# Discrete vs monolithic op amps for audio

#### galed

##### New Member
One of the buzzwords in the high-end audio industry is "discrete op amps." I'm wondering about their advantages over monolithic op amp circuits. The monolithic amplifiers would have to be better for high frequencies, right? Considering that audio tops out around 22kHz i don't think that's much of an issue.

What advantages do discrete op amps have over monolithics?

#### bychon

##### New Member
The major advantage of discrete op amps is the same as oxygen free copper. It's a money maker. As P.T.Barnum said, "There's a sucker born every minute". Then, there are plenty of people who think transistors have no place in amplifiers, and monolythic op-amps are chock full of transistors.

Op-amps are a feedback oriented way of thinking. That makes bandwidth and gain easier to work with compared to single stages all in a row. Still, you don't need to start with a differential pair to provide feedback. Where is the line between single stages with a feedback loop to the first emitter (or cathode) and an op-amp? If a feedback circuit is all that is required to form an op-amp, a Fender Princeton guitar amp is a discrete op-amp! If a differential pair is required to define an op-amp, Leo Fender didn't make those.

Personally, I know a guy who believes transistors have no purpose in audio amps. He actually has an amazing ear (he's a musician). I built an amp for him with 2 vacuum tubes, 4 monolythic op-amps and a discrete j-fet input stage. He installed that rack in his permanent setup. That was 15 years ago, he still plays through that amp, and he still doesn't know about the transistors.

Have you figured out what I think of, "buzzwords"?

#### audioguru

##### Well-Known Member
Monolithic amplifier ICs have a max output of only 50W into 8 ohms but are excellent (LM3886). Some people parallel and bridge them for 200W.
Discrete amplifiers have an output of thousands of Watts.

#### galed

##### New Member
Monolithic amplifier ICs have a max output of only 50W into 8 ohms but are excellent (LM3886). Some people parallel and bridge them for 200W.
Discrete amplifiers have an output of thousands of Watts.
Why not use a monolithic amp and a gain stage? I guess the point would be to do it all in one stage. I'm looking into a lot of audio electronics -- can you recommend any good resources?

Personally, I know a guy who believes transistors have no purpose in audio amps. He actually has an amazing ear (he's a musician). I built an amp for him with 2 vacuum tubes, 4 monolythic op-amps and a discrete j-fet input stage. He installed that rack in his permanent setup. That was 15 years ago, he still plays through that amp, and he still doesn't know about the transistors.

Have you figured out what I think of, "buzzwords"?
I think what you think of buzzwords is about the same as what I think of buzzwords: they're marketing. This says it all: xkcd: Free

And good work on that amplifier.

#### Ubergeek63

##### Well-Known Member
The major advantage of discrete op amps is the same as oxygen free copper. It's a money maker. As P.T.Barnum said, "There's a sucker born every minute". Then, there are plenty of people who think transistors have no place in amplifiers, and monolythic op-amps are chock full of transistors.

Op-amps are a feedback oriented way of thinking. That makes bandwidth and gain easier to work with compared to single stages all in a row. Still, you don't need to start with a differential pair to provide feedback. Where is the line between single stages with a feedback loop to the first emitter (or cathode) and an op-amp? If a feedback circuit is all that is required to form an op-amp, a Fender Princeton guitar amp is a discrete op-amp! If a differential pair is required to define an op-amp, Leo Fender didn't make those.

Personally, I know a guy who believes transistors have no purpose in audio amps. He actually has an amazing ear (he's a musician). I built an amp for him with 2 vacuum tubes, 4 monolythic op-amps and a discrete j-fet input stage. He installed that rack in his permanent setup. That was 15 years ago, he still plays through that amp, and he still doesn't know about the transistors.

Have you figured out what I think of, "buzzwords"?
that is because the transistor portion probably reproduces sound accurately... but it is the distortion that the tubes add that audiophiles adore.

a nice class d power stage reproduces accurately and a properly biased tube preamp is all it takes to please the musician.

Dan

#### unclejed613

##### Well-Known Member
a good monolithic op amp (TLE2072 for example) is far better than a discrete, and takes up a lot less pc board real estate. that said, building a discrete op amp can be more enjoyable and educational. there are also some things you can do with a discrete that you can't do with a monolithic. you can experiment with the compensation, input stage, voltage amp, output stage. you can try various design tricks such as current sources and current mirrors. i have seen some excellent discrete op amps. one in particular was one that has a very interesting diff amp for an input stage. it had a jfet for the + input and a bipolar transistor as the - input. obviously the output had a bit of DC offset and was coupled through a capacitor, but it was very low noise, and very low distortion, and the output stage was run with class A bias (another thing that's difficult to do with monolithics without damaging them). so if you want to try something different, go ahead and build a discrete. but modern monolithics are excellent devices that are near impossible to beat. years ago, it was possible to easily outclass a monolithic op amp with a discrete one. modern op amps have been improved to the point that, for all intents and purposes make them equivalent to "a piece of wire with gain"

the "myth" that discretes sound better has two possible reasons for existing:
1:during the 1970's and most of the 80's, a lot of audio gear was made with LM741, LM1458, and LM324 op amps (or other manufacturer's version of them), which are (except for how many amps are in each chip) about the same. the next big improvement was with the TL072 and LF353 types of op amps with 3Mhz bandwidth and very low noise. a well designed discrete might still have a slight edge, but these devices were much better for audio than those of the 741 family. the newest generation of op amps is vast and very diversified, today there are op amps with GBW's of 100Mhz and higher. for audio, the latest one i have looked at is actually a "hot-rodded" TL072, the TLE2072. it has a GBW of 10Mhz, and 17nV√hz noise voltage. to build a discrete that would beat this one would be extremely difficult.

