"Is there something on the data sheet related to the max current the op amp can source or sink? How can you figure this out?"
Not directly. This is because the prime application of an op-amp is in a linear loop with the output controlled by the feedback components.
"Maximum current" would involve saturating one half of the output stage and isn't easy to measure.
HOWEVER, the output swing under load gives us what we want! 12v at 2K means at least 6mA is available while staying linear.
Don't concern yourself with Pd much - it is just derived from output V / RL parameters.
Not directly. This is because the prime application of an op-amp is in a linear loop with the output controlled by the feedback components.
"Maximum current" would involve saturating one half of the output stage and isn't easy to measure.
HOWEVER, the output swing under load gives us what we want! 12v at 2K means at least 6mA is available while staying linear.
Don't concern yourself with Pd much - it is just derived from output V / RL parameters.
What kind of circuit are you thinking of here?
Generally you may want to Class A bias when the output stage is regularly pushed out of Class A operation. Since usually it is a push-pull affair, that happens when peak current exceeds 2x Iq. A number of opamps seem to run an Iq in the 300 µA .. 1 mA vicinity, though higher values do occur. You may be able to infer approximate Iq from Samuel Groner's measurements (THD over level with 600 ohm load).
Opamps well-suited to Class A bias are generally endowed with substantial output current ability at low Iq, and often rather fast parts of limited maximum open-loop gain. The technique was popularized in conjunction with the LM6172 by Jan Meier, I think, and possibly offers further benefits in these modern Class AB parts. I'd also give it a shot on TI's version of the MC33078 (which appears to be the most powerful of the bunch and shows high distortion in AB with low Iq), as well as the MC33178. NJM2068, may be worth a shot, possibly LM833 as well (in spite of rather limited output current). I would be interested in a list of well-suited parts as well.
The main drawbacks to the technique are reduced maximum current swing and adiitional power dissipation that heats up the part (+Vs * Ibias * #amps_per_package) and potentially degrades some of its other properties. Alternatively, you can also use a buffer to reduce output loading, if at the cost of higher parts count, some attention to stability and usually even higher current.
The point re: supply pollution certainly is a valid one. A heavily loaded opamp at high levels would be expected to require power amp-ish design techniques (separate signal and power ground, low-inductance rails etc.). That's part of what makes CFB parts so cranky, their output tends to be heavily loaded by default unless your circuit is a boring buffer.
Last edited:
Thanks for the info. If I Class-A bias the buffer chip prior to LT1363 at the output, then is the buffer forcing the LT1363 into Class-A as well?
I have done this with a portable amplifier and I haven't figured out yet why LT1363 sounds different when buffered, it's already stated as at least 50 mA per channel.
I have done this with a portable amplifier and I haven't figured out yet why LT1363 sounds different when buffered, it's already stated as at least 50 mA per channel.
Last edited:
op amp ccs
When I was looking into this a while back for a line level preamp I decided it would be a good idea to run two identical opamp circuits in parallel with a resistor between the outputs and purposely dc offset one of them, sounded good but didn't measure.
When I was looking into this a while back for a line level preamp I decided it would be a good idea to run two identical opamp circuits in parallel with a resistor between the outputs and purposely dc offset one of them, sounded good but didn't measure.
You mean a different IC altogether?
Thanks for the input, which I have not seen before, but LT1363 has many merits in it's design and I want to try it in Class-A.
I have tried OPA627 in Class-A which is a JFET but that is a different chip, different distortion spectrum, different speed and so forth.
if you are biasing the output of a input op amp which is only driving a buffer's Hi Z input then mostly you don't need to - even 100 uA internal op amp output stage bias is fine for 100s of kOhms, few pF
but if you only want a mA or 2 then many kOhm R from output to a rail could be good enough
but if you only want a mA or 2 then many kOhm R from output to a rail could be good enough
According to Douglas Self in his excellent Active Crossover book, just randomly biasing a given opamp to either +ve or -ve rail is hit and miss, and if your 50/50 gamble doesn't pay off and you bias to the wrong rail you are likely to be increasing THD rather than reducing it. This doesn't take into account potential changes in harmonic content by the elimination of crossover distortion, but hey either way I'd rather be reducing the THD than increasing it.
