Was reading Walt Jung, Op Amp Applications Handbook, and something blew my mind, as usual.
And sure enough, it seems like most op amps people like for audio are bias compensated. OP27 obviously, but it appears that OPA828, OPA134, OPA1611, LT1028... All have bias currents and offset currents in the same magnitude ballpark as Walt describes, and bias currents listed with +/-.
So what does this mean for impedance balancing resistors and even more curiously, balancing capacitors? Should these inputs simply be tied to ground instead, if you're using a bias compensated op amp?
As an example, I'm building an RJM phonoclone for a friend, and I am getting just a few mV less offset from it when I ground both R1 and R4. (In real life not SPICE) Any reason not to just ground them?
Also does
Thanks.
And sure enough, it seems like most op amps people like for audio are bias compensated. OP27 obviously, but it appears that OPA828, OPA134, OPA1611, LT1028... All have bias currents and offset currents in the same magnitude ballpark as Walt describes, and bias currents listed with +/-.
So what does this mean for impedance balancing resistors and even more curiously, balancing capacitors? Should these inputs simply be tied to ground instead, if you're using a bias compensated op amp?
As an example, I'm building an RJM phonoclone for a friend, and I am getting just a few mV less offset from it when I ground both R1 and R4. (In real life not SPICE) Any reason not to just ground them?
Also does
mean there is no point trying to balance FET op amps?
Thanks.
Last edited:
Starting upside down: using the 10R and the 2.2k input R's on the two stages does not create equal DC impedances, because there is also bias current coming from the output through the feedback R. These stages have high gain so this cannot be ignored. Your *actual* input R is the source resistance in parallel with the feedback resistance, and the latter has to be divided by the stage gain, and reversed in direction as the stage is inverting. It gets complicated fast ;-) .
Bias compensation makes the actual bias current going into/out of the input pins smaller, but not necessarily equal. An extreme example: if the bias currents are not compensated but are equal, you can cancel their impact by matching input R as you were doing.
If the bias currents were compensated with the resulting bias smaller in magnitude but NOT equal, then you get offset even with equal input R. So it needs careful consideration and detailed datasheet study.
FET input opamps have such extremely low input bias currents that even with very unbalanced input R there's no appreciable offset from the bias currents, so balancing resistances is unnecessary.
Jan
Bias compensation makes the actual bias current going into/out of the input pins smaller, but not necessarily equal. An extreme example: if the bias currents are not compensated but are equal, you can cancel their impact by matching input R as you were doing.
If the bias currents were compensated with the resulting bias smaller in magnitude but NOT equal, then you get offset even with equal input R. So it needs careful consideration and detailed datasheet study.
FET input opamps have such extremely low input bias currents that even with very unbalanced input R there's no appreciable offset from the bias currents, so balancing resistances is unnecessary.
Jan
Last edited:
Wow, OK, thanks.
So in this design, in the first stage, it's 10r||332R=9.7R, and second stage is 768K||2.2K=2.194, so I was just rounding up to a widely available resistor value, which I thought would be close enough, but apparently it's not so simple anyway.
Add to the fact, I just measured this Denon DL103r with the LCR meter... It's a "14 ohm" cart, but its DCR is of course, higher... And not very well matched... 15.7 on one channel, and 16.0 on the other. And then there's cables and mechanical connections, all of which affect the gain and DC offset of this design.
So in this design, in the first stage, it's 10r||332R=9.7R, and second stage is 768K||2.2K=2.194, so I was just rounding up to a widely available resistor value, which I thought would be close enough, but apparently it's not so simple anyway.
Add to the fact, I just measured this Denon DL103r with the LCR meter... It's a "14 ohm" cart, but its DCR is of course, higher... And not very well matched... 15.7 on one channel, and 16.0 on the other. And then there's cables and mechanical connections, all of which affect the gain and DC offset of this design.
For audio bias currents dont often cause problems unless high impedance front ends.
However I recently designed a front end for an oscilloscope.
That is 1 meg input impedance and possibly AC coupled and so any substantial bias current will cause quite an offset voltage on the input.
A bi polar op amp just doesnt cut it for that project.
However I recently designed a front end for an oscilloscope.
That is 1 meg input impedance and possibly AC coupled and so any substantial bias current will cause quite an offset voltage on the input.
A bi polar op amp just doesnt cut it for that project.
Found this relevant article enlightening...
Don’t Use the Resistor I Told You to Use | Analog Devices, Inc.
Jan, am I getting this math right then? First stage gain is 33.2X. Is it 10r || (332R/33.2) = 5R ??
Don’t Use the Resistor I Told You to Use | Analog Devices, Inc.
Jan, am I getting this math right then? First stage gain is 33.2X. Is it 10r || (332R/33.2) = 5R ??
actually the compensation is mentioned at the beginning of the datasheet.
https://www.analog.com/media/en/technical-documentation/data-sheets/op27.pdf
The low ohm resistors do not increase the noise significantly.
