Let's talk about SIMILAR Class A and Class B circuits and see what can be the difference, even without the math. Then, let us compare IC output stages, and the inclusion of an external resistor or current source to bias the Class A.
First, a similar comparison might be made with a complementary transistor follower output stage, with some sort of biasing between the 2 bases. IF we could get to this 'biasing' we could really make some IC's 'sing', but alas, we cannot.
These output stages, ever since the LM101 that most IC's, even today are patterned after, have been a class A-B, or more like B design. Today, it appears even more so, since OPERATING CURRENT is everything, and they try to get away with as little quiescent output current as possible.
IF you look at this stage, in simulation or reality, and you can increase the quiescent current to, let's say, 5 ma or so, you would be essentially class A throughout its normal operating region for audio, unless you loaded the design with 1K or less. In any case, push-pull Class A operation, CANCELS the even harmonic distortion, and as this is MOST of the output nonlinearity in each transistor, keeps the OPEN LOOP DISTORTION very low.
Now, what about Class B, with an 'ideal' crossover point? Now, the even harmonic distortion disappears, BUT it re-appears as 3'rd harmonic distortion. Some trade-off, huh? There is no real distortion cancellation, just harmonic up-conversion.
Now, what happens if the xover point is NOT ideal? Then, lots of extra harmonics are generated, AND the open loop gain is modulated, potentially creating PIM distortion. This is the normal case, with cheaper, offshore parts.
Now what happens when you add the external resistor or current source to the - supply for example. Then, you can get Class A operation, but NOT push-pull operation, so there is NO intrinsic distortion cancellation, and you get a lot more 2'nd harmonic than you used to have, BUT less 3'rd harmonic, and no added PIM. A good tradeoff, but not a perfect one.
First, a similar comparison might be made with a complementary transistor follower output stage, with some sort of biasing between the 2 bases. IF we could get to this 'biasing' we could really make some IC's 'sing', but alas, we cannot.
These output stages, ever since the LM101 that most IC's, even today are patterned after, have been a class A-B, or more like B design. Today, it appears even more so, since OPERATING CURRENT is everything, and they try to get away with as little quiescent output current as possible.
IF you look at this stage, in simulation or reality, and you can increase the quiescent current to, let's say, 5 ma or so, you would be essentially class A throughout its normal operating region for audio, unless you loaded the design with 1K or less. In any case, push-pull Class A operation, CANCELS the even harmonic distortion, and as this is MOST of the output nonlinearity in each transistor, keeps the OPEN LOOP DISTORTION very low.
Now, what about Class B, with an 'ideal' crossover point? Now, the even harmonic distortion disappears, BUT it re-appears as 3'rd harmonic distortion. Some trade-off, huh? There is no real distortion cancellation, just harmonic up-conversion.
Now, what happens if the xover point is NOT ideal? Then, lots of extra harmonics are generated, AND the open loop gain is modulated, potentially creating PIM distortion. This is the normal case, with cheaper, offshore parts.
Now what happens when you add the external resistor or current source to the - supply for example. Then, you can get Class A operation, but NOT push-pull operation, so there is NO intrinsic distortion cancellation, and you get a lot more 2'nd harmonic than you used to have, BUT less 3'rd harmonic, and no added PIM. A good tradeoff, but not a perfect one.
push-pull Class A does cost another op amp but it is entirely feasible, up to the op amp output current rating
in a dual on a single die the halves should match well for the even harmonic reduction (at least before feedback which may be more difficult to assure matching)
I highly advocate for some of the A/DSL driver op amps of the past decade for output in a composite op amp circuit
early on they competed on THD @ 1 MHz into 25 Ohms - some devices clearly use 5 ma or more output Q bias - they would drive that 1 K to +/-10 V in push-pull Class A individually, without extra tricks
in a dual on a single die the halves should match well for the even harmonic reduction (at least before feedback which may be more difficult to assure matching)
I highly advocate for some of the A/DSL driver op amps of the past decade for output in a composite op amp circuit
early on they competed on THD @ 1 MHz into 25 Ohms - some devices clearly use 5 ma or more output Q bias - they would drive that 1 K to +/-10 V in push-pull Class A individually, without extra tricks
Last edited:
few build 1 transistor RIAA preamplifiers, why should I be limited to 1 op amp?
