Hi!
I just have a couple of (very) basic questions regarding opamp I/V.
These are about overcompensation and impedance matching.
- Say i'm going to use an AD797: it features an "overcomp" pin and "high freq/gain distortion reduction".
We do not use these frequencies nor big gain. BUT it also improves bandwith, and may make it less sensitive to our conversion clock.
Unofrtunatly, I guess it's not made without any compromise. Which are they?
Anyway, this point isn't the worst as I still can have the connections for components on the board, and not solder them if they don't give any advantage.
- The other point is impedance matching. Zout of the DAC, the R in the feedback loop etc... Well I should calculate what would be the equivalent R, but I guess something around 150 ohms (but maybe I'm wrong).
Opamps perform better when they have matched input impedance. BUT total impedance then arises, and then noise arises. Not that much: it's a quite low impedance, and from the chart, I should gain more than I loose. But it's not very obvious, and wouldn't be a solid connection to ground be better?
Then, from a layout point of view, there's a dilemn to solve before deciding: should I put a solid connection to ground or the place for a resistor? I know I could use 0ohms but they are far from perfect and likely would disturb the signal. So, do you know if I should match the input impedances or not?
Thank you very much
P.S: Bonus question: what's your favorite part for doing the unbalancing stuff after it? Thanks
Nicolas
I just have a couple of (very) basic questions regarding opamp I/V.
These are about overcompensation and impedance matching.
- Say i'm going to use an AD797: it features an "overcomp" pin and "high freq/gain distortion reduction".
We do not use these frequencies nor big gain. BUT it also improves bandwith, and may make it less sensitive to our conversion clock.
Unofrtunatly, I guess it's not made without any compromise. Which are they?
Anyway, this point isn't the worst as I still can have the connections for components on the board, and not solder them if they don't give any advantage.
- The other point is impedance matching. Zout of the DAC, the R in the feedback loop etc... Well I should calculate what would be the equivalent R, but I guess something around 150 ohms (but maybe I'm wrong).
Opamps perform better when they have matched input impedance. BUT total impedance then arises, and then noise arises. Not that much: it's a quite low impedance, and from the chart, I should gain more than I loose. But it's not very obvious, and wouldn't be a solid connection to ground be better?
Then, from a layout point of view, there's a dilemn to solve before deciding: should I put a solid connection to ground or the place for a resistor? I know I could use 0ohms but they are far from perfect and likely would disturb the signal. So, do you know if I should match the input impedances or not?
Thank you very much
P.S: Bonus question: what's your favorite part for doing the unbalancing stuff after it? Thanks
Nicolas
Basic questions ---- Not Take it you have the Analogue Devices data sheet for the AD797 which actually shows a DAC implementation.
I suspect a lot depends on the DAC too, to be honest I think you need very sophisticated measurements and test gear to come to a correct final design.
Interesting though.
NeoY2k said:
I will not attempt to answer the questions about feedforward (distortion reduction) in the AD797, as I have never used the chip myself.
You need to remember that I/V only used for current output DACs(obviously) and these expect to see a very low impedance load. This is provided by the "virtual earth" at the op-amp feedback point, so no impedance matching is expected or needed. The only degree of design freedom is what value of feedback resistor to use, which is matter of setting the correct output voltage, so you get 2Vrms out for a full scale sine wave.
To determine the correct feedback resistor, find the maximum (peak) signal current, and calculate the resistor that this would develop 2.828V across. This will give the correct Red Book level.
Definitely do try the pin8 connection even in I/V or other gain-of-1 applications; the reduction in output impedance / decompensation really helps IME.
Try 47pF COG caps, and connect to the audio output node - i.e. at the point where the I/V stage feeds whatever follows, not close to the AD797 itself necessarily. If it looks like adding a small cap here might provoke instability I've had good results by adding a small (10-50ohm, SMT part) resistor between the pick-off point and the cap itself to add a touch of damping/decoupling.
