Using the AD844 as an I/V

I don't use the SK120 ccs circuit for dac dc offset nulling. I use the one Abraxilito gave me below. I don't know what proplems it creates but it works fine

As I see it , the ccs (sk170) serves to bias the ad844 in class A, because the AD844 have only a 6.5 ma quiescent current (the entire chip consumes only 6.5ma), this soluction will shut down half of the ad844 input stage and creates a D.C. offset at the output so you will need a output capacitor , another limitation is also the amount of current the ccs can source to the chip.

A better solution is to source current in input pin (2) and source the same amount of current in TZ pin (5), in this way we can bias the AD844 in Class A without D.C. offset at the output, and now the power dissipation of the AD844 is the only limitation for the amount of current we can source to the chip , I think 10ma is a good choice. In this basic circuit R5 serves to trim the output d.c. offset. You can use other types of current sources. This is only for demonstration.

The way you use it (without ccs). will also work because the DAC chip you chose has low output current and dont have output current offset. Is up to you , chose the solution that please you more. :)

The advantage of using a all discrete soluction is that we can better tailor a soluction for our needs.
 

Attachments

  • AD844.png
    AD844.png
    25.2 KB · Views: 2,295
C1 would get rid of the trash nicely if the GND were perfectly quiet. But what normally happens when dumping HF on to the GND - the GND becomes rather noisy because all GND connections have inductance. Putting a small resistance in series with the C can help keep the GND clean by soaking up the HF rather than merely dumping it. Something to try by listening.
 
abraxalito , what you describe is a snubber, and yes you right, is a good way of dissipate that high frequency energy. But i think that using like 100pf in parallel with the snubber you discribe is a good compromise.

A ground plane has almost no parasitic inductance, but is easier to make when one use surface mount devices.
 
Yes, no simple solutions to these issues - adding 100pF will move the LC resonance to a frequency much higher, perhaps there's less energy there for the LC circuit to ring from. Yes a snubber it is. I also use snubbers for decoupling, rather than pure caps - although its quite fashionable to aim for the lowest possible ESR, usually some series resistance helps to provide damping. Getting decent sound requires attention to such small details :)
 
Groundplanes in my understanding have resistive impedances rather than inductance, assuming you use them in pairs but usually that impedance is fairly high compared to the impedance we want from our power supply.

I keep a handy supply of 0805 SMT resistors from 1R to 4R7 to introduce as the R in snubbers as I almost always use ceramics (rather than electrolytics) for decoupling. Ceramics have scary-low ESRs, in contrast to small 'lytics.
 
abraxalito , what you describe is a snubber, and yes you right, is a good way of dissipate that high frequency energy. But i think that using like 100pf in parallel with the snubber you discribe is a good compromise.

A ground plane has almost no parasitic inductance, but is easier to make when one use surface mount devices.

I would think that the inductance added by the resistor would reduce the effectiveness of the bypass. Unless someone has actually investigated this in real life and produced some kind of evidence and data to show that there is a benefit, I'm not going to use a resistor between the bypass cap and ground. None of the data sheets or other sources I have read about bypassing have suggested this, but of course I am not extremely well read either.

Ground planes are no big deal for through hole devices in my experience. I use two, one on each side of the board.

One good design principle I learned from another source is to use a very wide trace between the IC power pin and the bypass cap.
 
on AD797's datasheet, there is a description such decoupling scheme.

Yes, but notice that the very small value carbon resistor is used to minimize the lead inductance of a large tantalum capacitor, not the smaller 0.1uF bypass capacitor. This technique appears to only be useful when using relatively large bypass capacitors and only if you use a carbon resistor, or some other type that has very low inductance at radio frequencies.
 
If the inductance contributed by the resistor concerns you then select a smaller package size - 0603 perhaps or 0402 if your eyesight is good enough. Its the total loop inductance that's important, not just that of the resistor - so if the resistor adds to the loop area it will increase the inductance. OTOH if the resistor's just replacing a wire that would be there anyway, it won't add inductance. This is assuming normal thick film SMT resistors of course, not wirewound ones.
 
If the inductance contributed by the resistor concerns you then select a smaller package size - 0603 perhaps or 0402 if your eyesight is good enough. Its the total loop inductance that's important, not just that of the resistor - so if the resistor adds to the loop area it will increase the inductance. OTOH if the resistor's just replacing a wire that would be there anyway, it won't add inductance. This is assuming normal thick film SMT resistors of course, not wirewound ones.

I think you misunderstand. This does not concern me at all. I won't be using this technique because I don't think there is enough benefit to bother. TI does not suggest the technique for their extremely high bandwidth BUF634 for example. Maybe they don't know as much as AD, I don't know.