I see no "rule" that discrete/op amp comparision has to be done against just 1 op amp for the circuit function
my point is that composite op amp circuits, using 2 or more op amps, each selected for its circuit position/role can greatly improve on the specs of any 1 op amp used for the same gain function
unity gain buffer is a poor choice of how to build a composite op amp - those paying attention will know Jung showed several ways to use gain from the high current output (preferably CFA) op amp in multiloop topologies
the circuit is still "op amp based" - more op amps are used for vastly improved overall performance
the high mid band gain and low impedance load of the low noise RIAA feedback network at high frequency make multiloop op amp circuits with gain in the CFA output op amp a good topology choice
I see no "rule" that discrete/op amp comparision has to be done against just 1 op amp for the circuit function
my point is that composite op amp circuits, using 2 or more op amps, each selected for its circuit position/role can greatly improve on the specs of any 1 op amp used for the same gain function
unity gain buffer is a poor choice of how to build a composite op amp - those paying attention will know Jung showed several ways to use gain from the high current output (preferably CFA) op amp in multiloop topologies
the circuit is still "op amp based" - more op amps are used for vastly improved overall performance
the high mid band gain and low impedance load of the low noise RIAA feedback network at high frequency make multiloop op amp circuits with gain in the CFA output op amp a good topology choice
Last edited:
For that matter, we could REALLY muddle the distinctions between "discrete" and "integrated" designs by connecting a pair of discrete transistors as the input stage of an IC opamp. I recall an Application Note (from somebody like Siliconix, Intersil or Teledyne Crystalonics circa 1980) that had a matched JFET pair wired into the compensation pins of a '5534 in such a way that the normal differential inputs were bypassed. As I recall, this increased speed and reduced noise at the expense of circuit complexity and offset voltage. (I can't seem to find the App Note right now, but perhaps somebody else remembers it.)
Regardless . . . would this be categorized as a "discrete JFET", or "IC opamp" design? Or, while we are quibbling over administrivial distinctions, should somebody be out making measurements and deciding whether the concept has merit?
Dale
Dale:
That was from the May 27 1981 Electronic Design News magazine "Composite Op-amp Outperforms FET Input IC's" by Tom Cale.
Walt Jung sent me the PDF last week. In his ADI Op-Amp Applications book, page 6.204, he talks about this with more modern parts.
These articles make me want to breadboard the Bugle with a discrete JFET front end. I just may do it to see how it performs.
Sorry I have not commented more, been a very busy weekend. I am interested in hearing more about the use of two push-pull op-amps for distortion reduction and the use of the ADSL drivers-any model numbers for me to read about?
I read up on the LME49710-this does look to be an excellent IC-may try it in the Bugle. It was interesting to see the IMD graph go vertical at 10v. If I use it, I had better turn the volume down!
I wonder what the open loop BW is?
Also starting to look for any subs for the 2sk170/2sj74 pair. Getting turned off on the speculation of these things.
Thanks for the discussion!
Dan
That was from the May 27 1981 Electronic Design News magazine "Composite Op-amp Outperforms FET Input IC's" by Tom Cale.
Walt Jung sent me the PDF last week. In his ADI Op-Amp Applications book, page 6.204, he talks about this with more modern parts.
These articles make me want to breadboard the Bugle with a discrete JFET front end. I just may do it to see how it performs.
Sorry I have not commented more, been a very busy weekend. I am interested in hearing more about the use of two push-pull op-amps for distortion reduction and the use of the ADSL drivers-any model numbers for me to read about?
I read up on the LME49710-this does look to be an excellent IC-may try it in the Bugle. It was interesting to see the IMD graph go vertical at 10v. If I use it, I had better turn the volume down!
I wonder what the open loop BW is?Also starting to look for any subs for the 2sk170/2sj74 pair. Getting turned off on the speculation of these things.
Thanks for the discussion!
Dan
Hello Dan!
Nice to see you here...
You have fallen through the RABBIT HOLE NOW! ha ha ha!
..-. -...
Anyhow, if you pull the Class B opamp into Class A, does one not loose some headroom?
Also potentially the discrete opamp has more headroom (higher rail voltages) and may also dump more current out...
Those newer ultra low distortion opamps CAN be very good, but I have heard them sound different from one another. It has been said that the cause may well be things happening *way* out of the audio band, but I have yet to detect that.
Also, RIAA is an art not a science. So I would not just assume that one RIAA circuit yields identical technical or sonic results as the next, they do not.
We've had extensive discussions on these sorts of topics, and it gets VERY FINE in the trenches... very very fine...
I'd suggest for light reading you read through the many THOUSANDS of posts in the John Curl Blowtorch Preamp I and II!! You'll learn to appreciate the 7th harmonic too...
Wait, this is not a RABBIT HOLE at all, this is a pit of Vipers!
Well, ur going to have fun now!
_-_-bear
PS. you can just buy some Tosh jfets in the Vendor's Bazzar here... jes fine.
Nice to see you here...
You have fallen through the RABBIT HOLE NOW! ha ha ha!
..-. -...
Anyhow, if you pull the Class B opamp into Class A, does one not loose some headroom?
Also potentially the discrete opamp has more headroom (higher rail voltages) and may also dump more current out...
Those newer ultra low distortion opamps CAN be very good, but I have heard them sound different from one another. It has been said that the cause may well be things happening *way* out of the audio band, but I have yet to detect that.