Try 47pF COG caps, and connect to the audio output node - i.e. at the point where the I/V stage feeds whatever follows, not close to the AD797 itself necessarily. If it looks like adding a small cap here might provoke instability I've had good results by adding a small (10-50ohm, SMT part) resistor between the pick-off point and the cap itself to add a touch of damping/decoupling.
Thank you all!
Mooly: I have my Tektronix DSO (a bit crappy, well, as every TDS2100 is) but also acced to an Audioprecision (portable one, and on a request the big expensive one! xD ) so I'll be able to measure all this. Lucky me
Martin Clark: Ok so I'll try the pin 8 stuff. Why should I connect it further not near the chip? To get some natural damping from the impedance of PCB traces?
What's your point of view on the impedance matching point?
Pigletsdad: Thank you, I knew that before but never had the chance to see someone explaining it that clearly.
I know that, from the dac's output point of view, we should see zero (or the lowest possible anyway). That's fine.
But from the - pin of the opamp, I do not consider this in- pin as a virtual ground. I look at the impedance around this, and it is the feedback loop R (say 3k) in parallel with the DAC's output impedance that is in series with the equivalent current source(that is, well, I don't really know).
So should I find this value to have the same on the noninverting input (between ground and in+)? I read on a datasheet that it improved DC performance (but I don't care about DC performance). On the other hand, as it adds impedance at the inputs of the opamp, noise rises. But having matched impedances reduces noise too. So from the noise point of view, it has to be chosen from graphs on the datasheet. If these impedances are very low, then we benefit (from a noise point of view) of matching impedances. Otherwise, me loose some performance.
But appart from noise issues that can be determined quite simply from the graph, does it improves/ changes anything? If it improves DC performance, chances are it will improve low frequency performance.
But it may have some drawbacks too.
And I fear using a jumper to keep this option on the printed board to choose between matched/direct ground, because a good direct ground is a close via or direct strip to ground
I hope I was clear, I'm a bit tired :/
Thanks,
Nicolas
Mooly: I have my Tektronix DSO (a bit crappy, well, as every TDS2100 is) but also acced to an Audioprecision (portable one, and on a request the big expensive one! xD ) so I'll be able to measure all this. Lucky me
Martin Clark: Ok so I'll try the pin 8 stuff. Why should I connect it further not near the chip? To get some natural damping from the impedance of PCB traces?
What's your point of view on the impedance matching point?
Pigletsdad: Thank you, I knew that before but never had the chance to see someone explaining it that clearly.
I know that, from the dac's output point of view, we should see zero (or the lowest possible anyway). That's fine.
But from the - pin of the opamp, I do not consider this in- pin as a virtual ground. I look at the impedance around this, and it is the feedback loop R (say 3k) in parallel with the DAC's output impedance that is in series with the equivalent current source(that is, well, I don't really know).
So should I find this value to have the same on the noninverting input (between ground and in+)? I read on a datasheet that it improved DC performance (but I don't care about DC performance). On the other hand, as it adds impedance at the inputs of the opamp, noise rises. But having matched impedances reduces noise too. So from the noise point of view, it has to be chosen from graphs on the datasheet. If these impedances are very low, then we benefit (from a noise point of view) of matching impedances. Otherwise, me loose some performance.
But appart from noise issues that can be determined quite simply from the graph, does it improves/ changes anything? If it improves DC performance, chances are it will improve low frequency performance.
But it may have some drawbacks too.
And I fear using a jumper to keep this option on the printed board to choose between matched/direct ground, because a good direct ground is a close via or direct strip to ground
I hope I was clear, I'm a bit tired :/
Thanks,
Nicolas
Regarding the opamp impedances:
When using negative feedback around an amplifier with infinite (very very large) open loop gain, the net result is a forced condition whereby the voltage between the + and - inputs is very small, essentially zero.
So if the + input is at circuit common, the - input is considered to be at what is called "virtual ground", because the amplifier itself forces this.
