[jboehle]

I want to build a simple CMOY-like headphone amp with the AD8397 chip. But I want it to be optimized for my Grados. I want to make the gain=2. So in the following schematic, I need to make R3 = R4, because gain = (1 + R3/R4). But how do I choose the actual values for R3, R4, and the rest of the parts?

I was going to take Tangent's new PINT schematic (http://tangentsoft.net/audio/pint/misc/sch-0.93.pdf) and just make the L/R channel portion of the amp, but the gain is not equal to 2. So why not just make R3 = R4? Because in the thread on the PINT, Tangent says:

Quote:

Originally Posted by tangent With bipolar-input op-amps, the C1, C3, R2, R3, and R4 values all interact significantly. Change one, and they all have to change to keep all else substantially equal. If R4 is cut in half, C3, R3, and R2 are also cut in half, and C1 must double. [...] An entirely different way to go is to short C1, make R2 high so its interaction is all but irrelevant, and try to balance the feedback R's against the pot's varying resistance. This can work with some op-amps, and obviously it gets the feedback Rs very low. But it requires very careful calculation to avoid getting the DC offsets too high. It might require adding a series resistance between the pot wiper and the op-amp input to make it practical. I'll have to do some calculations and get back to you on that.

Ignore the mention of C3, as I think Tangent determined a compensation cap across R4 was not necessary.

So, how do I know how C1, R2, R3, and R4 interact for bi-polar amps? Where should I start reading to learn how to determine the correct values for those, for a decently long-life battery powered amp, but avoid amplifying noise? I thought C1 was just to block DC from your source. Is that just for FET-input op-amps, and C1 plays a different purpose for bipolar op-amps?

I've read part of and skimmed over the rest of "Op Amps For Everyone", but it doesn't really cover the particular questions I'm asking (or I skimmed over it).

TIA.

 

[n_maher]

Quote:

Originally Posted by jboehle So, how do I know how C1, R2, R3, and R4 interact for bi-polar amps? Where should I start reading to learn how to determine the correct values for those, for a decently long-life battery powered amp, but avoid amplifying noise? I thought C1 was just to block DC from your source. Is that just for FET-input op-amps, and C1 plays a different purpose for bipolar op-amps?

I'm not sure there's a source to read this. Some things still just have to be tested, tested and then retested. I'd bet Tangent has spent a large part of his time designing the PINT testing different configurations. I believe that with cranky opamps like the 8397 that the most important information is probably gathered and not read.

So I guess the real answer is breadboard your circuit and test it, although some have reported difficulty doing this with the 8397, it's just that much of a bear to work with. But from the sounds of things all of the effort might be worth it

.

Nate

 

[MisterX]

http://tangentsoft.net/audio/hs-opamp.html

Read the part about balancing input and output offset currents and go from there.

 

[JahJahBinks]

R2 should be at least 10 times the Pot resistance.
For the feeback resistors R3 and R4, I would choose something between 10k to 50k. R3=R4 for a gain of 2.

Some of the information you can find in application notes.

 

[jboehle]

I read Tangent's guide on high-speed op-amps. His formula says that:

R2 = (R3*R4)/(R3+R4), which with a gain of 2, R2=(R3^2)/(R3*2)

But if R2=10*POTVALUE, the minimum R2 value I should use would be 100K (assuming my POT is 10K). But with R3=R4=50K, the above formula says R2=25K, which is way less than 10*POTVALUE.

With R3=R4=10K, R2 would be 5K, which is even further away from 10*POTVALUE.

 

[ericj]

So if R3 and R4 = 200k then R2 can = 100k and you can use a 10k pot.

If you feel bad about that, go get one of those extremely nice (and big) 10k Noble pots for $25 or so.

(Actually, I think I know where to get them for next to nothing. I'll let you know on monday if it turns out I'm right.)

Come on, at gain 2 it's not like you have much use for a volume knob in the first place.

Why 2, by the way? Most "for grados" designs I've seen are gain = 3db for exactly twice the audio output as the input.

Oh, and the input capacitor is needed on all opamp types but only if your source has a DC offset. If you know your source does not have any DC offset, you can omit it, but then you'll need to check for DC offset before connecting your headphones with each new source.

The problem is that your amp will amplify the DC offset, possibly increasing an otherwise harmless offset into something that will destroy your headphones.

The M3 for example doesn't have input couplers, and the instructions do carry a warning about that. You could add them, but you would have to use an off-board volume pot, and the the value of the capacitors would have to be pretty high unless you changed the values of the resistors around the opamps, and that would affect how the bass boost works, and may affect some of the other features of the board as well, and i didn't feel it was worth the hassle, myself.

