Hi fellas,
Got diffused by DC servo OPamp.
As Mr. Douglas said,
........It is essential to use FET-input op-amps, with their near-zero bias and offset currents. Bipolar op-amps have many fine properties, but they are not useful here. You will need a reasonably high-quality FET op-amp to beat non-servo power amplifiers......
However, as I noticed, those OPamps with low or ultralow input offset (Volt, Current) are not always JFET-input. Conversely, those JFET-input OPamps tend to have higher input offset though some of hem are still quite good, say OPA827 (Vos=75uV, Ios=3pA). However, non-JFET could have better figure in these properties.
So, When choosing DC servo OPamps, should it must be JFET-in, or should it be highest performance?
Also, Mr. Self, are you suggesting using JFET-input OPamps as servo despite that they are slightly higher in Vos and Ios? Why?
Thank you!
Got diffused by DC servo OPamp.
As Mr. Douglas said,
........It is essential to use FET-input op-amps, with their near-zero bias and offset currents. Bipolar op-amps have many fine properties, but they are not useful here. You will need a reasonably high-quality FET op-amp to beat non-servo power amplifiers......
However, as I noticed, those OPamps with low or ultralow input offset (Volt, Current) are not always JFET-input. Conversely, those JFET-input OPamps tend to have higher input offset though some of hem are still quite good, say OPA827 (Vos=75uV, Ios=3pA). However, non-JFET could have better figure in these properties.
So, When choosing DC servo OPamps, should it must be JFET-in, or should it be highest performance?
Also, Mr. Self, are you suggesting using JFET-input OPamps as servo despite that they are slightly higher in Vos and Ios? Why?
Thank you!
Moreover at elevated temperature JFET gate current could exceed the of the bjt input stage: it goes up 2 times every 10 degrees C. Back in a day there were no good CMOS, however good bjt did exists like LM108a. Unless you are forming frequency response with servo any amp with low input current is O'K. So I would check newer CMOS based O/A.
I'm guessing that the JFET input is a sort of "shorthand" for low offset/bias currents, because in low-cost opamps it's fairly safe to say that JFET is better than bipolar in this regard. Also, (again a guess) I think that high-BW fast opamps are actually potential troublemakers in a suboptimal layout, so you don't want to put something fancy in the servo position. Last but not least, the price difference between a pair of e.g. OPA827s and a pair of TL081s you can probably spend better elsewhere 🙂
So, if I don't care about the cost but performance, theoretically speaking, should I chose
1. JFET with "OK but not best" performance
2. Best performance figure BJT OPamps? (ex: LTC1151, Vos=0.05uV,Ib=5pA)?
In DC servo case, which one is more important, JFET or Vos/Ios? why?
1. JFET with "OK but not best" performance
2. Best performance figure BJT OPamps? (ex: LTC1151, Vos=0.05uV,Ib=5pA)?
In DC servo case, which one is more important, JFET or Vos/Ios? why?
If you are really, Really worried about this, you could use Mr. Self's newest recommended part: the TL051 JFET input opamp from Texas Instruments. It's got low offset voltage, it's got low input bias current, it sells for an extremely low price. Datasheet attached.
Mr. Self does not say this, but I will: the TL051 also has a feature that paranoid worry-warts will love. Its offset voltage is field trimmable via an external potentiometer connected to pins 1 and 5. So you can dial the output offset voltage of your audio amplifier down below 1 microvolt, if that's what it takes to relieve your insomnia. Be sure to purchase a calibrated voltmeter that will accurately read down below 1 microvolt, of course.
_
Mr. Self does not say this, but I will: the TL051 also has a feature that paranoid worry-warts will love. Its offset voltage is field trimmable via an external potentiometer connected to pins 1 and 5. So you can dial the output offset voltage of your audio amplifier down below 1 microvolt, if that's what it takes to relieve your insomnia. Be sure to purchase a calibrated voltmeter that will accurately read down below 1 microvolt, of course.
_
Worry about the Input bias current (Ib) first, then offset current (Ios) and voltage offset (Vos). JFET devices have very low Ib to start with. It's Ib through the inevitably high value servo input resistor that causes the biggest error.
Most servo has high impedance LP network, typically 1Mohm -- 1µF. The opamp therefore has to have at least 100x ~ 1000x higher input resistance.
Most FET input opamp has >1G input resistance. Seldom the case for bipolar input opamps.
And if you would test a 1s time constant servo on a typical circuit, you will be amazed how much signal you see at its output, well above 100Hz. So for a servo to be totally transparent, it has to be low distortion. No point saving 2 USD and ruin the sound.
As with coupling caps, best servo is no servo.
Cheers, Patrick
Most FET input opamp has >1G input resistance. Seldom the case for bipolar input opamps.
