Best sounding transistors

Many people talk of tube rolling and the sonic differences they impart.  Output transistor are not easily rollable since they are soldered in place, but I'm curious about the sonic signatures different transistors might have.  Who makes the "best" ones?  Which amplifiers use them and what is the native/optimal output impedance on these devices?  I know there is a lot more to an amplifier than just the output transistors, but I'm just curious about this one aspect.  I'm not an electrical engineer either, so please keep it simple.  What do the Pass amps use?
"What do the Pass amps use? " Nelson is using NOS Toshiba JFETs(input stage) and NOS Toshiba MOSFETs(output stage) in his latest amps(the .8 series).
What is NOS?  (It usually means "New Old Stock" when you buy transistors on ebay, but I'm sure Nelson Pass doesn't do it)
 Nelson Pass has a supply of NEW OLD STOCK Toshibas(ie: 2SK170/2SJ74/2SK2013/2SJ313), that he's using in his newest amps.  NOS doesn't only mean, "new old stock" on ebay!

Like you assumed, the sonic signature of the amp has more to do with the surrounding circuits as well as the capability of the power supply. However, there are some aspects to the type of transistor and number.

Generally speaking, a FET type transistor (like JFET or MOSFET) will typically have more of a warm/rich character. FET transistor are usually not precise and require a resistor or pot to balance for DC offset (when used in an input stage).

A bipolar transistor will, typically, sound more clean and will generally have lower distortion. Bipolar are manufactured more precise and don’t need DC offset adjustment. A bipolar transistor will also have more input bias current (which means there can be DC on the output).

The number of output transistor an amp uses can dictate how well it moves the speaker. For example, when you compare an amp with 4 output transistors (2 for positive waveform and 2 for negative waveform) with an amplifier that has 12 output transistors, the transistors on the 12 amp will work 1/3 as hard for the amp to push the same amount of current to the speaker. Each of the 12 transistors do not have to slew up the output as much when compared to the 4 transistor amp). As long as the power supply is capable, the 12-transistors will be more efficient in having more current on tap for the demanding music passages (even at low volume).

When having a large number of transistors, you also get into the linearity issue. A ultra-high end company like D’Agostino will hand match all the output transistors so that they are as close as possible - each one pushing through the same amount of current for a waveform. Most of the amplifiers are machine manufactured and do not have matched transistors. For example, you have 4 transistors for positive waveform -- these transistors could each be pushing different voltages for a given waveform (i.e. 1.7V, 1.88V, 1.65V, 1.75V). While not a bad thing overall, the accuracy of the matched transistors is probably more linear.

Whatever happened to MOSFETS? wheren't they going to fix everything transistor related?

Just asking. :)
@Goose-  Another type you might find of interest: Epitaxial Planar or Ring-Emitter transistors.  They've been used by Pass,  Burmeister,  Symphonic Line, Electrocompaniet, and other respected marques.   Seems the source now is Sanken.
Thanks for your posts.


I would like to comment on your response above.

The offset in a power amp has nothing to do with whether it uses bipolar or MOSFET or any other transistors. It depends on circuit topography and how much DC feedback is used (or if servo amps are employed as in Parasound amps). If you are referring to Vbe matching, bipolars might be better than MOSFETs but that depends on the type and manufacturer.

While using a large number of output transistors reduces the load on each transistor, the downside is that a large number of transistors create a high capacitance load for the driver stages, slowing down the amplifier. This high capacitance reduces the slew rate and can create TIM distortion. In my amp designs, I prefer to use the smallest number of power transistors possible that will get the job done and not experience secondary breakdown under full output current or high bias conditions. This is particularly true for bipolar output stages and less to MOSFET output stages.

You are correct in that a large number of bipolar output transistors can experience current hogging if the thermal environments are not identical, even if they are closely matched for hfe or other parameters. This is another reason to stay away from large numbers in the output stages. Also, with a large number of output transistors, if one fails it can take all the others with it, as the remaining transistors cannot handle the increased current load and fall into secondary breakdown. This is what often happened in the old Phase Linear 400 amps back in the 70's ands 80's.

