It was king of a joke, on my side, keantoken. Using an emoticon always seems to me like an insult to my interlocutor ;-)
Just re-read your two sentences:
"There is no mechanism which subtracts the common mode error of the LTP ."
"Inverting mode simply has the advantage that it doesn't generate common mode distortion in the LTP."
when you wanted to explain that inverting mode was not the solution ?
This said, i can imagine some solutions to lower the distortion of the input stage including the fact that LTP is not the one and unique way.
Last edited:
The fact that inverting mode _usually_ has lower HF distortion and noting that the Otalla
amp was inverting is what started this conversation with andy_c years ago:
Thoughts Concerning Cordell, Otala, and Gilbert papers
I wrote this in post #16 in that thread back in 2005:
"We did a lab in college showing some of the non-ideal characteristics of OP amps. We measured slew rate and power bandwidth by the first sign of distortion and triangle method and were told to take note of the difference between inverting and non-inverting. We looked at the 741, 748, 351 and others. There was also a popular OP amp that "snapped" to the other rail on clipping, can't remember which one that is? Made a lasting impression contrasting inverting and non-inverting cases.
I first read about the issues with non-zero common-mode inputs in an article by Eric F. Taylor Wireless World, 1977. Edit: I mentioned the August issue in the first post here, there was a second part in the September issue. Taylor states that the varying Vcb with a common mode signal the diff amp input impedance is non-linear. I believe that this is due to the non-linear Miller capacitance.
Edit: We should note that this means the performance of the amp will depend on the driving source impedance."
amp was inverting is what started this conversation with andy_c years ago:
Thoughts Concerning Cordell, Otala, and Gilbert papers
I wrote this in post #16 in that thread back in 2005:
"We did a lab in college showing some of the non-ideal characteristics of OP amps. We measured slew rate and power bandwidth by the first sign of distortion and triangle method and were told to take note of the difference between inverting and non-inverting. We looked at the 741, 748, 351 and others. There was also a popular OP amp that "snapped" to the other rail on clipping, can't remember which one that is? Made a lasting impression contrasting inverting and non-inverting cases.
I first read about the issues with non-zero common-mode inputs in an article by Eric F. Taylor Wireless World, 1977. Edit: I mentioned the August issue in the first post here, there was a second part in the September issue. Taylor states that the varying Vcb with a common mode signal the diff amp input impedance is non-linear. I believe that this is due to the non-linear Miller capacitance.
Edit: We should note that this means the performance of the amp will depend on the driving source impedance."
Last edited:
Hmm. I meant subtracts literally. There is nothing in a conventional amp that subtracts the common mode error from the LTP, in the same way that the feedback loop of said amp subtracts error from the input (or fails to do so). If you wanted you could add a feedback loop around the LTP to do this, but inverting mode isn't how it is done (it does work, but for a different reason).
The second sentence you refer to taken by itself might be mistaken as a recommendation for inverting mode, but grammatically it is not and it would be bias on the part of the reader to think so (or just bad reading comprehension).
In the first line of the previous post I described the compromise that usually makes inverting mode impractical. If the reader were attentive he would know that it is the dominant statement.
Against someone who abuses my writing in these ways, I have no defense. I try to write for those who are most likely to hear my message. It's impractical to write for people who want to think you meant something different from what you actually said. I will happily skip that audience if I can help someone who has similar goals to me.
The second sentence you refer to taken by itself might be mistaken as a recommendation for inverting mode, but grammatically it is not and it would be bias on the part of the reader to think so (or just bad reading comprehension).
In the first line of the previous post I described the compromise that usually makes inverting mode impractical. If the reader were attentive he would know that it is the dominant statement.
Against someone who abuses my writing in these ways, I have no defense. I try to write for those who are most likely to hear my message. It's impractical to write for people who want to think you meant something different from what you actually said. I will happily skip that audience if I can help someone who has similar goals to me.
