Ideal feedback resistor vales
I want to know how to determine what the ideal feedback resistor values for a given circuit would be.
I have an op amp based line amp (OPA627) that I use with no buffer. It has a gain of 3 or 9db. I have chosen feedback resistors of 1000 and 2000.
My question is, would higher value resistors at the same ratio be better? Would they raise the output impedance?
These feedback resistors are in parallel with the input of the next stage (amplifier), correct?
By using low values am I causing the op amp to drive to low of an input impedance?
I also have a line amp using the LME49710 with a BUF634 in the feedback loop. I am using the same values for the resistors here for the same amount of gain (or slightly lower with the buffer).
Are the values more important here or less important?
I want to know how to determine what the ideal feedback resistor values for a given circuit would be.
I have an op amp based line amp (OPA627) that I use with no buffer. It has a gain of 3 or 9db. I have chosen feedback resistors of 1000 and 2000.
My question is, would higher value resistors at the same ratio be better? Would they raise the output impedance?
These feedback resistors are in parallel with the input of the next stage (amplifier), correct?
By using low values am I causing the op amp to drive to low of an input impedance?
I also have a line amp using the LME49710 with a BUF634 in the feedback loop. I am using the same values for the resistors here for the same amount of gain (or slightly lower with the buffer).
Are the values more important here or less important?
Remember that essentially no current flows in the inputs, so the load to ground is just the series sum of the resistors, plus the actual load. Most op-amps drive 2K nicely, and some do well with 600 ohms. At very low signal levels, they can go even lower, thus phono input stages with a few hundred ohms. IMO, your values are fine. If you go a lot lower, you might get more distortion, and if you go (quite a bit) higher, stray capacitance and pickup get involved. There's really a huge range of acceptable values, with little difference between them. I typically use about 10K as a starting point, though with bipolar amps you want to also think about equalizing the values seen by both inputs.
Hi,
yes, the Zin of the next stage appears in parallel to the Rfb of this stage.
So take care to not go too low, particularly at maximum output voltage. Have a look at the distortion Vs load impedance and Output voltage Vs load impedance.
Both these will show a variety of closely spaced load curves and then a curve where performance is markedly different. I would suggest that the one that is different has gone too low. Your parallel impedance should be about the same value (or higher) as the lowest of the closely grouped bunch.
yes, the Zin of the next stage appears in parallel to the Rfb of this stage.
So take care to not go too low, particularly at maximum output voltage. Have a look at the distortion Vs load impedance and Output voltage Vs load impedance.
Both these will show a variety of closely spaced load curves and then a curve where performance is markedly different. I would suggest that the one that is different has gone too low. Your parallel impedance should be about the same value (or higher) as the lowest of the closely grouped bunch.
With FET-input op-amps in non-inv mode it is good practise to exactly match the impedances at both input pins, Z(+IN)==Z(-IN). That is, not only the resistive part. This keeps distortion low and helps to minimize offsest. Of course one usually strives to use as low an impedance as one can afford (without getting other drawbacks, e.g. overloading the output), from noise and stability standpoints. Walt Jung (www.waltjung.org) has elaborated on this.
For your example, we would get a 1k//2k=667Ohms impedance at the -IN, so the source impedance should also be 667Ohms. With normal sources like a CD player (~100Ohms Z_out) one would need a series resistor in the input, which would result in a neglegible noise penalty for the OPA627 (below 3k total Z there is no improvement). Not a bad idea anyway...
- Klaus
For your example, we would get a 1k//2k=667Ohms impedance at the -IN, so the source impedance should also be 667Ohms. With normal sources like a CD player (~100Ohms Z_out) one would need a series resistor in the input, which would result in a neglegible noise penalty for the OPA627 (below 3k total Z there is no improvement). Not a bad idea anyway...
- Klaus
KSTR said:
Interesting, I have never given the impedance that the inverting input sees any consideration.
I was more concerned with what load the OPA627 sees due to not using it with a buffer.
The non-inverting input is preceded by a 25K discrete attenuator, how would one adjust for that?.
With an input impedance of 100K on my amplifier that this line amp is driving and the feedback resistors I'm using, the load the op amp sees is 2.913K? I'm I figuring this correctly?
