for most audio amplifiers
normal two TO92 put together is performing more than well enough
as current mirrors
the precision will be in 1/10-s of one precent
selecting one reasonably matched pair
and eventually using two 100 Ohm emitter resistors
and in a typical Global-Feedback amplifier, you can't go wrong
looking for more precision
and you will have to consider all other parts in your amplifier
- Using 0.1% resistors instead of 1% or 5%
- Using 1% capacitors
- Regulate your Voltages of PSU to 0.1%
- Remove any ripple is PSU
In short:
Being paranoid about one mirror down to last decimal
won't get you anywhere .. unless you match everything else in your amplifier
to the same decimal
--------------------------------------------
People hang up on details of a system
and lose the overall understanding.
This is typical for many who lose their way, while staring at only one little spot.
This goes for even very clever audio designers, with half a century (50 years) experience.
Lineup
normal two TO92 put together is performing more than well enough
as current mirrors
the precision will be in 1/10-s of one precent
selecting one reasonably matched pair
and eventually using two 100 Ohm emitter resistors
and in a typical Global-Feedback amplifier, you can't go wrong
looking for more precision
and you will have to consider all other parts in your amplifier
- Using 0.1% resistors instead of 1% or 5%
- Using 1% capacitors
- Regulate your Voltages of PSU to 0.1%
- Remove any ripple is PSU
In short:
Being paranoid about one mirror down to last decimalwon't get you anywhere .. unless you match everything else in your amplifier
to the same decimal
--------------------------------------------
People hang up on details of a system
and lose the overall understanding.
This is typical for many who lose their way, while staring at only one little spot.
This goes for even very clever audio designers, with half a century (50 years) experience.
Lineup
lineup said:for most audio amplifiers
normal two TO92 put together is performing more than well enough
as current mirrors
the precision will be in 1/10-s of one precent
selecting one reasonably matched pair
and eventually using two 100 Ohm emitter resistors
and in a typical Global-Feedback amplifier, you can't go wrong
looking for more precision
and you will have to consider all other parts in your amplifier
- Using 0.1% resistors instead of 1% or 5%
- Using 1% capacitors
- Regulate your Voltages of PSU to 0.1%
- Remove any ripple is PSU
In short:
won't get you anywhere .. unless you match everything else in your amplifier
to the same decimal
--------------------------------------------
People hang up on details of a system
and lose the overall understanding.
This is typical for many who lose their way, while staring at only one little spot.
This goes for even very clever audio designers, with half a century (50 years) experience.
Lineup
:bs:
Current mirrors only work with exact equivalent transistors.
QSerraTico_Tico said:
Everything Lineup has said seems to be true, why do you insult and say it's BS?
I'm a rookie, a noob and I have used current mirrors with discrete transistors that I didn't match. They work just fine. Current and voltage through the "legs" of the LTP was very well matched.
I used a super-matched pair for differential pair before and it gave no better measured or sonic results over a discrete pair. A waste of money.
I found that the standard two transistor mirror never matched the output current to the input current even when the transistors were well matched.
Deliberately maladjusting one of the emitter resistors improved the current mirroring but only at the set point current and voltage.
I think a two transistor current mirror is incapable of good current matching over a range of voltages and currents.
Deliberately maladjusting one of the emitter resistors improved the current mirroring but only at the set point current and voltage.
I think a two transistor current mirror is incapable of good current matching over a range of voltages and currents.
AndrewT said:
Hi Andrew,
Do you see an improvement over the simple resistor source? Do you think a resistor or two will do a more accurate job?
I know you have a negative attitude toward the current mirror in general, believing it has a bad effect on sound quality.
I measured the voltage and current at idle, not during operation with a signal, so I can't comment any further.
Not much on how you find the sound of current mirrors, so I'll stick my chin out, and I'm expecting a couple of slaps.
I have limited experience from using current mirrors on the LTP, but I didn't like what I heard. The sound was just worse in every way - harsh, undfriendly, colored. Go on, hit me!
I have limited experience from using current mirrors on the LTP, but I didn't like what I heard. The sound was just worse in every way - harsh, undfriendly, colored. Go on, hit me!
I too am only capable of measuring mirror performance at the DC level and that's where I found it failed.MJL21193 said:
The three transistor version is far better, although all those years ago I was unaware of that extra resistor and have never tested what it is capable of.
I believe the current mirror if truly matching the mirrored currents is completely defeated in it's purpose by the unbalanced way that VAS is taken from one collector only.
This sourcing and sinking of current below the mirror into one half of the LTP guarantees that the LTP is operating with unequal currents as soon as any AC signal is processed.
Improving the mirror function can never cure that drawback.
Reducing the VAS load can help as proved by Self and others.
What is really needed is balanced operation on both sides of the LTP.
Symasym gets much closer to what is required and Roender has gone a bit further.
AndrewT said:
Symasym uses a second LTP with a current mirror as the VAS, no? Though both "legs" are loaded are they loaded equally? Simulation says no.