Absolutely. Reason why some people found some advantages in CFA configuration.
Is that kind of performance good enough ?
Functional Block Diagram for LM4562
That we could use in two different ways in an inverting power amp:
1- input buffer to reduce the input impedance.
2- Comparing output signal and input signal of the amp to extract distortions and create an extra feedback loop. Tried-it, found- it is a little too tricky ;-).
It is amazing how the active devices we can find are never fast enough, when we consider phase shift in Audio.
keantoken, please, don't take this too seriously. As I said, I was joking. Sorry you don't understand what i found funny. On my side, I'm not interested by distortion numbers under -100dB (0.001%) at 1KHz as I'm unable to hear-it. Trying to improve-it is just for fun.
Hi Bob:
Over the years, you've made comments about things you might/would do differently for the driver stage of your 50w MOSFET amp.
(For another example, the exchange between you and John Curl when he challenged you to provide balanced inputs
)
Just wondering if you have a pointer to a single schematic that has collected your current thoughts on this?
thanks!
mlloyd1
Over the years, you've made comments about things you might/would do differently for the driver stage of your 50w MOSFET amp.
(For another example, the exchange between you and John Curl when he challenged you to provide balanced inputs
)Just wondering if you have a pointer to a single schematic that has collected your current thoughts on this?
thanks!
mlloyd1
This is most certainly true. However, the extent of this effect in practice depends on the particular circuit and the numbers. In a power amplifier without a cascoded input LTP, the input transistors have a high reverse bias collector-to-base, often on the order of the rail voltage. This strongly depletes the collector-base junction and minimizes Ccb. At the same time, the modest input voltage (and hence common-mode voltage), on the order of a volt, is small in comparison to the depleting Vcb, making the percentage changes in Ccb fairly small. This also makes the percentage changes in beta due to Early effect small, so input current modulation is also fairly small. Similarly, the tail current source, if properly implemented, will see a small signal-dependent change in Vcb in comparison to the voltage depleting its collector-base junction.
The point here is that these effects certainly do exist, but, depending upon the details of the circuit, may be smaller than one might expect. The Devil is usually in the details.
I would also point out that op amps tested in non-inverting mode usually are often tested at unity gain with substantial signal amplitudes, while the amount of voltage depleting the input stage collector-base junctions may not be all that much greater in comparison. This is a recipe for much stronger common-mode effects.
Cheers,
Bob
I was going to try to help out Kean and explain what you just wrote about non-inverting
OP amps - now I don't have to.
Let me just state outright for people who might be new to the discussion that for the unity
gain non-inverting case that the base voltage on both sides of the diff pair obviously follow
the output voltage and this is where there is a large Vcb change in most common OP amp
topologies, even with a fixed bias cascode. This is when there is a high common mode
base signal on the diff pair.
In contrast, the non-inverting case has the positive input at ground and this creates, for a
good OP amp, a virtual ground at the negative input and in this case, ideally, there is no
voltage signal on the base of the diff pair and therefore little if any Vcb change.
Yes, of course it is application dependent.
The difference that we saw in the lab was shocking but it was unity gain, with perhaps 5V out
and early OP amp varieties.
OP amps - now I don't have to.
Let me just state outright for people who might be new to the discussion that for the unity
gain non-inverting case that the base voltage on both sides of the diff pair obviously follow
the output voltage and this is where there is a large Vcb change in most common OP amp
topologies, even with a fixed bias cascode. This is when there is a high common mode
base signal on the diff pair.
In contrast, the non-inverting case has the positive input at ground and this creates, for a
good OP amp, a virtual ground at the negative input and in this case, ideally, there is no
voltage signal on the base of the diff pair and therefore little if any Vcb change.
Yes, of course it is application dependent.
The difference that we saw in the lab was shocking but it was unity gain, with perhaps 5V out
and early OP amp varieties.
Last edited:
Corrected.
Just scanned my posts in case I also made the same mistake, it's been known to happen.
