Current Mirror Discussion

Member
Joined 2011
Paid Member
... the helpered arrangement allows the operating voltage difference between the collectors of the current mirror to be very small. That, in turn, allows the use of diode clamps across those collectors, which can improve overall behavior in clipping, including some reduction in sticking. That also can reduce the peak current in the VAS helper transistor under clipping conditions.
Yep. BUT the diode clamps across the collectors steal some of the tail current during slewing, so the actual slew rate is less than Solomon's equation predicts (SR = Itail / Cc). In my experience the clamp diodes divert about 20% of the tail current away from Cc, so the observed slew rate is approx 80% of the theoretical value.
 
If we assume that each red LED has a noise level of ~0.3uV (from the 5mA data in Forr's link)
Then what is the noise level of three in series?

The 5.6V Zener is shown as roughly 2.9uV
How do we compare 3 LEDs to a Zener of similar voltage?

And in the location shown in posts327 & 338, do we need to consider noise, or is it irrelevant?
 
Last edited:
Member
Joined 2011
Paid Member
What's the transfer function when you apply an AC voltage on Q3's collector and measure the resulting AC current flowing in Q2's collector? Emitter followers do have quite high PSRR; perhaps the effect is completely negligible.

I predict you'll find (Iout/Vin) is teeny tiny.

_
 

Attachments

  • schema2.png
    schema2.png
    59.5 KB · Views: 451
  • prediction.zip
    223 bytes · Views: 51
If we assume that each red LED has a noise level of ~0.3uV (from the 5mA data in Forr's link)
Then what is the noise level of three in series?

The 5.6V Zener is shown as roughly 2.9uV
How do we compare 3 LEDs to a Zener of similar voltage?

And in the location shown in posts327 & 338, do we need to consider noise, or is it irrelevant?

Bear in mind that the Zener diode is bypassed; that will reduce the noise a great amount anyway. The corner frequency of the resulting low-pass filter should be at about where the dynamic impedance of the Zener equals the reactance of the bypass capacitor.

The noise should not matter much in this particular application, but if the resulting reference voltage is used as part of a current source (such as an LTP current source), it might matter more.

Cheers,
Bob
 
If so, then it obviates setting up a separate biasing network for the helpers. . .
If the voltage/amplitude is any different, the separate resistors are required for ballast, to soak up the differences. Well, else there's an effects box.
Also, several different inputs to amplifiers (gain); well, that also does require minimizing differences, else, yet more effects box results.
I'm claiming that you'd need more board space and not much more than 18 cents more resistors.
If it is a big deal then the topology has probably missed the mark; but, otherwise, a couple of resistors is probably okay.
 
How do we compare 3 LEDs to a Zener of similar voltage?
The noise level can be considered entirely on device capacitance (which is 3x lower in your example).
However, if the example had gone out to a much grander extreme then we'd also have to consider current/loading variance.

It is the very same thing if a wee little, low capacitance, low noise, amplifier had been used to blow off some drywall and knock down some ceilings, because had it been cranked up that far then the linearity would not suit the application. It is at that point when higher capacitance and necessarily higher noise devices could have done the job more suitably.

I would like the easy answer that higher capacitance (aka higher noise) devices aren't suitable until you needed them.
But, that's not quite it.
The higher capacitance (aka more linear) devices require less ballast current bias, work better in a broader variety of applications and are more durable.
That is approximately how the minimum noise / minimum capacitance device, never hits the mark relevantly.

It is possible to hit the mark with the low capacitance / low noise, device even in a real application. That involves running the device tandem/parallel. To absorb manufacturing variety, that also includes resistance, such as base stoppers (and more other language) to absorb the differences. Having your cake and eating it too, in that way, does take more board space.

Well, then there's this: If you wanted to spare the board space for a couple of resistors, then it will cost far more board space and circuitry to make up for the omissions.

P.S.
I'm not disagreeing with AndrewT. Instead, I wanted to explore the same except with different language.
 
Last edited:
Hello Alex

Have you done any updates or improvements in your method of balancing the ltp ?

Thank

Bye

Gaetan

homemodder.

I believe that it isn't the added diode in one method, that makes an improvement due to any effect on the CM, but due to the added voltage drop in the unloaded side that brings the currents through both sides of the LTP much closer.This will also be true with many CFP input designs. Incidentally, the reason I now use the Series Schottky diode plus the 20T trimpot, is that the vertically mounted Schottky diode thermally tracks the VAS, which in a preamp or Class AB, enables maintenance of a +-2mV difference between the collectors of the closely matched LTP devices. (yes, you can hear the difference with this degree of matching) Even with a 15W /Ch. Class A design ,with a total dissipation of around 100W, the tracking will stay within +-5mV if carefully adjusted when warm.
"Blameless" designs, with all due respect to the very talented Douglas Self, are often reported to sound bland, and with a small soundstage. Balancing the front end as described, (and I am sure that there are also even better ways to do this) will transform a "Blameless" design into an amplifier with a truly excellent soundstage, and with heaps of separation between the vocalists and instruments. IMO, Douglas Self has of course done all the hard work, by getting the Blameless to such a state of perfection, that these changes are readily apparent.

SandyK
 
Member
Joined 2011
Paid Member
It would be easy to test. Just install the super duper balancing magical circuitry, and then connect a 25-turn, 5 megohm trimmer potentiometer (link) across one of the two halves (but not the other). Now you can dial in as much or as little imbalance as you want, and see what are the sonic results, if any.
 
