Thank you.
I've been trying to understand just how a PP Class AB op amp would revert to class AB operation when biased to class A and a load is applied, and can't see it.
It is my understanding that negative DC voltage will offset the output in the negative direction. To compensate, the op amp will favour the use of the components attached to the postive rail. If there is sufficient DC bias applied, then the op and will use only the components attached to the positive rail, which defines single ended/Class A operation.
I have to agree though, that distortion is already vanishingly small, then, what is there to gain? Is slew rate or some other measure of transient response somehow affected? Is the distortion spectrum changed ie the balance between 2nd and 3rd order harmonics?
Perhaps the latter, because once out of PP operation the even order harmonics will no longer cancel.
I've been trying to understand just how a PP Class AB op amp would revert to class AB operation when biased to class A and a load is applied, and can't see it.
It is my understanding that negative DC voltage will offset the output in the negative direction. To compensate, the op amp will favour the use of the components attached to the postive rail. If there is sufficient DC bias applied, then the op and will use only the components attached to the positive rail, which defines single ended/Class A operation.
I have to agree though, that distortion is already vanishingly small, then, what is there to gain? Is slew rate or some other measure of transient response somehow affected? Is the distortion spectrum changed ie the balance between 2nd and 3rd order harmonics?
Perhaps the latter, because once out of PP operation the even order harmonics will no longer cancel.
The AB stage biased to class A will fall back to AB when the voltage across the load impedance rises to a given level. That level is determined by the constant current flowing but you are limited as to how big that current can be by the opamps output stage limits. The more current the opamp sinks, the less is available to the load.
If you tried this on a 741 or a 4558 then you might actually get some benefit provided you kept signal levels low and impedances high, but on state of the art devices dedicated to audio I think it just compromises performance.
If you tried this on a 741 or a 4558 then you might actually get some benefit provided you kept signal levels low and impedances high, but on state of the art devices dedicated to audio I think it just compromises performance.
I don't highly recommend the fet ccs because of normal production parameter variation, some times large minimum V for actual "pentode mode" operation where you get current regulation rather than just a dirty Si nonlinear resistor
the "ring of two" bjt ccs gives more accurate, calculable I, low V operation and high Z out for pennies in parts
My current hypothesis for how classA loading works is that its making the power supply cleaner, not that its reducing OPS crossover distortion. Adding a lot more caps to the rails subjectively gives similar results to CCS biassing. If you go along with this hypothesis then to get the cleanest rails a balanced approach is called for where the load current flows between two SE classA biassed stages and hence doesn't modulate the supply rails at all (to a first order).
Abraxalito, I will let you know shortly, having found a couple of jfets I can use (2N5484). Now, I just need 72 ohm resistors (or thereabouts). The interesting part is that the power supply is regulated.
Yes, it will move back to class AB when the voltage output exceeds that of the bias. But that is going to be a terrific amount. I'm talking about the output of a phono section here that won't push more than a volt and a half max into a 10K pot (DCB1 buffer). So, even with 2 V output, 3 mA ought to cover it....
Yes, it will move back to class AB when the voltage output exceeds that of the bias. But that is going to be a terrific amount. I'm talking about the output of a phono section here that won't push more than a volt and a half max into a 10K pot (DCB1 buffer). So, even with 2 V output, 3 mA ought to cover it....
Last edited:
Does it matter where on the supply rail the current source is powered? we always see drawings with the CCS source near the op amp. Yet, I have seen advice suggesting that the CCS should be AFTER a coupling resistor, to lower the effect the capacitance of the jFET has on the op amp.
An Idss "matched" pair of 2N5486's with about 525 ohms between source and gate will give 3.15 mA from a -15 V rail attached to the gate.
Will be putting this together tomorrow.
An Idss "matched" pair of 2N5486's with about 525 ohms between source and gate will give 3.15 mA from a -15 V rail attached to the gate.
Will be putting this together tomorrow.
The data sheets say that the output capacitance 0f the 2n5456 is 2 pF at this Vgs. Can't see this having an effect at audio frequencies.
As for the location of the power supply take off, it should not really matter, but I will make sure it does not bypass the force/sense point of the shunt regulated PS.