Also, a current output DAC is an (ideally) infinite impedance in parallel with an ideal current source, equivalent to a voltage source in series with a zero impedance. So, the current the DAC forces out, in order to work optimally, wants a zero ohm load resistance. The higher the load, the more voltage must be developed across the load, and the more current is "lost" through the parallel source impedance. This equates to a form of distortion, in laymen's terms.
When the DAC outputs a current, the amplifier forces the - node to circuit common potential. The only way this is possible is for the amplifier to put out a voltage equal to I*Rf, where Rf is the feedback resistor. So whatever the DAC sources out as current, the opamp sinks by outputting the necessary voltage. This is how you force the DAC to see a zero ohm load; it doesn't have to perform any work; the opamp performs it.
As far as source impedances, you offset input bias currents by inserting a resistor in the + input, such that it is equal to the source impedance as seen by the - input. But the Thevenin equivalent source impedance of the DAC is zero ohms (or very low). In parallel with the feedback resistor, you end up with zero. So the resistor value at the + input should be zero. Most input bias currents are small enough anyway that you could negate this procedure anyway, especially if using a FET input opamp.
When using negative feedback around an amplifier with infinite (very very large) open loop gain, the net result is a forced condition whereby the voltage between the + and - inputs is very small, essentially zero.
So if the + input is at circuit common, the - input is considered to be at what is called "virtual ground", because the amplifier itself forces this.
Also, a current output DAC is an (ideally) infinite impedance in parallel with an ideal current source, equivalent to a voltage source in series with a zero impedance. So, the current the DAC forces out, in order to work optimally, wants a zero ohm load resistance. The higher the load, the more voltage must be developed across the load, and the more current is "lost" through the parallel source impedance. This equates to a form of distortion, in laymen's terms.
When the DAC outputs a current, the amplifier forces the - node to circuit common potential. The only way this is possible is for the amplifier to put out a voltage equal to I*Rf, where Rf is the feedback resistor. So whatever the DAC sources out as current, the opamp sinks by outputting the necessary voltage. This is how you force the DAC to see a zero ohm load; it doesn't have to perform any work; the opamp performs it.
As far as source impedances, you offset input bias currents by inserting a resistor in the + input, such that it is equal to the source impedance as seen by the - input. But the Thevenin equivalent source impedance of the DAC is zero ohms (or very low). In parallel with the feedback resistor, you end up with zero. So the resistor value at the + input should be zero. Most input bias currents are small enough anyway that you could negate this procedure anyway, especially if using a FET input opamp.
Hi,
Thank you very much. When I said "current source in series with eq resistor" I meant voltage in series with eq resistor, sorry, was tired. Other than that, I'm fine with opamps, fortunately xD
But I wasn't aware that a DAC's impedance would be as low as you said!
This nonideal zero impedances was my problem.
I had heard about a hundred of ohms or more!
But you seem to say I'm wrong.
Do you know what order of magnitude they are actually?
And yes, you're probably right in saying that, as the impedance is low and bias currents should be very low too, this in+ resistor would likely be useless.
Thank you very much for your very clear explanation!
Now just to know: if some people here have ideas about the output impedance of some DACs.... (not given in my datasheet, of course...).
Thank you very much. When I said "current source in series with eq resistor" I meant voltage in series with eq resistor, sorry, was tired. Other than that, I'm fine with opamps, fortunately xD
But I wasn't aware that a DAC's impedance would be as low as you said!
This nonideal zero impedances was my problem.
I had heard about a hundred of ohms or more!
But you seem to say I'm wrong.
Do you know what order of magnitude they are actually?
And yes, you're probably right in saying that, as the impedance is low and bias currents should be very low too, this in+ resistor would likely be useless.
Thank you very much for your very clear explanation!
Now just to know: if some people here have ideas about the output impedance of some DACs.... (not given in my datasheet, of course...).
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.