To connect the capsule directly to the virtual gnd is useful, better would be a OP37 here, where you could increase the 330 ohm to 1k.
Even better the LT1028 would cut the noise by factor 3.
https://www.analog.com/media/en/technical-documentation/data-sheets/op27.pdf
The low ohm resistors do not increase the noise significantly.
To connect the capsule directly to the virtual gnd is useful, better would be a OP37 here, where you could increase the 330 ohm to 1k.
Even better the LT1028 would cut the noise by factor 3.
Yes I think it is. The voltage across the 332 is much larger than across the 10R so it takes more of the bias current 'per ohm' so to say. How much more? The gain factor more.
Another way to look at this is to imagine a bias current generator in the opamp, looking into a load of 10R to say ground, and 332R to a large negative voltage. So the bias current divides through the two resistors and generates the offset voltage across them while doing so.
A quick test in your sim could be to change R1 to 5R and see what happens to the offset.
Can someone confirm this is sensible?
Edit: I think I have this wrong. The bias current divides over the input and feedback resistor. The feedback resistor terminates in a much higher voltage than the input resistor (by the gain), but that feedback resistor is also much larger (again, by the gain), so I think the correct way is that the offset voltage would be the bias current x Rin//Feedback. Sorry guys, senior moment ...
Jan
Last edited:
The bias current is not dependent on the offset voltage.
the +/-50nA will flow half though the MC and the other half though the 332 ohm. but this varies when there is a 0-crossing all will flow through the coil, so it is modulated with the music signal, I do not know what effect this has sonically.
the coil output current into 0 ohm is 17.9uA, that is 60dB above the bias modulation current.
the +/-50nA will flow half though the MC and the other half though the 332 ohm. but this varies when there is a 0-crossing all will flow through the coil, so it is modulated with the music signal, I do not know what effect this has sonically.
the coil output current into 0 ohm is 17.9uA, that is 60dB above the bias modulation current.
Bias compensation circuitry injects current noise into the inputs, but often this is closely correlated for the two inputs.
If the impedance seen by each input is different (impedance, not resistance), then this extra current noise doesn't cancel and will be converted into voltage noise across the inputs.
The datasheet current noise spec often is done using a circuit which balances the impedance seen by each input, in order to hide this extra source of current noise from you...
When noise matters (high gain MC phono amp, etc), its sometimes the case you can't balance the input impedance (without adding more sources of noise...), so the datasheet current noise spec won't apply.
Its really important to know this compensation current magnitude in order to estimate the current noise actually seen in realistic circuits, but often this is hidden from you.
If the impedance seen by each input is different (impedance, not resistance), then this extra current noise doesn't cancel and will be converted into voltage noise across the inputs.
The datasheet current noise spec often is done using a circuit which balances the impedance seen by each input, in order to hide this extra source of current noise from you...
When noise matters (high gain MC phono amp, etc), its sometimes the case you can't balance the input impedance (without adding more sources of noise...), so the datasheet current noise spec won't apply.
Its really important to know this compensation current magnitude in order to estimate the current noise actually seen in realistic circuits, but often this is hidden from you.
The optimum choice of op amp and resistor selection depends on the specific circuit use, such as input impedance, output load, input/ouput signal swing, required gain, bandwidth, signal levels, noise and distortion requirements.
If one is working with low impedance audio circuits, say less than 2k impedance, then the BJT input op amps may work out better due to lower input voltage noise.
If the circuit impedance is say above 10K ohm, then a FET input op amp may be a better solution, since it most likely has lower input current noise. The BJT input op amp with large input impedance resistors will have the effective input voltage noise swamped out by input current noise through the large input resistance.
Bias current through a resistance will drop a voltage across the resistance, if the input resistance seen by the input bias current are largely different, then there will be an input offset voltage due to the input bias current through the different effective resistances seen from the two input nodes of the Op Amp. The DC input impedance imbalance is the issue for input DC offset.
Using a low input offset op amp effectively, also requires careful attention to input impedances, value resistors used, input bias current, DC gain and whether the output gets AC coupled.
All diode junctions have a non-linear voltage dependent capacitance (parasitic varactor) which can lead to harmonic distortion with a large source impedance and large voltage swings (such as in an input voltage follower application). For those applications, a dielectric isolated FET input op amp may be appropriate choice (one example is the OPA1642).
Also, if using FET input op amps, one must be careful about imbalance in the input impedance driving large nonlinear input capacitance with large unbalanced voltage swings.
In a MOSFET input op amp, the input bias current is coming from the ESD diodes on the input. In a JFET input opamp, input bias current is a combination of ESD diode and gate junction diode current.
Many low input bias current BJT input op amps either use super beta transistors or bias current compensation on the front end.