I'm not meaning to take anything too seriously, BUT, I do like to give everything it's due. Some people read that as me getting riled up, but I just like to end things neatly once they have been started.
This discussion has pointed out also that DUT should be conducted under the conditions it is being used. Mfr spec have standard arrangements and data applies only to those arrangements for the tests.
In each person's home is usually a preamp with a particular output Z vs freq. When I am balancing the PA diff IPS, I have it connected to the preamp or source I will be using that amp with. I drive the PA thru that particular preamp and measure the CMR and adjust/trim PA IPS bal for best CMRR etc.
Take note please that many PRO equipment allows for this sort of bal adjustment... connected to the other equipment and then balanced adjusted for LF and HF bal.
I keep saying this..... we listen to the entire system connected together and not one piece of it in isolation at a time. So it should be tested as a system as much as possible.
BTW - this affect of Zo interacting with PA IPS is one of several sources of audible variation from home to home... reviewer to reviewer.
THx-RNMarsh
In each person's home is usually a preamp with a particular output Z vs freq. When I am balancing the PA diff IPS, I have it connected to the preamp or source I will be using that amp with. I drive the PA thru that particular preamp and measure the CMR and adjust/trim PA IPS bal for best CMRR etc.
Take note please that many PRO equipment allows for this sort of bal adjustment... connected to the other equipment and then balanced adjusted for LF and HF bal.
I keep saying this..... we listen to the entire system connected together and not one piece of it in isolation at a time. So it should be tested as a system as much as possible.
BTW - this affect of Zo interacting with PA IPS is one of several sources of audible variation from home to home... reviewer to reviewer.
THx-RNMarsh
Last edited:
Corrected.
Just scanned my posts in case I also made the same mistake, it's been known to happen.
I'm not meaning to take anything too seriously, BUT, I do like to give everything it's due. Some people read that as me getting riled up, but I just like to end things neatly once they have been started.
Thank you, I'm getting old.
Are there any really good, complete models of a loudspeaker to use as load for SIM of PA's? Including motional back emf ?
I have seen the generic LRC models.
Alternative is to test distortions of PA with the real loudspeaker it is going to be used with (within power limits of the speaker system, of course).
THx-RNMarsh
I have seen the generic LRC models.
Alternative is to test distortions of PA with the real loudspeaker it is going to be used with (within power limits of the speaker system, of course).
THx-RNMarsh
Last edited:
Back EMF is just a way of saying that motion and resonance on the mechanical circuit side
of the motor is reflected back into the primary. If you "pull" the mechanical components
through to the electrical side it will behave the same way ignoring non-linearities. Simply
matching the input impedance with an electrical circuit will also get you the same thing.
Ken Kantor had a suggested load circuit, but obviously there are as many as there
are speakers:
Real-Life Measurements | Stereophile.com
The amplifier circuit you showed as exhibiting input current distortion, Figure 6.19 in APAD 6, shows the shared IPS and VAS current sources I pointed out as being a bad idea. That was the primary thing I was commenting on. However, virtually all of your amplifier examples use a current source where the reference voltage is shared among the IPS and VAS current sources, with either the IPS or the VAS current source having the master transistor that provides the feedback and make the transistor that it works with into a feedback current source. This includes the amplifier of Figure 15.1 in the slew rate chapter. The load-invariant power amplifier in Figure 10.12 also employs a feedback current source for the VAS, where that reference voltage is also used by the LTP current source transistor. However, at least there, you placed some R-C decoupling in the path of the reference voltage back to the LTP current source, presumably preventing the feedback of signal and distortion back to the LTP current source that I complained about and wherefrom significant input current distortion originated.
There is no discussion of curing any problems by using independent current sources in the slew rate chapter. There you bemoan the loss of slew rate due to coupling of steep rising edges into the base of the VAS current source through Ccb of the VAS current source transistor. Had you not shared the bias line there from the LTP current source, and not introduced a 2.2k isolating resistor in the base of the VAS to make troubleshooting of the shared arrangement easier (lipstick on a pig), that problem would not have occurred if you simply used enough current flowing into the control transistor of an independent VAS feedback current source. None of these bandaids is needed if you just do it right in the first place by using independent current sources.