Self only considers the 5532/5534 as an example, and with these devices he shows a significant decrease in THD when the output is biased to the +ve rail only, with CCS giving better results than a simple resistor due to the increased output loading a resistor causes. He found about 5mA to be an optimum current.
I decided to try this in one of my preamps that uses two stages of 5534 per channel. I connected a CCS consisting of an LED and a BC547 to the +ve rail on each opamp output. The difference in sound shocked me, as I wasn't expecting it to be so huge. The top end "grain" is completely eliminated and the sound is now quite "sweet" and very enjoyable. To me the 5534 biased to +ve rail sounds better than an LM49710 (the only other opamp I can live with in my pre) without biasing. I'd like to try biasing the LM49710 but don't have the means to measure distortion at present, so would not know which rail will work best. If anyone has tried this and worked it out, please let me know!
Self only considers the 5532/5534 as an example, and with these devices he shows a significant decrease in THD when the output is biased to the +ve rail only, with CCS giving better results than a simple resistor due to the increased output loading a resistor causes. He found about 5mA to be an optimum current.
I decided to try this in one of my preamps that uses two stages of 5534 per channel. I connected a CCS consisting of an LED and a BC547 to the +ve rail on each opamp output. The difference in sound shocked me, as I wasn't expecting it to be so huge. The top end "grain" is completely eliminated and the sound is now quite "sweet" and very enjoyable. To me the 5534 biased to +ve rail sounds better than an LM49710 (the only other opamp I can live with in my pre) without biasing. I'd like to try biasing the LM49710 but don't have the means to measure distortion at present, so would not know which rail will work best. If anyone has tried this and worked it out, please let me know!
Generally the output stage is shorted to the rail to bypass the PNP device of an complementary EF output stage.
But if the output stage is a complementary CFP then the PNP is in the "top" half and its the top half that needs to be bypassed by taking Out to +ve rail.
If the output is a quasi complementary then the lower NPN is the one to be bypassed to the -ve rail.
It seems from this that we need to know what output stage is in the opamp that is being converted to ClassA single ended.
You're right, we can't guess, but we should be able to work out which rail if the manufacturer shows a simplified sch with the output stage correctly drawn.
I've had a quick look at Philips, Ti and ONsemi sch in each of their datasheets.
As far as I understand it, all show the same quasi output stage.
This would indicate that out to -ve rail via a CCS is the way to bypass the lower output device.
This seems to be at odds with D.Self's findings.
But if the output stage is a complementary CFP then the PNP is in the "top" half and its the top half that needs to be bypassed by taking Out to +ve rail.
If the output is a quasi complementary then the lower NPN is the one to be bypassed to the -ve rail.
It seems from this that we need to know what output stage is in the opamp that is being converted to ClassA single ended.
You're right, we can't guess, but we should be able to work out which rail if the manufacturer shows a simplified sch with the output stage correctly drawn.
I've had a quick look at Philips, Ti and ONsemi sch in each of their datasheets.
As far as I understand it, all show the same quasi output stage.
This would indicate that out to -ve rail via a CCS is the way to bypass the lower output device.
This seems to be at odds with D.Self's findings.
Last edited:
5534 ops isn't quasi. It's a single common-emitter NPN (on pin 5). Its collector is both the output (via a diode and limiting R) and level-shifted with a vbe mult which feeds an NPN EF to allow sourcing current from the output higher than the standing current, again with limiting R plus associated transistor.
With the 5534 you don't need an external current source, just connect a resistor from output to pin 5 to increase standing current. Pretty close to a current source because the voltage is (almost) constant. Tektronix did that trick alle the time.
With the 5534 you don't need an external current source, just connect a resistor from output to pin 5 to increase standing current. Pretty close to a current source because the voltage is (almost) constant. Tektronix did that trick alle the time.
Dare I say it, and presuming here that you accept DS's measurements, doesn't this just prove that:
a) your statement of fact is actually only a theory that might not be on the money, and
b) no amount of postulating will make up for actually getting off your backside and trying things out?
Even if your theory was 100% correct, schematics are not shown for any of the more modern opamps, so experiment and listening/measurement will still be needed.
That resistor to pin 5 on the 5534 sounds like a neat trick, but like with an R to rail it will be loading the output stage.
Discrete opamps still sound better than any IC opamp, biased or not. So there. Not that I'm generalising or anything.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.