I assumed that the datasheet portion under discussion applied to through hole devices (since the AD797 is available that way) which have more lead inductance than SMD's. The datasheet specifies a carbon resistor.
 
I think you misunderstand. This does not concern me at all. I won't be using this technique because I don't think there is enough benefit to bother.

That's totally fine by me - your reasoning though wasn't particularly sound because in many cases the loop area isn't increased by adding the resistor, so the 'extra inductance' argument doesn't hold water. That's what concerned me in giving my response. If the part being decoupled is in a DIL8 package for example, both the cap and resistor comfortably fit beneath the chip without adding inductance. That's how I decouple TDA1545 DACs when in DIL packages.

The use of snubbers is less important for decoupling purely analog circuits that aren't handling fast rise time signals and hence aren't taking big, fast gulps of current like CMOS digital chips do. I do recommend them for such applications, on pure analog handling audio band signals its a toss-up whether there's any benefit.
 
That's totally fine by me - your reasoning though wasn't particularly sound because in many cases the loop area isn't increased by adding the resistor, so the 'extra inductance' argument doesn't hold water. That's what concerned me in giving my response. If the part being decoupled is in a DIL8 package for example, both the cap and resistor comfortably fit beneath the chip without adding inductance. That's how I decouple TDA1545 DACs when in DIL packages.

The use of snubbers is less important for decoupling purely analog circuits that aren't handling fast rise time signals and hence aren't taking big, fast gulps of current like CMOS digital chips do. I do recommend them for such applications, on pure analog handling audio band signals its a toss-up whether there's any benefit.

For THD's, the leads of the added resistor have inductance at RF, so I can't see how they would be of much benefit in that particular case. Of course, SMD's are a different story. I prefer THD's because my hands are not nearly as steady as they used to be.

Obviously, digital circuits have different power supply requirements than analog ones, and it's debatable whether or not an opamp in an IV converter for a DAC sees enough digital surges to warrant this kind of snubber circuit. I would like to see some hard data on that before bothering with it. Do you have any?
 
For THD's, the leads of the added resistor have inductance at RF, so I can't see how they would be of much benefit in that particular case. Of course, SMD's are a different story. I prefer THD's because my hands are not nearly as steady as they used to be.

Fair enough, I only use SMT passive parts these days (other than where I need major power dissipation in resistors or large-valued inductors) and was only speaking of SMT. If the resistor is through-hole then yes its going to add inductance so I can see why you'd not be interested.

Obviously, digital circuits have different power supply requirements than analog ones, and it's debatable whether or not an opamp in an IV converter for a DAC sees enough digital surges to warrant this kind of snubber circuit. I would like to see some hard data on that before bothering with it. Do you have any?

What kind of 'hard data' did you have in mind? I'm not interested in making measurements or scope shots of any differences from adding series resistors if that's your wish. After all it would be totally irrelevant data to you as you're not into SMT.
 
What kind of 'hard data' did you have in mind? I'm not interested in making measurements or scope shots of any differences from adding series resistors if that's your wish. After all it would be totally irrelevant data to you as you're not into SMT.

I may be forced to use SMD's in the near future, so I have some interest in this idea. The LME49990 is particularly difficult to implement, according to their data sheet (and comments by others), for example, and this technique may be useful for that part. I'd be interested if you know of someone who has made this comparison if you don't have it. I assume that Analog Devices didn't invent this idea out of thin air but instead came up with it to address a specific problem they were having with the AD797. I understand the theory of course, but would like to see data on how much of a difference it makes. I can try to do a Google search myself but I am not hopeful about finding something useful.
 
Power supply decoupling would be the least of your worries if you're considering either of those opamps in close proximity to anything digital as I understand they don't co-exist very happily with RF. Walt Jung (who works for ADI) even designed out the AD797 in his regulator as it misbehaved when powering digital stuff.

<edit> Here's a paper which anyone considering an opamp with input LTP for I/V would do well to read and digest : http://www.uemc.polito.it/papers/opampsusc_01.pdf
 
Power supply decoupling would be the least of your worries if you're considering either of those opamps in close proximity to anything digital as I understand they don't co-exist very happily with RF. Walt Jung (who works for ADI) even designed out the AD797 in his regulator as it misbehaved when powering digital stuff.

<edit> Here's a paper which anyone considering an opamp with input LTP for I/V would do well to read and digest : http://www.uemc.polito.it/papers/opampsusc_01.pdf

Thanks for the paper.

No, I'm not considering either opamp for an IV converter for a DAC. The IV converter I have is based on the LME49710.