Thoughts on discreet tubs, transistors vs. IC chips, op-amps.,.etc


What has been your experience with components using tubs or transistors vs., chips/op-amps..etc?  It has been some time since I have owned anything based on chips/op-amps and wondered if there had been major improvement sonically or if discreet transistors or tubes are still at the forefront?


whatjd
There are trade-offs to all of them, and my guess is that the "tube sound" will continue to enthrall many, right or wrong.  But here's the funny thing: in my design career i manage to get "similar" sound out of all of them.  Don't read too much into that - i still, overall, prefer discrete solid state devices to most chips, and to tubes, but its amazing how the major design goals transcend the packaging.
Also -- don't necessarily confuse "opamps" with "chips"  opamps, after all, started with tubes in analog computers of the 1940s and 1950s.  One one can buy monolithic transistor pairs etc.  I even worked with a researcher once who had invented a solid state tube/ Really - a vacuum tube. It never when anywhere if anyone's interested.
In general, chips allow for complex designs that would be large and costly in discrete form.  They also allow for decent device matching as a side product of the monolithic process. These are good things.
On the other hand the very geometry and small size leads to compromises, and things like connecting resistors etc. must, in reality, be semiconductors -- not metal film, carbon film, etc.
I also question whether all the complexity chips allow is good for the sound.  Ok, i don't question it, I think its NOT.

In the end though, good design is good design.  I've made differential front-end amps on a  chip sound almost tube-like. Except they are quiet, have bass, and don't degrade.
By the way today's best opamps and instrumentation amps are vastly, vastly better than they were 20 years ago.  One very well respected designer int he rockies uses them at the heart of a stupid-expensive preamp. And it sounds good: i've used the same device.
Happy thanksgiving.  Everything soudns better after good wine and good food with good friends.

G


On the other hand the very geometry and small size leads to compromises, and things like connecting resistors etc. must, in reality, be semiconductors -- not metal film, carbon film, etc.

This isn't true, I use thin film smd resistors which are very easy to find and there's a much wider range of C0G/NP0 capacitors available in smd than through hole. If you look at a datasheet for an op amp the design is likely to be an entirely straight forward transistor based amplifier in a small package.

Small packages have significant advantages in audio applications, the feedback loop from the output to the inverting input of an op amp can be reduced to a few millimetres which results in improved accuracy. Also IC's are very accurate in their own right as components such as resistors can be laser trimmed and very precisely matched. The drawbacks really come down to heat dissipation and the voltage handling capability. Small packages have a small surface area which requires effective heat sinking... which if designed properly will work fine. The other drawback for the designer is that the maximum rail voltage is likely to be dictated by the IC, so if you want +-60V rails you have to go for discrete.

There are many ways of optimising the performance of ICs, at the moment my favourite is to use a power amp IC in the feedback path of a precision audio op amp.

pragmasi,

I think itsjustme was specifically referring to resistors right on the integrated circuit, but what was stated, that they "must be semiconductors" is wrong.  They can be built from a diffused region where you take advantage of the bulk resistivity of the diffused semiconductor material (but not acting as we think a semiconductor), you can build it in the epitaxial layer, and you can diffuse thin film resistors right onto the chip. The former 2 are not very accurate and tend to be rather temperature sensitive. Thin film on chip is not very accurate unless trimmed and that would typically only be done for op-amps with on chip resistor dividers.

I completely agree with your short path arguments, though you probably need to better define accuracy.

I think itsjustme was specifically referring to resistors right on the integrated circuit

@atdavid

Ah... okay, that makes more sense. As you imply 'accurate' is a bit of lazy shorthand... I meant an accurate rendering of the output signal into the inverting input i.e. relatively unchanged by the parasitic characteristics of the copper traces. Probably more importantly shorter traces improve the stability of the IC by presenting a low inductive load. I have generally found that compactly designed layouts are more linear.

a bit of an apology,  my keyboard "e" does not always respond...hence a word like tube can end up being tub..if I don't catch it.  And yes, my keyboard is younger than I, but does not function as well. 

On the other hand the very geometry and small size leads to compromises, and things like connecting resistors etc. must, in reality, be semiconductors -- not metal film, carbon film, etc.

This isn't true, I use thin film smd resistors which are very easy to find and there's a much wider range of C0G/NP0 capacitors available in smd than through hole. If you look at a datasheet for an op amp the design is likely to be an entirely straight forward transistor based amplifier in a small package.

You are missing my context. I am referring to monolithic device - chips -- that contain all semiconductors, resistors etc in a package. Example - 8 pin opamp. This is true regardless of SMD or through hole.  yea, sure if we're in discrete resistor land the part selection is often BETTER in with SMDs because that's the modern packaged, designed for automated placement.

I was very clear: ( I wrote)

In general, chips allow for complex designs that would be large and costly in discrete form. They also allow for decent device matching as a side product of the monolithic process. These are good things.
On the other hand . . .
But you have to read in context! Context matters.
'later
G