Even if yes, how much of an impact will this have on linearity and/or sound quality? This arrangement is known to suffer from stability problems.
Simulation shows that current from the LTP to an EF VAS is very tiny. A few uA typical. How much impact will this very small fluctuation have on the overall result?
There is an interesting simulation of current mirrors at
http://www.analog-innovations.com/SED/EnhancedCurrentMirrors.pdf .
Also has the Thompson current mirror which I have never seen used in audio.
Ray
http://www.analog-innovations.com/SED/EnhancedCurrentMirrors.pdf .
Also has the Thompson current mirror which I have never seen used in audio.
Ray
Expert level
Quote from the MAT03 datasheet:
"A fundamental requirement for accurate current mirrors and
active load stages is matched transistor components."
For further insight see the MAT03 datasheet:
http://www.analog.com/en/other/matched-transistors/mat03/products/product.html
In view of the expert level on this forum I no longer take part in the discussion.
MJL21193 said:
Quote from the MAT03 datasheet:
"A fundamental requirement for accurate current mirrors and
active load stages is matched transistor components."
For further insight see the MAT03 datasheet:
http://www.analog.com/en/other/matched-transistors/mat03/products/product.html
In view of the expert level on this forum I no longer take part in the discussion.
check wikipedia for types of mirrors and differences. Mostly higher output impedance and closer i matching in 3 trans mirrors. Current sources and mirrors are slightly different, and used differently: source in the LTP tails and VAS loads, mirrors as LTP loads (on there collecters). For VAS loads the extra impedance increases the gain by an order of mag. For LTP pairs the tail source makes most difference and dosnt seem to matter if its only two transistors. In the LTP load it ballances the load but some people think the extra noise the trans. add to the input cct. isnt worth it (Dr. Leach for one). matching is more important for the mirror but also matters for the sources. Look at your amp as stages and an improvement in a stage will be an improvement in your amp. At what % matching makes a diff not sure, but improveing the input stage dist. with mirror inthe LTP tail is pretty standard now.
Re: Expert level
Why don't you read the whole datasheet?
The MAT03 transistors are matched within +/- 3 %. You can get a lot better than that by manually matching single transistors. The advantages of matched duals, like the MAT03, are better thermal coupling and that in production they might in the end be cheaper than matching transistors manually.
QSerraTico_Tico said:
Quote from the MAT03 datasheet:
"A fundamental requirement for accurate current mirrors and
active load stages is matched transistor components."
For further insight see the MAT03 datasheet:
http://www.analog.com/en/other/matched-transistors/mat03/products/product.html
In view of the expert level on this forum I no longer take part in the discussion.
Why don't you read the whole datasheet?
The MAT03 transistors are matched within +/- 3 %. You can get a lot better than that by manually matching single transistors. The advantages of matched duals, like the MAT03, are better thermal coupling and that in production they might in the end be cheaper than matching transistors manually.
Tico,
Please don't leave, it is just getting intresting. In my opion matching is important but there are other parameters that can be just as or more important in a mirror, saying that matching is all important might be in error. As I stated earlier matching does not into account the Early effect and heating of the junctions that might cause the parameters of the transistors to change.
I hope this thread does not turn into a war, after all we are here to learn and all opions are welcome.
Regards,
Jam
Please don't leave, it is just getting intresting. In my opion matching is important but there are other parameters that can be just as or more important in a mirror, saying that matching is all important might be in error. As I stated earlier matching does not into account the Early effect and heating of the junctions that might cause the parameters of the transistors to change.
I hope this thread does not turn into a war, after all we are here to learn and all opions are welcome.
Regards,
Jam
Re: Expert level
I admitted that I'm not an expert. That puts me beneath you, I know. If you were to enter my world of expertise I'm sure I could muster a similar level of derision for your lack of knowledge.
You were probably not an expert at one time, or were you born to it?
QSerraTico_Tico said:
In view of the expert level on this forum I no longer take part in the discussion.
I admitted that I'm not an expert. That puts me beneath you, I know. If you were to enter my world of expertise I'm sure I could muster a similar level of derision for your lack of knowledge.
You were probably not an expert at one time, or were you born to it?
in LTPs its the current not the voltages thats important, idealy the current in the legs should be the same with zero input and equal 1/2 the current of the source at the LTP tail. With a load atached to one leg a non zero input changes the current ballance in the LTP trans. Since the curents in the mirror are the same the current difference in the LTP must come from the load. think of this LTP Mirror combo as a current pump. It will go from pumping all the mirror current into the load to draining the same amount of current into the LTP The advantages are 2 or was 4 times the drive into a capacitve load among others. Mirrors and i sources are valubale tools that can help cancel out transistor nonlinearities like in the above example which is very linear for small enough input voltages Its so hard (and long winded to explain this with words instead of equations) And the equations for this are pretty basic and easy to find on the net so go look