http://www.analog.com/static/imported-files/data_sheets/AD797.pdf
I'd like to use the line receiver in Figure 47.
However, I'd like it quieter than 9 nv/sqrt(Hz). Which, I think, is possible.
And, I'd like a gain of 2 in this design, which is also possible.
Questions:
1. Does anyone have experience with AD-797 in line receiver applications? If so what do I need to know?
2. At a gain of 2, what is the proper combination of de- and de-decoupling capacitors to use? The Analog circuit shows 20 pf, which is similar to the suggestion in Table 6 - HOWEVER, note that Figure 47 is claiming a noise level that's actually LESS THAN that shown in Table 6, which makes no sense to me. Can anyone explain how adding two 1k resistors will somehow get me less noise?
3. Why are the resistors 1k? I'm looking at Figure 51 which is showing me a gain of 1 with 26.1 ohms in the feedback loop! Using the same circuit as in Figure 47, can I lower the resistor values to get even less noise? With 26.1 ohms on the input, what kind of load demand would that place on my driving stage? Would 26.1 ohms work being driven by another ordinary op amp?
I'd like to use the line receiver in Figure 47.
However, I'd like it quieter than 9 nv/sqrt(Hz). Which, I think, is possible.
And, I'd like a gain of 2 in this design, which is also possible.
Questions:
1. Does anyone have experience with AD-797 in line receiver applications? If so what do I need to know?
2. At a gain of 2, what is the proper combination of de- and de-decoupling capacitors to use? The Analog circuit shows 20 pf, which is similar to the suggestion in Table 6 - HOWEVER, note that Figure 47 is claiming a noise level that's actually LESS THAN that shown in Table 6, which makes no sense to me. Can anyone explain how adding two 1k resistors will somehow get me less noise?
3. Why are the resistors 1k? I'm looking at Figure 51 which is showing me a gain of 1 with 26.1 ohms in the feedback loop! Using the same circuit as in Figure 47, can I lower the resistor values to get even less noise? With 26.1 ohms on the input, what kind of load demand would that place on my driving stage? Would 26.1 ohms work being driven by another ordinary op amp?
none of these questions can be answered without knowing your source impedance, load impedance, drive capability of previous stage, power supply quality, noise sources etc. you seem to be mixing up measurements/results in the datasheet of several applications, differing amounts of gain, inverting vs non-inverting feedback, single ended vs differential input. we cant just give you an open ended answer, especially at these noise levels.
the resistors are 1kΩ because its not such a crazy load for the amplifier. it has low noise, because it has differential input, so the 1kΩ on the input are common mode and the other resistors are wrapped in the loop, so are corrected for, for the most part. the noise in table 6 is given at minimum 10x gain, while fig 47 has only 2x gain
the circuit with 26Ω, doesnt have 26Ω input impedance, the 26Ω divider is in the inverting feedback loop, but the input goes to the non-inverting input directly, so the source doesnt have to drive the 26Ω. no you cannot just use 26Ω in a circuit like fig 47, that would produce a very difficult load and current noise would increase as well
it seems like you need to work out what you actually need more clearly, as you have just given examples for circuits that cover the entire gamut of applications. the AD797 datasheet is very thorough, but you need to know what your application is, you cant just pick numbers or values from different circuits. have you done any type of circuit analysis before? can you read a schematic? do you understand feedback mechanisms and basic passive filtering? do you have experience with designing and implementing less critical opamp circuit topologies?
to get the noise levels you are shooting for, you need to tick all the boxes and it must be a complete design that takes its source and load impedance into account, as well as a decent power supply. you cant just use the AD797 and expect to get datasheet performance by dropping it in and at low gains it needs some care in application/PCB layout.
the resistors are 1kΩ because its not such a crazy load for the amplifier. it has low noise, because it has differential input, so the 1kΩ on the input are common mode and the other resistors are wrapped in the loop, so are corrected for, for the most part. the noise in table 6 is given at minimum 10x gain, while fig 47 has only 2x gain
the circuit with 26Ω, doesnt have 26Ω input impedance, the 26Ω divider is in the inverting feedback loop, but the input goes to the non-inverting input directly, so the source doesnt have to drive the 26Ω. no you cannot just use 26Ω in a circuit like fig 47, that would produce a very difficult load and current noise would increase as well
it seems like you need to work out what you actually need more clearly, as you have just given examples for circuits that cover the entire gamut of applications. the AD797 datasheet is very thorough, but you need to know what your application is, you cant just pick numbers or values from different circuits. have you done any type of circuit analysis before? can you read a schematic? do you understand feedback mechanisms and basic passive filtering? do you have experience with designing and implementing less critical opamp circuit topologies?
to get the noise levels you are shooting for, you need to tick all the boxes and it must be a complete design that takes its source and load impedance into account, as well as a decent power supply. you cant just use the AD797 and expect to get datasheet performance by dropping it in and at low gains it needs some care in application/PCB layout.