Also, RIAA is an art not a science. So I would not just assume that one RIAA circuit yields identical technical or sonic results as the next, they do not.
We've had extensive discussions on these sorts of topics, and it gets VERY FINE in the trenches... very very fine...
I'd suggest for light reading you read through the many THOUSANDS of posts in the John Curl Blowtorch Preamp I and II!!
You'll learn to appreciate the 7th harmonic too...Wait, this is not a RABBIT HOLE at all, this is a pit of Vipers!
Well, ur going to have fun now!
_-_-bear
PS. you can just buy some Tosh jfets in the Vendor's Bazzar here... jes fine.
Hi Bear!
I was wondering how soon you'd find me!
Yes you are right....this is a hole I will not be able to get out of!
So this does get deeper into our discussion years ago about sound versus measurements. I was very surprised at how the Borbely preamp sounded much clearer than my op-amp preamps. So this is the quest to discover technically why this is so. I have read much so far and learned some.
I have read reports about different IC op-amps sounding different. My AD797 sounds better than my 5534, but to me the 5534 does sound pretty good and has had a long life.
Since our discussion, I have learned that much of this sound may be due to IM products because of limited open loop bandwidth. There are other parameters affecting this.
Now sometime down the road, I should be able to bread-board a couple of circuits and measure them with my AP system 1. I can also then measure my older circuits and hopefully point to test numbers that help define a sound.
Did you know Erno well? As I was busy at school and work, I missed his closing by about 6 months. It would have been fun to chat with him. His articles reveal much of the technical reasoning behind his designs. This board shows me that people have progressed further.
Thanks,
Dan
I was wondering how soon you'd find me!
Yes you are right....this is a hole I will not be able to get out of!
So this does get deeper into our discussion years ago about sound versus measurements. I was very surprised at how the Borbely preamp sounded much clearer than my op-amp preamps. So this is the quest to discover technically why this is so. I have read much so far and learned some.
I have read reports about different IC op-amps sounding different. My AD797 sounds better than my 5534, but to me the 5534 does sound pretty good and has had a long life.
Since our discussion, I have learned that much of this sound may be due to IM products because of limited open loop bandwidth. There are other parameters affecting this.
Now sometime down the road, I should be able to bread-board a couple of circuits and measure them with my AP system 1. I can also then measure my older circuits and hopefully point to test numbers that help define a sound.
Did you know Erno well? As I was busy at school and work, I missed his closing by about 6 months. It would have been fun to chat with him. His articles reveal much of the technical reasoning behind his designs. This board shows me that people have progressed further.
Thanks,
Dan
One note about seriously high performance opamps like the LME. They do have high open loop bandwidth and cannot be treated like a 5534 or such. If you don't pay attention to lead lengths on bypass caps and generally good layout, they may misbehave. IMHO, that's why some people have had trouble just dropping them into PCBs of the past, that were designed for slower parts. I've had no trouble with hand wired Vector board, and don't use or like ground planes, but I take a lot of care with both lead lengths and the physical layout, particularly the avoidance of inadvertent positive feedback paths due to poor choice of ground connections and such. The parts have excellent drive capability, but you still have to watch for reactive loads and pad accordingly.
I see no reason why everything I've just said doesn't apply equally to fast discrete designs!
I see no reason why everything I've just said doesn't apply equally to fast discrete designs!
Hi Bear!
I was wondering how soon you'd find me!
Yes you are right....this is a hole I will not be able to get out of!
So this does get deeper into our discussion years ago about sound versus measurements. I was very surprised at how the Borbely preamp sounded much clearer than my op-amp preamps. So this is the quest to discover technically why this is so. I have read much so far and learned some.
I have read reports about different IC op-amps sounding different. My AD797 sounds better than my 5534, but to me the 5534 does sound pretty good and has had a long life.
Since our discussion, I have learned that much of this sound may be due to IM products because of limited open loop bandwidth. There are other parameters affecting this.
Now sometime down the road, I should be able to bread-board a couple of circuits and measure them with my AP system 1. I can also then measure my older circuits and hopefully point to test numbers that help define a sound.
Did you know Erno well? As I was busy at school and work, I missed his closing by about 6 months. It would have been fun to chat with him. His articles reveal much of the technical reasoning behind his designs. This board shows me that people have progressed further.
Thanks,
Dan
Hi Dan,
You are in trouble now.
Unfortunately there is no known metric that correlates what we hear with measurements - yet.
You might want to search for and read about D.E.L. Shorter (BBC) and more recently the "GedLee Metric" that sheds some light on the subject.
In light of that minor revelation, Dan I suggest to you that the AP unit you have is of marginal use. I want to help you. So, I can drive out to Boston in under 3 hours and get that AP unit out of your way. Think of what you could do with the extra space on that bench??
Email or PM me on Erno...?