 

[jboehle]

I don't mind using a 10K pot. My comment on the pot wasn't so much that I didn't want to use a 10K pot, it was that Tangent's formula for R2 didn't line up with JahJahBinks reply.

For C1, that's what I thought, that it was only needed to block DC from your input. I just didn't know if it was needed for more than just blocking DC. Tangent's guide says C1 & R2 form a high-pass filter. Is that needed in the circuit for some reason (other than to block DC)?

What gain value = 3db? I wanted to use gain=2 for the AD8397 because it doesn't behave well at unity gain, but the AD8397 has good current output, which is good for Grados. But the Grados don't need much gain, just lotsa current.

 

[MisterX]

Do you still have your RA1 clone?

That is about as subtle of a hint as hitting you with a brick.

 

[jboehle]

I could never get that damn thing to be hiss-free.

 

[ericj]

Quote:

Originally Posted by jboehle I don't mind using a 10K pot. My comment on the pot wasn't so much that I didn't want to use a 10K pot, it was that Tangent's formula for R2 didn't line up with JahJahBinks reply. For C1, that's what I thought, that it was only needed to block DC from your input. I just didn't know if it was needed for more than just blocking DC. Tangent's guide says C1 & R2 form a high-pass filter. Is that needed in the circuit for some reason (other than to block DC)?

It Coincidentally forms a high-pass filter with R2. That's what you get when you pass a signal through a capacitor followed by a resistor to ground.

You can visually look at what the values do here:

http://www.st-andrews.ac.uk/~www_pa/...hpass/hpf.html

It's pretty easy to see that .1uf and 100k is less than ideal, but you only need .22 or so before things start looking better.

But with higher values of resistors in the feedback loop there is much more potential for hiss. You want to keep them as far below 1 megohm as you can while still keeping things stable. Low values are good, but with bipolar amps and an input coupler cap you really can't go very low, without huge input caps. This has to be some of the reason for the 5uf solens in the RA1.

Quote:

Originally Posted by jboehle What gain value = 3db? I wanted to use gain=2 for the AD8397 because it doesn't behave well at unity gain, but the AD8397 has good current output, which is good for Grados. But the Grados don't need much gain, just lotsa current.

Gain = decibels. Every 3 decibels is 2x signal. So gain 1 = nothing, unity, . gain 2 = 1.5x signal, gain 3 = 2x signal, gain 6 = 4x signal, etc.

The thing about the input cap interacting with R2 is that if the values are wrong, you attenuate audible bass.

Not sure how these interact in a meaningful way with the compensation cap across R4. Small numbers (low two digits) of picofarads here will roll off rf noise to gain 1, which can stabalize high speed opamps.

There are probably ideal values, but it's probably one of those situations where the audible difference between calculating the 'ideal' mathemetically or just using the SWAG* method is arguably nothing.



*(S.W.A.G. = Scientific Wild-Ass Guess. You'd be shocked how many business cases are developed using this method.)

 

[tangent]

Quote:

Originally Posted by MisterX http://tangentsoft.net/audio/hs-opamp.html

I just finished rewriting that article. It's been on the to-do list for a long time, and it came to a head with this PINT work. Your question was the straw that broke the camel's back.

The article is now 3 times longer than before, and covers several all new topics. It covers the pot issues, which I completely ignored before, and it's the first time on my web site that I've tried to explain oscillation. I tried to avoid writing a textbook, but it's still a bit of a brain dump. In the first week or so after a big effort like this, I keep a more open mind for changes, so do let me know if there are parts that are unclear or otherwise problematic.

Soon to come: an addition to my audio calculator page that will calculate DC offset due to input offset current imbalance in simple CMoy-like configurations, with and without input caps, with and without pots.

Quote:

Every 3 decibels is 2x signal.

Only if you're talking about power. Here, we're talking about voltage, so 2x gain is 6 dB. Voltage dB = 20 * log(V1/V2). Power dB = 10 * log(P1/P2).

 

[bigmike216]

Quote:

Originally Posted by tangent Soon to come: an addition to my audio calculator page that will calculate DC offset due to input offset current imbalance in simple CMoy-like configurations, with and without input caps, with and without pots.

That will be a very helpful resource, I'm looking forward to that!

 

[tangent]

The first version of the calculator is done. It currently only considers offset due to input offset currents. Someday I may add things like input offset voltage, source-referenced offset, etc.

 

[MASantos]

I know this is off toppic, btu what it a PINT amplifier? Is it between cmoy and mint?

 

[MisterX]

Pint thread---------> http://headwize.com/ubb/showpage.php?fnum=3&tid=5806

(the PINT is a successor to the MINT amp)