Regarding the sound aspects, perhaps they exist as with other components, but the design parameters such as the power supply, driver circuits, amount of feedback, etc probably have a much bigger impact. You very rarely hear of power transistor "rolling" as with tubes.
This is what often happened in the old Phase Linear 400 amps back in the 70's and 80's. "  Their 700 had the same issue.   Those of us that repaired so many, often referred to the company as, "Phlame Linear". 
Another thing I wanted to mention regarding power transistor "rolling". The high frequency characteristics of the power transistors can directly affect the open loop bode response and therefor the stability (ie the tendency to oscillate with the feedback loop closed) of the amplifier. Unless the replacements are close to the originals in high frequency characteristics (input capacitance and cut off frequency), you can negatively affect the frequency response with replacements and make the amplifier unstable. This dilemma often makes repairing vintage SS amps difficult as the OEM power transistors are no longer being made. If you have an amp with Motorola power transistors (some of the best ever made), many of these part are currently unobtainium, with no comparable modern substitutes. Add into the equation the need for a power amp to drive complex impedance loads with high inductance and capacitance, messing with output transistors can be a recipe for disaster unless you really know what you are doing.

@dhl93449 - thanks for the added information. I did preface my comments on the offset adjustment for FET transistors was specifically when used in an input stage - such as comparator circuit or a discrete operation amp circuit, where you need an exact mirror pair of transistors (whether FET or bipolar type). Monolithic FET op amps (such as OPA627, OPA2134, etc.) are laser trimmed during manufacturing and the offset is excellent. I was only trying to give the OP more detail on transistor types.

I didn’t know about the larger number of transistors slowing down the slew rate. That is useful information when I look at different amps. Though, I have listened to the Emotiva XPA-1, with 24 transistors total (12 per differential side). There’s nothing slow about this particular amp. :)  Though, a Class A bias might slow down the slew rate - possible more specifically on the negative slew direction.


The slew rate is determined by typically two parameters (simplistically); the driver circuit bias current, and the Cob of all the collective (bipolar) power transistors the driver is coupled to. Think of it as the bias current being the water flow from a hose and the capacitance being an empty bucket. The higher the bias, the higher the water flow. The bigger the bucket, the larger the collective parallel capacitance. The slew rate is determined by how fast the bucket fills with water from the hose. So, you can have a larger collective capacitance (ie more power transistors) if you have a high class A bias current in the driver stage, which can still give you an acceptable slew rate. So just counting the number of transistors in an output stage is insufficient for comparison. The type of transistor also makes a difference, as the Cob is generally lower for very high frequency transistors, but they typically have lower max current ratings so you need more of them for a given power. Also, you need consider whether the design uses a bridge type configuration (which is actually TWO amplifiers in a channel, inverted in phase) which will give you double the power transistors without the slew rate limitations.

Regarding offset voltages, in a power amp built from discrete transistors, the DC offset is not just determined by the input stage, as you seem to assume. If left open loop with no feedback, you will find the total DC offset may be due to unbalanced bias currents in the driver stages, for example. This may also be true for Op-amps as well, but engineers are commonly used to "reflecting" the net offset to the input stage when the op amp is closed loop with a net gain of 1. So many think all the offset voltage is due to the input stage when actually it is not. This is mathematical abstraction so that a real op amp can be compared to the hypothetical "ideal" op amp. So I return to my original point, which is that DC offset in a power amp has little to due with the type of transistor used, but is more design dependent. In another example, some power amp designs use inter-stage coupling capacitors which remove any DC offset, at the expense of adding a coupling cap in the amplification line. Op amps NEVER use inter-stage coupling caps, primarily because you cannot build capacitors of sufficient quality and size on a monolithic integrated circuit die; and because most industrial applications require op amps with DC gain. Audio amps require only AC gain.