Jacco, I would say it is an art because all audio is an art, but in the case of the RIAA preamp, even more so. Have you ever heard two phono preamps that sounded identical? I have yet to do so. Also while the RIAA curve is spec'd the implementation of it is nearly impossible to get exact, despite those who claim high precision - that effects all sorts of other sonic factors, imo. Ymmv
_-_-bear
Bear:
I upgraded the System 1 to do only 8 bit measurements.
I was gonna put it out to the curb, but I found the case was good for a 5 tube AC-DC radio.
So I bet at least the major sonic qualities have some measurement correlation. I do not believe in building something just based on adjetives.
On another note, I feel the J74 is obsolete in the fact it is not made anymore and manufacturers cannot directly source them.
I'll look up the TPA6120.
To get back to the basics here, I agree with John's idea of comparing apples to apples. Discrete Class A versus AB1 or B. I have been reading on that and need to read further.
Conrad: Yes I would guess that bypassing and grounding will help prevent oscillations, and I think i need to pay more attention to an RF style layout than an audio form. Does anyone know what the open loop bandwidth of the LME is? 10kc, 20kc?
As far as RIAA networks go, I am leaning toward separate passive EQ (buffered between active stages). From the Borbely, that may be one of the keys to its sonics.
Thanks!
Dan
I upgraded the System 1 to do only 8 bit measurements.
I was gonna put it out to the curb, but I found the case was good for a 5 tube AC-DC radio.
So I bet at least the major sonic qualities have some measurement correlation. I do not believe in building something just based on adjetives.
On another note, I feel the J74 is obsolete in the fact it is not made anymore and manufacturers cannot directly source them.
I'll look up the TPA6120.
To get back to the basics here, I agree with John's idea of comparing apples to apples. Discrete Class A versus AB1 or B. I have been reading on that and need to read further.
Conrad: Yes I would guess that bypassing and grounding will help prevent oscillations, and I think i need to pay more attention to an RF style layout than an audio form. Does anyone know what the open loop bandwidth of the LME is? 10kc, 20kc?
As far as RIAA networks go, I am leaning toward separate passive EQ (buffered between active stages). From the Borbely, that may be one of the keys to its sonics.
Thanks!
Dan
The LME49710 is a 55 MHz GBP part. The open loop gain is 140 dB. It's a killer part, but only if you treat it like the high speed device it is.
LME49710 - High Performance, High Fidelity Audio Operational Amplifier
LME49710 - High Performance, High Fidelity Audio Operational Amplifier
Hi Conrad!
Thanks for the link. I had read the LME datasheet over the weekend (and got really excited!), but did not see it's open loop bandwidth. So the closed loop bandwidth is 55Mhz. Gee I could use that for my shortwave receiver!
From what I read in a few papers, the bandwidth of an amplifier without feedback seems to affect the final IM of the amp with normal feedback. The higher the bandwidth, the lower the IM. Borbely was able to get something like a 10Khz open loop bandwidth before he applied feedback on his discrete stuff. The NE5534 has a GBP of 10Mhz. I also like the power bandwidth specification. If an op-amp cannot make at least 20Khz power bandwidth, I do not consider it. The NE5534 has a power bandwidth of 10Mhz, and an open loop bandwidth of 10 Khz.
The LME's open loop bandwidth I bet is much better, but wonder what it is....
Can I use it for my ham transmitter?
This may be the next IC I try.
I also looked at the TPA6120. Much better than the LM1877N I used in the 1980's (it was very quiet an clean).
Thanks,
Dan
Thanks for the link. I had read the LME datasheet over the weekend (and got really excited!), but did not see it's open loop bandwidth. So the closed loop bandwidth is 55Mhz. Gee I could use that for my shortwave receiver!
From what I read in a few papers, the bandwidth of an amplifier without feedback seems to affect the final IM of the amp with normal feedback. The higher the bandwidth, the lower the IM. Borbely was able to get something like a 10Khz open loop bandwidth before he applied feedback on his discrete stuff. The NE5534 has a GBP of 10Mhz. I also like the power bandwidth specification. If an op-amp cannot make at least 20Khz power bandwidth, I do not consider it. The NE5534 has a power bandwidth of 10Mhz, and an open loop bandwidth of 10 Khz.
The LME's open loop bandwidth I bet is much better, but wonder what it is....
Can I use it for my ham transmitter?
This may be the next IC I try.
I also looked at the TPA6120. Much better than the LM1877N I used in the 1980's (it was very quiet an clean).
Thanks,
Dan
Last edited:
I think this was talked about when the parts were first released- what you have to do is get the LM4562 data sheet, which is just the dual version. At the end of all those graphs, that one has the bode plot that the single version data sheet is missing.
Yes, I'm sure you could make a small household AM radio transmitter out of it.
Yes, I'm sure you could make a small household AM radio transmitter out of it.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.