In Figure 6.22, you cascode the LTP current source to improve PSRR (that you lost by not using an independent feedback current source), but that does not do anything to mitigate the tail current distortion caused by feedback from the VAS via the shared bias line.
Here is how the problem of the shared bias line works when the VAS uses a feedback current source and the LTP current source hitches a ride off of the same bias line. The VAS current source transistor has Early effect that wants to reduce its output impedance and cause distortion. The use of the feedback current source almost completely eliminates this problem by having the control transistor control the base drive of the slave current source transistor in such a way as to counteract those Early effect consequences that would otherwise occur.
Herein lies the mechanism. In order for that correction to take place, the base drive of the slave transistor must have a distorted version of the VAS collector signal on it. Any SPICE simulation will show that, even given the fairly simplistic modeling of Early effect by SPICE. When that bias line (the base of the VAS slave current source transistor) is also used by the tail current source, the tail current will be modulated by that correction signal, propagating that distortion into the input current via the finite beta of the input stage transistors. This is obviously an unwanted interaction between the VAS and the IPS. One additional cheap transistor and one resistor solves the whole problem by implementing two independent feedback current sources for the tail current and VAS pull-up current. This is an example of KISS - don't try to be cute by sharing a bias line between the IPS and the VAS. Simple as that.
Cheers,
Bob
I've seen similar ---- all SIM/CAD should pick one and use for model load. nice if the non-linearity portion could be added though. get better results in SIM of power amp.
The preamps -- many have a 'build-out' or series R added on their output. Often to limit output current of preamp to safe level. Tube preamps do this most often. AR in one preamp I had used 4.7K. That will mess up an other wise nice CMR of the PA. Not too quiet and helped it to be susceptible to RFi/EMI pickup on attached cable.
THx-RNMarsh
Last edited:
Good point about pre-amp build-out resistance. In general, I believe that most solid state preamps have a build-out resistor typically something like 100 ohms to isolate the output stage, be it an emitter follower or op amp, from capacitive loading effects of the interconnect to insure HF stability.
Some designs may use values more like 50 ohms or 75 ohms, sometimes even in an attempt to make a compromise impedance match to the characteristic impedance of typical interconnect (even terminating a transmission line at only one end in its characteristic impedance can be a good thing - whether this matters much at audio frequencies is debatable, though - we should be mindful that the interconnect is indeed a transmission line at EMI frequencies, so EMI susceptibility might be affected in some way).
In other cases, a designer might try to build out to a "standard" impedance, like 600 ohms.
Cheers,
Bob
Some designs may use values more like 50 ohms or 75 ohms, sometimes even in an attempt to make a compromise impedance match to the characteristic impedance of typical interconnect (even terminating a transmission line at only one end in its characteristic impedance can be a good thing - whether this matters much at audio frequencies is debatable, though - we should be mindful that the interconnect is indeed a transmission line at EMI frequencies, so EMI susceptibility might be affected in some way).
In other cases, a designer might try to build out to a "standard" impedance, like 600 ohms.
Cheers,
Bob
Quite. Though I would mention that the Blameless configuration/philosophy achieves very low distortion without tracking cascodes or whatever.
The great snag with inverting-mode is that if the input becomes disconnected the closed-loop gain drops to unity which usually means massive oscillation and general mayhem. A permanently wired-in input buffer is always required, which might be regarded as at best inelegant.
I would also point out that the measurements in APAD 6th edition (p140) show that common-mode input-stage distortion is very low indeed even in non-inverting mode. Too see common-mode distortion you need a power amplifier with a very low closed-loop gain, such as 1.5 times, which puts a big common-mode voltage on the input stage.