As for the location of the power supply take off, it should not really matter, but I will make sure it does not bypass the force/sense point of the shunt regulated PS.
the internals of the op amp make the difference
some may like pull-up better than pull-down
there may be an optimum current
you really have to measure the distortion with your parts, load to know if you've made any difference, to tune it for best performance when it does make a difference
people could actually look at the reference I quoted above
some may like pull-up better than pull-down
there may be an optimum current
you really have to measure the distortion with your parts, load to know if you've made any difference, to tune it for best performance when it does make a difference
people could actually look at the reference I quoted above
I am just finishing off a new preamp. All the opamp stages are bootstrapped into class A and buffered with a single ended class A OPS running inside the opamp feedback loop. The opamps only see a very small load like this because the load induced distortion component is completely removed.
I am using a 'relaxed' PSU architecture where each stage is decoupled with a 22 ohm and 100 uF cap ( both rails). The class A constant current load means you only see low order ( 2 nd and 3rd) harmonics on the rails. The filter kills regulator noise and isolates each of the stages.
If you run an opamp in normal class AB mode, the wide band hash on the rails (current) runs into the 100s of kHz.
I'll be posting up a small write up in a few weeks. Mechanics are taking a bit longer than I anticipated.
I am using a 'relaxed' PSU architecture where each stage is decoupled with a 22 ohm and 100 uF cap ( both rails). The class A constant current load means you only see low order ( 2 nd and 3rd) harmonics on the rails. The filter kills regulator noise and isolates each of the stages.
If you run an opamp in normal class AB mode, the wide band hash on the rails (current) runs into the 100s of kHz.
I'll be posting up a small write up in a few weeks. Mechanics are taking a bit longer than I anticipated.
Last edited:
So, it is done, and I cannot tell much, but it seems a bit more relaxed, in the sense that I relax a bit more when I hear it. I know, totally subjective, but I've yet to order the distortion analyzer. Only 3.15 mA negative bias on the output OP27.
Could this not also be done on the input opamp? If so, this ought to be more effective here, no?
Bonsai, I'm looking forward to your write up.
Could this not also be done on the input opamp? If so, this ought to be more effective here, no?
Bonsai, I'm looking forward to your write up.
Last edited:
I did have a quick skim through actually. Good to see real testing done... mind you the author will have the best of the best test equipment wise available to him.
Good article though
If you are pleased with the results then that is all that matters. I find you need to listen over a long period to reliably identify subtle differences. Keep coming back to it, and listen over and over again. If you honestly sense a consistent improvement after say 7-10 days then that indicates it was beneficial.
the CCS has a varying voltage when signal is passing.
If the chip has +-15Vdc supplies and the chip outputs 0Vac then the CCS sees ~15Vdc
If the output increases to 12Vpk (8.5Vac), then the CCS sees voltage varying from ~27Vdc to ~3Vdc
The capacitance at 27Vdc will be very different from the capacitance at 3Vdc.
That change in capacitance loading will introduce distortion.
I seem to recall that the added distortion varies as the square of the voltage.
Keeping the output voltage low will reduce the distortion.
Increasing the opamp supply rails will reduce the distortion.
I often recommend that opamps should have 10dB to 20dB of overhead, over the usual maximum signal, to reduce distortion generally. The above would be another case where increased overhead would offer a benefit.
eg a CDP usually has a maximum signal of 2.2Vac (6Vpp)
10dB of overhead requires the amplifier to accept ~6.5Vac before distortion becomes significantly impaired.
20dB of overhead requires the amplifier supply rails to be increased substantially to allow it to handle 22Vac signal levels.
Last edited:
An OP27 typically dissipates 90mW. Given rail voltages of +/- 15, this implies that the maximum current draw will be 3 mA. ( 3mA x 30 V = 90 mW )
Does this mean the upper limit for the bias current is also 3mA? or should it be strictly less..... I have a couple CCS that were 3.15 mA. Is that going to degrade op amp performance?
Does this mean the upper limit for the bias current is also 3mA? or should it be strictly less..... I have a couple CCS that were 3.15 mA. Is that going to degrade op amp performance?
3ma is the current draw of all the internal circuitry, the output stage will be running at considerably less than that. The maximum current the output can source or sink is in the region of 17ma for an OP27. So your CCS value needs to be in the range 0 to 17ma, and in practice less than that, but enough to be able to meet the demands of your loading on the opamp.
Sometimes a max current figure is specified although the data sheet I had for the OP27 doesn't quote this directly. You can deduce a value by looking at what voltage swing it can develop across 600 ohms which is the "lowest" impedance value commonly quoted. I think it was around 11 volts give or take for the OP27. The power dissipation quoted in this case relates to quiescent conditions. I'm not sure if there was a max dissipation for the package quoted as well although I don't just remember seeing it.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.