And if you would test a 1s time constant servo on a typical circuit, you will be amazed how much signal you see at its output, well above 100Hz. So for a servo to be totally transparent, it has to be low distortion. No point saving 2 USD and ruin the sound.
As with coupling caps, best servo is no servo.
Cheers, Patrick
Glad it helped. I think Patrick has a good point too - try and avoid them, coupling caps aren't that bad, a little offset isn't fatal!
That TO051 is a nice OpAmp, isn't it? 3.5 MHz unity-gain bandwidth, excellent pulse response, 150,000 open loop gain. FET input, low picoamp input bias current. So low in fact that the datasheet specifies differential metrology for measuring it on individual testbed devices.
Nice.
GoatGuy
Nice.
GoatGuy
Don't forget to attenuate the servo output by a factor of ((R2+R3)/R2) = 201X (46dB) because your oscilloscope is looking at node A, but the amplifier is looking at node B. See Figure 23.2 from Self 6th edition. Errors / distortions introduced by the servo, are attenuated by a factor of 201X.
_
Attachments
Last edited:
Thank you for reminding; I am fully aware of that.
I do not want to see any signal at 100Hz, however small or low distortion, from the servo. Hence no servo, unless I cannot avoid. Else might just as well build opamp circuits or chipamps. No need to do discrete.
One can also build servos using discrete JFETs of course.
But each to his own.
Cheers, Patrick
I do not want to see any signal at 100Hz, however small or low distortion, from the servo. Hence no servo, unless I cannot avoid. Else might just as well build opamp circuits or chipamps. No need to do discrete.
One can also build servos using discrete JFETs of course.
But each to his own.
Cheers, Patrick
Then you might want to investigate T. Christiansen's 3rd order servo, which attenuates 100 Hz response (along with servo opamp distortion) by another 45 to 60 dB depending on details of pole location. It's probably the highest order circuit that can be easily realized with a single JFET-input opamp.
For an amplifier, DC Offset below +-25mV is fine. I use JRC072 for DC Servo, I get DC Offset below 1mV.
If Input bias current is low, we can use small value of capacitor and high value resistor for the integrator (usually 1uF and 1M). But if input bias current is high, we can use high value capacitor and small value resistor (50uF bipolar and 18K).
Of course small value capacitor is better, because we can use high quality capacitor. But, can it influence the quality of amplifier, I do not know.
If Input bias current is low, we can use small value of capacitor and high value resistor for the integrator (usually 1uF and 1M). But if input bias current is high, we can use high value capacitor and small value resistor (50uF bipolar and 18K).
Of course small value capacitor is better, because we can use high quality capacitor. But, can it influence the quality of amplifier, I do not know.
Some circuit topologies do a great job with dealing with DC offset automatically. No op amp servo needed. Take for example, the ALPHA designed by Hugh Dean. With zero adjustment trimpots, the output is within 25mV (1mV and 6mV on my particular Alpha 55w). And this is from cold startup to fully warmed up Class A operation. It’s pretty neat but it tracks as temps rise. There are no “sensors” mounted on heatsinks or MOSFETs either.
Aksa Lender P-mos Hybrid Aleph (ALPHA) Amplifier
The other one that also has self correcting zero offset is the TimS Simple haeadphone amp. My particular examples (3 amps built so far) are all under 6mV and track from cold to fully warmed up.
On both amps, I match the input transistors (BJT for Hfe) or JFETs (Iq at operating voltages). Nothing fancy testing and matching wise.
Simple High Performing Headphone Amp
Aksa Lender P-mos Hybrid Aleph (ALPHA) Amplifier
The other one that also has self correcting zero offset is the TimS Simple haeadphone amp. My particular examples (3 amps built so far) are all under 6mV and track from cold to fully warmed up.
On both amps, I match the input transistors (BJT for Hfe) or JFETs (Iq at operating voltages). Nothing fancy testing and matching wise.
Simple High Performing Headphone Amp
I'm kind of missing why “the snottiness”, Patrick. You've asserted some pretty strong objections to design strategies that actually “solve” any perceived distortion issues one might nominally have. But your response has been dismissive. Why?
Just asking,
GoatGuy
Try ADA 4522-1 as brought to our attention by Monty McGuire.
Very low offset and drift, very low 0.1 to 10 hz noise , and cheap
ADA 4522 -2 if you need a dual
And Evul is right and John Curl avoids opamps if he can or does his best to get the most out of them.
Very low offset and drift, very low 0.1 to 10 hz noise , and cheap
ADA 4522 -2 if you need a dual
And Evul is right and John Curl avoids opamps if he can or does his best to get the most out of them.
Last edited:
- Status
- Not open for further replies.