The op amp choice and circuit design approach is always a compromise of design requirements & tradeoffs. Every op amp is good for some type of circuit, the proper choice is dependent on the specific case.
If one is working with low impedance audio circuits, say less than 2k impedance, then the BJT input op amps may work out better due to lower input voltage noise.
If the circuit impedance is say above 10K ohm, then a FET input op amp may be a better solution, since it most likely has lower input current noise. The BJT input op amp with large input impedance resistors will have the effective input voltage noise swamped out by input current noise through the large input resistance.
Bias current through a resistance will drop a voltage across the resistance, if the input resistance seen by the input bias current are largely different, then there will be an input offset voltage due to the input bias current through the different effective resistances seen from the two input nodes of the Op Amp. The DC input impedance imbalance is the issue for input DC offset.
Using a low input offset op amp effectively, also requires careful attention to input impedances, value resistors used, input bias current, DC gain and whether the output gets AC coupled.
All diode junctions have a non-linear voltage dependent capacitance (parasitic varactor) which can lead to harmonic distortion with a large source impedance and large voltage swings (such as in an input voltage follower application). For those applications, a dielectric isolated FET input op amp may be appropriate choice (one example is the OPA1642).
Also, if using FET input op amps, one must be careful about imbalance in the input impedance driving large nonlinear input capacitance with large unbalanced voltage swings.
In a MOSFET input op amp, the input bias current is coming from the ESD diodes on the input. In a JFET input opamp, input bias current is a combination of ESD diode and gate junction diode current.
Many low input bias current BJT input op amps either use super beta transistors or bias current compensation on the front end.
The op amp choice and circuit design approach is always a compromise of design requirements & tradeoffs. Every op amp is good for some type of circuit, the proper choice is dependent on the specific case.
Attachments
Last edited:
Yeah, I actually considered the OPA1641 for a phono preamp I'm designing, based on Walt Jung's two-stage, passive shunt EQ phono preamp. But I dumped it in favor of the OPA828 for it's trimmed offset voltage and high GBW. Current noise slightly higher at 1.2f versus 0,8f, but voltage noise is lower, 4nV versus 5. Either is dwarfed by noise from the 47K and cartridge anyways.
The 1642 has admirably flat input capacitance though... And the first stage only runs at 28db gain, so that it doesn't clip on high-frequencies before the second stage does. The way it's balanced, they both clip at the same amplitude at 20KHz.
So this first stage is working with a small signal and low gain. Is that flat capacitance important in this application?
And hmm... the 1641 swings closer to the rails than the 828, and that's important in this design, more headroom is better...
I might be talking myself into the 1641 haha... The offset's not a big deal I guess.
The 1642 has admirably flat input capacitance though... And the first stage only runs at 28db gain, so that it doesn't clip on high-frequencies before the second stage does. The way it's balanced, they both clip at the same amplitude at 20KHz.
So this first stage is working with a small signal and low gain. Is that flat capacitance important in this application?
And hmm... the 1641 swings closer to the rails than the 828, and that's important in this design, more headroom is better...
I might be talking myself into the 1641 haha... The offset's not a big deal I guess.
Just to be clear, its the imput _impedance_ that matters, not the resistance. Current is multiplied by any impedance, not just the resistive part, to give voltage. This is why low current noise is important for MM cartridges as they can have a henry or so of inductance, turning 1pA/√Hz of noise into maybe 50nV/√Hz+ at 10kHz - making it pointless to have ultra-low voltage noise figures for an MM amp, the current noise should be addressed first.
1H is 63k ohms of reactance at 10kHz, MM cartridges 1k or so of source resistance is insignificant by comparison across much of the audio spectrum.
Even the professionals get this wrong, I've seen some datasheet for an opamp showing a phono preamp and miscalculating the noise performance by assuming an MM cartridge is purely resistive...
So anyways, getting back on topic...
Would it be correct to say, that if low offsets aren't required, then balancing the inputs is still preferred for noise reasons?
As Mark mentioned, the current compensation magnitude is unknown so it's impossible to make accurate estimates of the noise.
So maybe I'm being thick, but what is the formula for determining the balancing resistor? I've always read it's just R2 || R1, but does R2(Feedback) need to be divided or multiplied by the gain in some way?
Would it be correct to say, that if low offsets aren't required, then balancing the inputs is still preferred for noise reasons?
As Mark mentioned, the current compensation magnitude is unknown so it's impossible to make accurate estimates of the noise.
So maybe I'm being thick, but what is the formula for determining the balancing resistor? I've always read it's just R2 || R1, but does R2(Feedback) need to be divided or multiplied by the gain in some way?
If you don't really care about input offset voltage, then one does not need to balance the input resistors for input bias current induced Vos. The DC impedance looking out from the input pin of the op amp to the connected circuit elements would normally be balanced to minimize input bias current induced input offset.
Input noise is a different issue.
Input noise is a different issue.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.