[Andrea]

Hi,

I've been pondering using an AD45048 at unity gain as an output buffer (directly connecting the outputs to the inverting inputs, without resistors).


My question is -- unlike the AD8397, can its AD45048 variant be succesfully used at unity gain, and put in the feedback loop of the controlling op-amp?

And if yes, will it likely be stable if driven by a AD823 or OPA2107 or OPA2134 ?



Thanks to anyone who should be able and willing to answer

 

[jcx]

I believe a analog devices' engineer claimed they are identical over on Headwize

in unity gain mode they don't swing rail-to-rail due to the input common mode range limit - a problem with any buffer chip

if you are able to deal with the extra variables Jung's example multiloop amplifiers use gain in the output stage, this would let the ad45048/8397 swing to the rails without exceeding its input cm range

I would start with a local loop gain around the ad45048/8397 of ~ sqrt(Av) where Av is the overall gain of the composite amp

they are moderately sensitive to Cload on the output, smallish series R >~ 10 Ohms, R||L or lossy ferrite bead can help isolate the cable Cload from the amp output - please don't follow tangents suggestion of placing the series impedance inside the feedback loop

 

[Andrea]

Well, it seems more complex than I had wished (a drop in replacement in an amp I have, that's using a unity gain op-amp for output stage).

I guess I'll go the already proven to work BUF634 on BrownDog way, then.


Thanks

 

[ericj]

Yeah. The guy who designed it says these two chips are one and the same, cut from the same wafer, and marketed for different uses.

Some people have hypothesized that the 45048 is perhaps sorted at a lower quality threshhold than the 8397 - which would probably be something like offset current differences between channels - but we don't know that for sure.

 

[tangent]

Look at the G=+1 curve in the graph in the upper right of page 5 in the AD45048 datasheet. It indicates the same problem as the one in the upper left of page 9 in the AD8397 datasheet does: this chip is peaky at unity gain at high frequencies, leading it to more easily become unstable.

This can be fixed by using either a resistor or suitable inductor in series with the output, but it can be a trial-and-error process. It's not something I would recommend to someone looking for drop-in replacements.

 

[Andrea]

Quote:

Originally Posted by tangent Look at the G=+1 curve in the graph in the upper right of page 5 in the AD45048 datasheet. It indicates the same problem as the one in the upper left of page 9 in the AD8397 datasheet does: this chip is peaky at unity gain at high frequencies, leading it to more easily become unstable. This can be fixed by using either a resistor or suitable inductor in series with the output, but it can be a trial-and-error process. It's not something I would recommend to someone looking for drop-in replacements.

Thanks, tangent, that was enlighting.


And what if the AD45048 (at unity gain) is put inside the feedback loop of a low bandwidth op-amp like the OPA2107, with an overall gain of 5 ?

Will that peaky high frequency behavior matter anymore? And should a, say, 33 ohm output resistor suffice to make the whole thing perfectly stable, then?


Thanks

 

[Andrea]

Lastly, if the above was to be true, could I use an AD8397 instead with equal (or better?) results?

 

[amb]

Heed tangent's advice. The AD45048 and AD8397 are cranky chips that should not be used as opamp-rolling candidates into an existing design, regardless of topology. The circuit and layout must be carefully tailored to these chips, or you'd be blowing them up (they have no output current limit protection, and self-destruct easily when they oscillate) or at best have a marginally-unstable amp with severe DC offset. If you plan to design your own amp around these chips, you better have the necessary know-how and test equipment...

 

[Andrea]

Quote:

Originally Posted by amb Head tangent's advice. The AD45048 and AD8397 are cranky chips that should not be used as opamp-rolling candidates into an existing design, regardless of topology. The circuit and layout must be carefully tailored to these chips, or you'd be blowing them up (they have no output current limit protection, and self-destruct easily when they oscillate) or at best have a marginally-unstable amp with severe DC offset. If you plan to design your own amp around these chips, you better have the necessary know-how and test equipment...

I see, thanks. I'll drop the idea.

So back to the former plan of using either 2 BUF634 or 2 OPA551 on a dual DIP browndog.

There's another doubt I have, though, that is, is the OPA551 too slow (15 V/us, 3 MHz) to buffer an OPA2107 or an OPA2227 (at G = 5) ?

 

[amb]

Quote:

Originally Posted by Andrea So back to the former plan of using either 2 BUF634 or 2 OPA551 on a dual DIP browndog. There's another doubt I have, though, that is, is the OPA551 too slow (15 V/us, 3 MHz) to buffer an OPA2107 or an OPA2227 (at G = 5) ?

There is little point in using a OPA551 as a buffer for any other opamp because it is already itself an opamp (unlike the BUF634 which is only an open loop buffer). If you're going to use the OPA551, just build a cmoy-like amp with it and be done.

 

[Andrea]

Quote:

Originally Posted by amb There is little point in using a OPA551 as a buffer for any other opamp because it is already itself an opamp (unlike the BUF634 which is only an open loop buffer). If you're going to use the OPA551, just build a cmoy-like amp with it and be done.

Yes I understand that. I was just attracted by the OPA551's low specified distortion.


For last (really!), I'm a little bit concerned about placing two BUF634's on a dual DIP BrownDog (together with 220 ohm bandwidth R's). Yes it has already proven to work fine, but I don't have a way to check for potential oscillations.

The amp's pcb is double layer, well designed and very compact, and the power supply caps (470uF Panasonic FM's and .39uF Rubycon film) are close to the active devices. Both op-amps (gain and 'buffer' stage) are socketed.


Thanks to anyone who answers

 

[tangent]

Quote:

Originally Posted by Andrea And what if the AD45048 (at unity gain) is put inside the feedback loop of a low bandwidth op-amp like the OPA2107, with an overall gain of 5 ?

That wouldn't help. The 2107 would be operating at G=5, but the 45048 would still be at G=1.

Now, if you were feeling perverted, you could put the 2107 in the 45048's feedback loop, and probably have a much stabler combination, even if you gave it an overall gain of 1. Why? Because then the 45048 would be isolated from the load capacitance. But then you would be taken out and shot by the secret audiophile police for abusing fine silicon.

Quote:

And should a, say, 33 ohm output resistor suffice to make the whole thing perfectly stable, then?

Quite likely, but then you're giving up one of the 45048's most attractive attributes, its low output impedance.

 

[Andrea]

Thanks again. I'm not perverted enough to give it a try

so,


How about the other way, do I run a risk of oscillation with buffers on a browndog? Maybe it's pure paranoia, but I don't have enough practical experience to tell.