you need mixed PNP and NPN devices for CFP designs - that would not be possible on a single chip.
I think one of the reason chips are so common using quasi is because true PNP is difficult to manufacture in a chip?
Two BJT? Unlike Rod, Wiki doesn't mention if the pair is of the same sex (e.g. both PNP or both NPN) or mixed? When the same sex it is darlington, when mixed it is compound (CFP), then when we mix the pairs we call it COMPLEMNTARY. If the complementary is NOT of the same type (e.g. darlington on the top and compound/CFP at the bottom) we call it QUASI complementary. That makes sense and easy to remember/understand.
Not being able to make both npn and pnp on the same die is a good guess. But...
Here is a blast from the past:
741 construction
I've still got some in my parts box. Too sentimental to throw them away.
Anyhow, these had npn and pnp on the same die.
Also, notice how much area the Miller cap takes up. Chip makers don't like big capacitors.
Here is a blast from the past:
741 construction
I've still got some in my parts box. Too sentimental to throw them away.
Anyhow, these had npn and pnp on the same die.
Also, notice how much area the Miller cap takes up. Chip makers don't like big capacitors.
Attachments
Last edited:
Not being able to make both npn and pnp on the same die is a good guess. But...
Here is a blast from the past:
741 construction
I've still got some in my parts box. Too sentimental to throw them away.
Anyhow, these had npn and pnp on the same die.
Also, notice how much area the Miller cap takes up. Chip makers don't like big capacitors.
Unfortunately, the laws of physics cannot be cheated and a capacity as well as an inductor, I still remain unquestionably linked to the physical dimensions.
More likely that PNP is 3 times less performant due to lower hole mobilities, so NPN is preferred use of valuable Si real-estate. If you cracked open a complementary pair of transistors you'd likely find the PNP die size is larger to get similar current rating.
Well its my take given the ratio of electron to hole mobilities.
Re: PNP devices in uA, LM741 etc.
Understanding silicon circuits: inside the ubiquitous 741 op amp
Have a look at the illustrations and comments on the subject of PNP transistor fabrication in the above article. It certainly is done but you can't assume that the technique is also compatible with power transistor fabrication.
On the other hand, Philips made complementary chipamps such as TDA1514. They proved unreliable but they did work, though I know nothing about the chip construction which could well have been radically different to LM3886 quasi chips.
Understanding silicon circuits: inside the ubiquitous 741 op amp
Have a look at the illustrations and comments on the subject of PNP transistor fabrication in the above article. It certainly is done but you can't assume that the technique is also compatible with power transistor fabrication.
On the other hand, Philips made complementary chipamps such as TDA1514. They proved unreliable but they did work, though I know nothing about the chip construction which could well have been radically different to LM3886 quasi chips.
Why npn are preferred
I don't think this a strong reason not to make pseudo-darlingtons. "All pnp" darlingtons are available to buy as well as all npn. There are other devices that contain npn and pnp on a single die such as thyristors and triacs, although uniquely structured.
I don't think this a strong reason not to make pseudo-darlingtons. "All pnp" darlingtons are available to buy as well as all npn. There are other devices that contain npn and pnp on a single die such as thyristors and triacs, although uniquely structured.
I based my simplified answer on the likely construction of PNP darlingtons and problems of mixing them on NPN power transistor dies. These would be heating effects and extra processing, so I might be in some agreement with Adrian Popa's answer 2 in traderbam's linked article.
.
For many years, Sanken have made heaps of NPN and PNP Darlington audio and industrial grade devices with thermal tracking diodes too, so the Darlington construction is or at least was good and profitable for audio but the extra complexity and heating issues of the CFP variants probably is a problem and one best avoided. For example, how do you effectively control the bias current when the power transistor and driver are thermally integrated? As latala suggested, I have a feeling that would be unstable or just unworkable in power devices but no evidence to show yet
.
Last edited:
With a Darlington you have the turn-on voltage set by two junctions. With CFP, one junction. A minor difference.
You could argue the latter is better because it can be compensated for with one external diode rather than two. On the other hand, in a push-pull output stage the Darlingtons give about 2.4V across the Vbe multiplier and the CFPs only 1.2V, which is more constraining. This may be why the 5i is missing that 27ohm resistor.
How about outside of audio? What sort of circuit is the single package Darlington intended to cost-reduce?
You could argue the latter is better because it can be compensated for with one external diode rather than two. On the other hand, in a push-pull output stage the Darlingtons give about 2.4V across the Vbe multiplier and the CFPs only 1.2V, which is more constraining. This may be why the 5i is missing that 27ohm resistor.
How about outside of audio? What sort of circuit is the single package Darlington intended to cost-reduce?
Last edited:
The 27R collector resistor in the NAP250 design, I believe, is used to fine-tune the tempco of the bias generator. For CFP output stages, it may be a low value, say 22R and for EF types anything up to maybe 180R. In a quasi stage, the 27R value may be just right in the shoebox case but perhaps the shallow Nait box is a different story.
It's also possible that the designer didn't need to add the resistor for adequate tracking of the output stage with the devices and thermal coupling used in the 5i.
It's also possible that the designer didn't need to add the resistor for adequate tracking of the output stage with the devices and thermal coupling used in the 5i.
Audio is usually the last thing semiconductor manufacturers care about. The money is in control systems. Motor drivers, activators, relays, PWM controllers, regulators and so on. In data sheets they tend to say "intended for general purpose and low speed switching".
The Darlington transistor encapsulates resistors and a back-emf diode to replace the parts usually needed in discrete circuits of these types. Often driven directly by logic chips, like TTL. Being able to turn the darlington on and off without additional parts is desirable.
The Darlington transistor encapsulates resistors and a back-emf diode to replace the parts usually needed in discrete circuits of these types. Often driven directly by logic chips, like TTL. Being able to turn the darlington on and off without additional parts is desirable.
Problem with packaged darlingtons is that the current sharing between driver and output is already defined inside the package, and this sharing is not intended/suitable for audio. There are some Sanken darlingtons for audio where the internal resistor is more suitable for audio but for 'high-end' sound this bias seems to be still too small. I don't know how far the sound achievement could be when the Vbe multiplier is critically designed to handle the temperature issue and the packaged darlington is chosen for the good sounding one (old TO-3 packages, not the one from Sankens).
Member
Joined 2009
Paid Member
Let's be clear about what the term "complementary Darlington" refers to here. Systematically, a Darlington transistor is an integrated EF or emitter-follower pair, fabricated on a single die. They come in complementary all-NPN or all-PNP flavours. There is no Sziklai or CFP (complementary feedback pair) counterpart but even for discussion, they aren't Darlingtons.
Some of us also use the term "complementary" in reference to a CFP or "complementary feedback pair". This could also mean a dual but with PNP+NPN or NPN+PNP paired transistors. These are not common and as mentioned earlier, possibly don't exist as commercial products. Hence I think, Mark's enthusiasm to get some. There are more contrived terms for these non-existent semis so I imagine the confusion will go on.
Some of us also use the term "complementary" in reference to a CFP or "complementary feedback pair". This could also mean a dual but with PNP+NPN or NPN+PNP paired transistors. These are not common and as mentioned earlier, possibly don't exist as commercial products. Hence I think, Mark's enthusiasm to get some. There are more contrived terms for these non-existent semis so I imagine the confusion will go on.