Last edited:
Yes to all your questions. Been using the AD-797 for at least 5 or 6 years, but always at the upper end of its gain range (50-60 dB). Yes I'm thoroughly familiar with all the issues related to bypassing and such. I'm also aware it's a quirky chip, and that's why I'm asking questions.
Here's the information you're asking for: power supply +/- 15 volts. The driving stage is also an AD-797, actually two of them, in a "transformerless mic preamp" arrangement. I want balanced in, unbalanced out. The driving stages are already working, they're AD-797's at 40 dB fixed gain.
My test setup is VERY simple: I have a Digi-002, and I use the TRS input, and I turn the pad on and the gain all the way up. Then I listen. If I hear hiss, it's unacceptable. Right now, I'm hearing just a LITTLE bit of hiss, but it's too much. I want about HALF of what I'm hearing (I'm being reasonable, right? I don't need it to "go away entirely", I just need less of it).
And I do have one other question, which is probably also very simple - I'm contemplating using a BUF634 to enhance the drive capability of THIS stage, and my question is should the 20 pf cap remain in parallel with the feedback resistor or should it go directly across the op amp?
I don't know what your system really looks like but worrying about even half decent op amp noise in the stage after a 100x Av AD797 is pretty pointless
the gain of 100 amplifes the AD797 noise too, the 0.9 nV/sqrt(Hz) input V_noise becomes 90 at the output
and if the noise is mostly white as it is at audio midband and above then it adds "vectorially"
so a "noisy" fet input op amp next stage at 15 nV/sqrt(Hz) changes the 90 to 91 nV/sqrt(Hz) - way below the spread in AD797 "typical" noise
the gain of 100 amplifes the AD797 noise too, the 0.9 nV/sqrt(Hz) input V_noise becomes 90 at the output
and if the noise is mostly white as it is at audio midband and above then it adds "vectorially"
so a "noisy" fet input op amp next stage at 15 nV/sqrt(Hz) changes the 90 to 91 nV/sqrt(Hz) - way below the spread in AD797 "typical" noise
Last edited:
I'm well familiar with the theory. I'm asking a different question, it's a question "in practice". What I'm telling you is I can hear the hiss with an OPA, and not with an AD. It is MORE than noticeable, it is DRAMATIC. Not a 1% increase, more like a 1000% increase. So, I'll just stick with my plan, and use the AD in the second stage as well.
So, how about my specific questions? One, can I use lower value resistors, and two, if I use a buffer should the cap remain across the feedback resistor or should it go directly across the op amp?
its hard to help without the schematics, gain settings...
but I really don't understand how there could be audible difference after amplifying the AD797 input V_noise by 100x
the numbers are literally orders of magnitude off from your description of what you hear
in such situations I really go over everything verify circuit, source, operating conditions - noise vs frequency plots, histograms
I would be happy to bet pictures of presidents all day long that flat frequency response amplifiers, properly working, no visible/measurable PS/EMI/oscillations/clipping problems with 2x Av OPA627 added noise to 100x Av AD797 can't be heard with the noise kept to < 30 dB SPL in AB/X DBT, 95% confidence level
but I really don't understand how there could be audible difference after amplifying the AD797 input V_noise by 100x
the numbers are literally orders of magnitude off from your description of what you hear
in such situations I really go over everything verify circuit, source, operating conditions - noise vs frequency plots, histograms
I would be happy to bet pictures of presidents all day long that flat frequency response amplifiers, properly working, no visible/measurable PS/EMI/oscillations/clipping problems with 2x Av OPA627 added noise to 100x Av AD797 can't be heard with the noise kept to < 30 dB SPL in AB/X DBT, 95% confidence level
Last edited:
'Kay, well, I did describe the test setup, right? There is LOTS of gain in this setup, so for instance, when I have a microphone plugged into this thing, and I crank the gain on the test setup, I'm actually HEARING the resistor nosie, there's that much gain. But more importantly, I'm hearing the self-noise in the second stage, which is fixed gain, and it's operating at a LOW gain where noise is much more significant. With this much gain in the test setup, if I blow air out of my mouth 45 degrees AWAY from the microphone I get a signal that's so painfully loud I practically have to yank the headphones off. I'm trying to listen to a loud kind of waterfall-sounding thing called "total noise", and part of that changes when you put your hand over the mic, and part of it doesn't.