Does that seem low to anyone?
hags said:
one of the biggest problems for a passive attenuator is worsened if the following amp has DC gain.hags said:
The amplifier expects to see a lowish but constant source resistance to allow correct input DC conditions to be set up.
If the attenuator is not DC blocked, the input sees a variable source resistance as the volume is adjusted. This variable source resistance changes the DC operating conditions with the result that the input offset current causes an output offset voltage.
If the amp is also DC coupled then the input offset voltage gets multiplied by the DC gain. Change the volume and the output offset changes and sometimes quite markedly.
The variable resistance must be made to have a very small range of source resistance or else DC blocked to remove it from the DC set up. FET inputs have a very low input offset current (although I believe it increases with increase in temperature) and can tolerate a larger range of source resistance.
AndrewT said:
Uh, I can understand that but the OPA627 allows for the addition of a dc offset null circuit. I have implemented the null circuit and have zero dc offset through the entire range of the volume control, as measured at the outputs. I have no clicks or pops when turning the attenuator either.
I am concerned about the non-inverting input seeing a changing impedance for reasons of distortion though.
That explains the Jim Williams advice in one of the LT databooks- if you have the freedom to do so, invert.
As for sources and RF caps, I like everything defined. It always drove me crazy to see people adjust a gain stage by adjusting the feedback. BW, slew, noise and who-knows-what-else changes. Thus, I don't like things that vary my offset and BW with gain or different input sources. I suppose it can't be helped at some level, but the more obvious variables should be minimized. But... in this case we were talking about an OPA627. The thing is made for electrometers and the like, and the input currents and offsets are so close to zero as not to matter. This is an opamp where the pcb layout, leakage, and thermoelectric effects will almost always determine the performance of the part, not the part itself. Thus the cost. I used to use them for input amps in scanning tunneling microscopes, where the input specs are about as critical as it gets. So, IMO the enemy is the outside world, not input effects. If there's a chance of DC getting in, raise the amplifier impedance and DC block with a nice cap. You won't pay much penalty for higher feedback resistors. RF should be taken care of prior to a gain control, in fact, IMO it should be taken care of right at the input connector and not allowed in the box at all ever, because once it gets past the connector it's that much harder to eliminate.
As for sources and RF caps, I like everything defined. It always drove me crazy to see people adjust a gain stage by adjusting the feedback. BW, slew, noise and who-knows-what-else changes. Thus, I don't like things that vary my offset and BW with gain or different input sources. I suppose it can't be helped at some level, but the more obvious variables should be minimized. But... in this case we were talking about an OPA627. The thing is made for electrometers and the like, and the input currents and offsets are so close to zero as not to matter. This is an opamp where the pcb layout, leakage, and thermoelectric effects will almost always determine the performance of the part, not the part itself. Thus the cost. I used to use them for input amps in scanning tunneling microscopes, where the input specs are about as critical as it gets. So, IMO the enemy is the outside world, not input effects. If there's a chance of DC getting in, raise the amplifier impedance and DC block with a nice cap. You won't pay much penalty for higher feedback resistors. RF should be taken care of prior to a gain control, in fact, IMO it should be taken care of right at the input connector and not allowed in the box at all ever, because once it gets past the connector it's that much harder to eliminate.
Conrad Hoffman said:
Haven't had a need yet for a DC blocking cap in my application. I don't have any problems with RF or EMI, or at least I haven't noticed it.
I do have problems with a LME49710 with a BUF634 for a buffer and noise from a CD player.
Almost every design I see, or application note, shows a pot/attenuator right before the op amp.
I'm using a 100K input resistor after the attenuator at the non-inverting pin.
You could use buffers or an attenuator with a constant output impedance, but that would complicate the design and not necessarily improve the sound.
Please don't misread my tone, but, does the 25K volume control really have that big of an impact on the performance/sonics?
I think many of you have lost track of the original queston.
"My question is, would higher value resistors at the same ratio be better? Would they raise the output impedance?"
The answer is that so long as the resistances are reasonable, they will not effect the closed loop output impedance of the amp.
"My question is, would higher value resistors at the same ratio be better? Would they raise the output impedance?"
The answer is that so long as the resistances are reasonable, they will not effect the closed loop output impedance of the amp.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.