L&R channels it cannot be explained by increased current capability. If it was a factor (and it usually isn't as your calculation demonstrates) it would have the same influence in the ground channel, as L/R channel and ground channel are in series.
The big difference between the channels is the higher amount of feedback in the unity gain ground channel. This should reduce the beneficial effects of stacked buffers, like lower output impedance or reduced distortion - maybe that's the reason?

 

[tangent]

Quote:

But that means that if there's an audible improvement from increasing the number of buffers in the L&R channels it cannot be explained by increased current capability.

See my EDIT above. Stacked buffers affect the amp differently on the L&R channels because of differences in how they operate.

 

[zxc]

Quote:

Sounds like it's oscillating. Is C4 installed?

Yes, both C4's (6.8uF Wima) and C6 (10pF ceramic) are installed. Quote:

If the amp was this way from the start, your problem probably always existed. It just stayed off the radar until now.

Actually, sometimes I heard signs of distortion in female vocales and high pitched sounds with my CD player at low volumes... Is there anything that I can do at this step to get rid of this oscillation?

 

[dip16amp]

Quote:

Originally Posted by tangent All of what current? Let's say zxc is trying to make himself deaf rightquick, and is putting 2V through Grados. That'll give you well over the 90dB maximum you should listen to, but it only causes 63mA of current to go through each channel. And this only happens when listening to sine waves, not music! A single buffer can easily handle 126mA. The idea that the ground channel has to have some number of buffers (whether equal to the number in each audio channel, or twice, as some would say) is fallacious.

Any instantaneous sound such as a drum snap would require an instantaneous voltage change. It doesn't have to be at a real loud volume, but that instant voltage change on the load (headphone) requires infinite current flow for that power pulse. Since infinite current flow can not be provided, the voltage change (slew rate) will be slowed down by how much current can be provided. Normally the ground channel will be the return path of the current flow from both left and right channels. This is why you see the lowest crosstalk with more buffers on the ground channel. If one channel exceeds the ground channel capacity, then it will end up with return current in the other channel. Then it has more crosstalk because the channels are crossfeeding each other. It may sound better with crossfeed and this may be cheaper than adding crossfeed circuits, but that is where the current flow is going.

 

[amb]

Quote:

Originally Posted by tangent Not at all. I can hear the benefit from stacking the L&R channels. I cannot hear a benefit from stacking the ground channel. Before I tested it, I believed as you seem to. (Assuming you're not just being the devil's advocate.) EDIT: One improvement we get from stacking buffers in improved slew rate, but on the ground channel, the buffers are not slewing. The other is lower output impedance, but on the ground channel this shows up as lower crosstalk, which is very hard to hear past a certain point with most music. So, it's not surprising that improvements are hard to discern past a certain point.

I was playing the devil's advocate, and wanted to know why you don't believe that there's a benefit in stacking buffers on the ground channel. I haven't actually listened/compared a Pimeta or PPA with/without stacked buffers on the ground channel as you have, so I couldn't comment on the audible aspects. I am interested in knowing whether you have done some quantitative measurement to demonstrate no improvement with additional buffers on the GC. I know about the PPA RMAA benchmarks you have, but I presume those tests were referenced to signal ground, not output ground (even though the dummy load is to the output ground). So in a way we're not seeing the whole picture there.

As for the slew rate, did you actually measure an improvement with stacked buffers on the L and R channels?

 

[tangent]

Quote:

I presume those tests were referenced to signal ground, not output ground (even though the dummy load is to the output ground). So in a way we're not seeing the whole picture there.

True enough, but without buliding elaborate test setups (two battery powered computers, one running the test signal the other measuring, etc.) I think it will be very difficult to get useful numbers.

Quote:

As for the slew rate, did you actually measure an improvement

I didn't have to. In the datasheets, you'll usually find a slew rate vs. load graph. Since the load is shared among the buffers, slew rate must go up when you parallel buffers, all else being equal.

 

[tangent]

Quote:

Originally Posted by zxc Is there anything that I can do at this step to get rid of this oscillation?

Theoretically, there's always something you can do. The question is, what is the actual cause? You have to realize that there is something atypical about your amp: if all PIMETAs had this problem we'd be hearing a lot of screaming here on the boards about it. Since I don't know all of what you've done differently than all the others, it's hard for me to guess.

Since you've already planned on building a second amp, I say go ahead with that and see what comes of it. If it also has problems, then investigate.

 

[PeterR]

Hmm. So how much slew rate do you need? A 2Vrms sine wave at 20kHz requires 0.36V/us. Slew rate of a BUF634 is 2000V/us, good enough for 113MHz...

 

[zxc]

Quote:

Since you've already planned on building a second amp, I say go ahead with that and see what comes of it. If it also has problems, then investigate.

Thanks Tangent; and everybody else who wrote back

By the way; I just checked and no TLE left at mouser, digikey... Couldn't locate TLE at Newark, Allied... Does anybody know of a vendor that carries TLE? I know that mouser shows the chrome nose 1/4" neutrik jack (with solder lugs) out-of-stock for a month now, I hope TLE won't turn out like that...

 

[dip16amp]

Quote:

Originally Posted by zxc Thanks Tangent; and everybody else who wrote back By the way; I just checked and no TLE left at mouser, digikey... Couldn't locate TLE at Newark, Allied... Does anybody know of a vendor that carries TLE? I know that mouser shows the chrome nose 1/4" neutrik jack (with solder lugs) out-of-stock for a month now, I hope TLE won't turn out like that...

Digikey has the TLE2426CLPR (tape reel packaged) and TLE2426ILP (Industrial temperature version).

 

[dip16amp]

Quote:

Originally Posted by PeterR Hmm. So how much slew rate do you need? A 2Vrms sine wave at 20kHz requires 0.36V/us. Slew rate of a BUF634 is 2000V/us, good enough for 113MHz...

On a 32 ohm load the 2000V/uS would need 62.5 A/uS.

 

[tangent]

Quote:

Does anybody know of a vendor that carries TLE?

There's another current thread that covers this.

Quote:

A 2Vrms sine wave at 20kHz requires 0.36V/us.

Well, you can go and use uA741s then.

Seriously, I think dip16amp might have touched on it. We end up using higher-spec'd chips than you might calculate as absolutely required, because distortion still goes up due to other factors.

When you boil it all down, we stack buffers because it sounds better. Except when it doesn't.

 

[Pars]

Quote:

Originally Posted by tangent When you boil it all down, we stack buffers because it sounds better. Except when it doesn't.

That one just made my day! Thanks Tangent!