So, when I crank the headphones and listen "without" the microphone, I can move the output from the first stage to the second stage, and compare. With the first stage only, I can actually hear the wind. If I point the mic in a particular direction and leave it there, and there happens to be a gust of wind over in the neighbor's clump of trees, the mic pics it up and I can hear the details of it. With the second stage in place, the details of that wind sound are hidden by the extra noise. It's not "detailed" anymore, it sounds muddy and it's beginning to be indistinguishable from the background noise.
The AD is a very nice chip when it's used properly, I'm convinced the designers were on the right track with this thing. It's not 32-bit perfect, noise-wise, but it's pretty darn good. You can hear a lot of stuff with this chip that you can't really hear the details of with others. There's maybe a little spectral tweak that could occur here and there, but basically it sounds pretty good, lot of details in the upper mids (and conveniently right in the sibilance range). So, what I first wanted to try was pairing it with a "different" sounding chip, like maybe an OPA, something a little warmer with maybe more of a fatter bottom - or some combination that might even out the "AD sound". Two AD's in a row is definitely the "AD sound", it's hard to get rid of or change after that. But one AD followed by something else is workable, at least so far that's what I've been hearing. So I'm trying to find low noise chips with different sounds, so far I've tried the usual culprits like LM4562 and so on (two of those in a row is also... close to becoming unusable in a real tracking context). Anyway, I digress.
The thing is, self-noise at a gain of 2 can be significant. Depends on the chip. You can hear the waterfall rushing at "less than full gain" on the test setup, which is not a good thing. The EIN never tells the whole story, right? I always want to listen with my own ears. That OPA627 has other problems too, I've been discovering that it can't handle anything below a few hundred ohms in the gain leg without the DC output going wacky. It's nice when it's used properly, yes? I have a little DI box with an OPA627, I wouldn't trade it for anything in the world. (But that's a 150k input impedance
)Yeah, wish I had that stuff. Nah, I'm flying blind here, all I have is an M-Box.
BUT - yes, I can hear some VERY interesting stuff with those AD-797's. First of all, at low gain, without any decoupling, I can hear the servos. They're obvious you can hear them very clearly. There's also a sound that resembles a VCO driven by an LFO, kind of like wooeeeoooeeooo like that. And of course if you leave off the bypass circuitry you can hear all kinds of noises - and, they do change with and without a buffer.
So, I've gotten rid of all the switching noise with proper bypassing, and I've got the basic stage-1 preamp noise down to where it's "insignificant" relative to a gust of wind in my neighbors' trees. Now I hear the VCO/LFO sound pretty clearly, and it seems my problem there is I need to move the decoupling cap or something. I found a Walt Jung thing about AD's and buffers, so I'll have to read through that.
Yeah, theory isn't always the same as practice. For instance, I've found an OPA602 is actually much quieter than an OPA627 in this circuit. Not sure why, I have no idea, but the character of the noise is also audibly different.
Thank you. Yes, there is 100 ohm resistor directly on the output.
Here is why I'm asking these questions: I've seen schematics using an AD-797 and a BUF-634. In some cases, people keep the decoupling cap across the chip, and in other cases they move it to the output of the buffer. Not sure which one of these is "correct". The 47pf distortion-reducing cap should go on the output of the buffer, that is my understanding and I've verified that it sounds better and "correct" that way. However moving around the decoupling cap produces some very strange results.
Anyway, Figure 42 is as good an example as any. How should this look when a BUF-634 is added? I'm guessing the 33r resistor goes away, no?
Sorry, but that's the whole point. Didn't you read my description of the test setup? Yes, you can in fact HEAR noise. That's kinda why we people try so hard to get rid of it!
Actually the noise spec for the 49990 is just about identical to the 797. The biggest advantage the 49990 has is a lot less low-frequency drift (we're talking in the range of 0.1 to 0.01 Hz, like that). Some of the 797's are very drifty, is a production issue though, it varies according to lot number.
Had similar issues in my Zishan DSD, AK4497 based DAP, exactly same kind of noises and sometimes some radio programming too.
Tried just about everything including every suggestion from datasheet, but ultimate solution was to add RC filter on AD797 output, 100Ohm resistor in series + 330pF to ground.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.