Thanks, but you're still completely wrong. The only criteria necessary:

"Groupthink is a type of thought exhibited by group members who try to minimize conflict and reach consensus without critically testing, analyzing, and evaluating ideas."

And no “critically testing, analyzing, and evaluating ideas” does not include making a current mirror modification that may yield a 0.00001% linearity improvement to a 0.05% amplifier and then reporting a hugely improved “soundstage” and much better imaging, in agreement with some other guys subjective evaluation.

Now that almost 13 years have passed, I really hope you gave up the majestic SPICE simulations and tested the real world by soldering the 1N4148 diode to the current mirror, as AKSA suggested in the first post on the thread.

Critical testing is not simulation, THD is somewhat irrelevant when a standard speaker distorts 1%.

IMD is also more or less irrelevant since "the ear canal's inherent IMD is around 1%, but IMD inside the cochlea is around 7%". Psychoacoustic/Ear-Related IMD | Audio Science Review (ASR) Forum

Do you still think that SPICE THD simulations worth anything when it comes to audio reproduction?

I think, you are using the simulator incorrectly because this is how to use it: :)
Memory Distortion Philosophies

BTW, I think one of the characteristics of bad hearing is the long term believe that the 'Blameless' amplifier sounds good. :D

Douglas Self in his AES 2015 WARSAW presentation finally admit that thermal distortion exists in discrete amplifiers after denying it for at least one decade. In the end of his presentation proposing that it's possible that there is a thermal distortion in the input differential amplifier. [Distortion D13 ??]
Progress In Power Amplifier Design - ppt video online download

:D Possible progress of Mr. Douglas Self:
in 2025: D14 Thermal distiortion of non cascoded simple BJT CCS
in 2030: D15 Heavy non-linear distortions caused by shitloads of unnecessary electrolytics in the circuit
in 2037: D16 Thermal distortion of simple emitter degenerated current mirror [use 4 transistor improved Wilson mirror instead]
 
Last edited:
Douglas Self in his AES 2015 WARSAW presentation finally admit that thermal distortion exists in discrete amplifiers after denying it for at least one decade. In the end of his presentation proposing that it's possible that there is a thermal distortion in the input differential amplifier. [Distortion D13 ??]


You confuse negligible with non-existent - you can only see the mysterious LF distortion with a very very good baseline amp and measurement system driven at full power and its clearly not just a simple thermal mechanism due to the roughly 2dB/octave slope (temperature deltas over a cycle are going to be 6dB/octave in nature). At levels of less than 0.001% at 10Hz/100W, its an elusive beast to study. Its certainly not something to stress about, its still negligible.
 
You confuse negligible with non-existent - you can only see the mysterious LF distortion with a very very good baseline amp and measurement system driven at full power and its clearly not just a simple thermal mechanism due to the roughly 2dB/octave slope (temperature deltas over a cycle are going to be 6dB/octave in nature). At levels of less than 0.001% at 10Hz/100W, its an elusive beast to study. Its certainly not something to stress about, its still negligible.

Yes, you're right: 'its clearly not just a simple thermal mechanism'

However, I'm not confusing negligible with non-existent. Even the principle of Douglas Self's measurement is inherently incorrect. That's basically not the way to look for these kinds of distortion. So just because memory type distortion can't be detected by THD measurement, that distortion is not negligible. As I described in the previous post, THD measurement (and simulation) or spectrum analysis is not worth much.

The THD vs Frequency curve will never show memory distortion. With false measurements, it is easy to say that something does not exist or is negligible. While the reality is that the problem is very serious, it's just this measurement that doesn't show it. Memory distorion is more than simply the thermal distortion.

A measurement principle worth to follow: Document sans titre

Music has constanly changing spectrum. Sine has a constant spectrum. Two sine (IMD measurement) is a constant spectrum signal. DIM, TIM are also constant spectrum signals.

With listening tests one can differentiate between good and bad sound amplifiers while by using only standard measurements one cannot.

Please see the attached image. In reality, when listening to music, we 'examine' the black boxed area shown in the attached image with our ears. On the other hand, for THD, IMD, DIM, TIM, SINAD measurements, we measure the green boxed range where there is no more memory distortion. In this way, there will never be a correlation between listening and measuring. Something else needs to be measured. The linked french article is an excellent example for alternative measurement that correlates with listening.
 

Attachments

  • n.png
    n.png
    92.2 KB · Views: 249
Last edited:
You are now conflating general memory distortion with LF thermal distortion, which is a specific subclass of memory distortion that is possible to characterize using THD.

Yes memory distortion is often swept under the carpet, partly because there are few standards for how to measure it meaningfully - many effects of memory distortion are outside the audio band, basically slow recovery graphs after the load is removed are usually effectively DC and irrelevant to audio amplifier performance. You can measure it if you want. A simple DC servo may remove this effectively.

Throwing of bias point during heating up and cooling down is a much more important memory effect and many people spend a lot of time trying to improve bias stability for this reason - if the amp gets tricked into under-bias by changes of loading then the thermal management needs improving. There's a whole chapter about this in D Self's book on amp design.


With listening tests one can differentiate between good and bad sound amplifiers while by using only standard measurements one cannot.
That's an assertion you want to be true. Have you tried to prove it? Or are you saying simply that some people don't test THD during load fluctuations - which, yes, is an omission.