2: "Golden Ears" like even harmonic distortion. that's why some people like tube sound better, and a discrete op amp when not well balanced has plenty of even harmonics in it's output to keep "Golden Ears" happy

Last edited:

#### galed

##### New Member
thanks for the replies. really informative.

so it's really completely hype and marketing? I guess discrete would let you tune the exact sound of the amp, but aside from that....?

#### Electronworks

##### New Member
I think it is historical. Doug Self (the UK audio guru) says that early integrated designs did not have the PSRR that later chips have, therefore the discrete design was always considered superior.

I think a mixed of both is needed: IC on the front end (to get good PSRR, a well matched differential stage on the front end, well defined gain and bandwidth) then have discretes on the output to get the power up.

Audio is (and always will be) a personal thing, so you will alway get emotional differences of opinion

Oh... and throw away the gold plated mains plug and directional speaker cable

#### Hero999

##### Banned
Personally, I know a guy who believes transistors have no purpose in audio amps. He actually has an amazing ear (he's a musician). I built an amp for him with 2 vacuum tubes, 4 monolythic op-amps and a discrete j-fet input stage. He installed that rack in his permanent setup. That was 15 years ago, he still plays through that amp, and he still doesn't know about the transistors.
Did you valves (tubes) do anything except glow and make it look pretty?

Monolithic amplifier ICs have a max output of only 50W into 8 ohms but are excellent (LM3886). Some people parallel and bridge them for 200W.
Discrete amplifiers have an output of thousands of Watts.
But the input stage is normally monolithic, the discrete transistors are just a power booster, right?

#### unclejed613

##### Well-Known Member
thanks for the replies. really informative.

so it's really completely hype and marketing? I guess discrete would let you tune the exact sound of the amp, but aside from that....?
a really good amp shouldn't have a sound of it's own... which is really the point here.

#### unclejed613

##### Well-Known Member
I think it is historical. Doug Self (the UK audio guru) says that early integrated designs did not have the PSRR that later chips have, therefore the discrete design was always considered superior.

I think a mixed of both is needed: IC on the front end (to get good PSRR, a well matched differential stage on the front end, well defined gain and bandwidth) then have discretes on the output to get the power up.

Audio is (and always will be) a personal thing, so you will alway get emotional differences of opinion

Oh... and throw away the gold plated mains plug and directional speaker cable
try to get 120db PSRR on a discrete op amp. without real careful matching of transistors (or monolithic pairs of transistors) and resistor values, you won't get anywhere near a monolithic's performance.
a good power op amp (like an Apex) is more expensive than your average complete amplifier. and you still have to add a power supply, chassis, circuit boards, etc...

don't forget the cryogenically de-stressed teflon insulated oxygen free silver coax cable with gold alloy connectors for the interconnects at over $100.00 per foot. #### audioguru ##### Well-Known Member Most Helpful Member this one is actually affordable at$8 for BTL 300W into 8 ohms: TDA8950.
Its datasheet says its output is 150 Whats per channel at a horrible-sounding 10% distortion or 100 Watts per channel at 0.5% distortion which is still heard. Maybe at 80 Watts per channel then it might sound good.

#### Ubergeek63

##### Well-Known Member
Its datasheet says its output is 150 Whats per channel at a horrible-sounding 10% distortion or 100 Watts per channel at 0.5% distortion which is still heard. Maybe at 80 Watts per channel then it might sound good.
I actually said BTL ... 200W 0.5% THD .... the 300W was the 10% and it would indeed sound ugly.

so you would probably want, as you say, to keep it under 150W but my point remains: it is a simple, inexpensive power amp at $8 as opposed to using 4 LM3886s at$7 each to get 120W and needing 4X the power supply.

#### schmitt trigger

##### Well-Known Member
High endAudio is, unfortunately, full of con artists. There is also a lot of extraordinary individuals who are always pushing the envelope. The key is to figure out who belongs to which category.

I rember a while back there was a company who you would send your CD collection. Then they would cool it with liquid nitrogen, to "align the molecules and eliminate the harsh digital modulation for a softer, truer to life musical experience".

Digital Guru Ken Pohlmann once launched an open invitation, which if you could prove that such claims were true, he would fill his Florida pool with LN and jump in it.
To my knowledge, Ken is still alive and kicking.....

#### audioguru

##### Well-Known Member
The class-D amplifier has the distortion of a cheap PA amp or guitar amp.
But the analog LM3886 is true hi-fidelity quality. Four produce 200W at extremely low distortion and wide bandwidth, not just 120W.

A guy in New York makes a stereo amplifier with two LM3886 ICs and sells many for $4000.00 each. They are called a Gain-Clone Amplifier in Google. #### Ubergeek63 ##### Well-Known Member The class-D amplifier has the distortion of a cheap PA amp or guitar amp. But the analog LM3886 is true hi-fidelity quality. Four produce 200W at extremely low distortion and wide bandwidth, not just 120W. A guy in New York makes a stereo amplifier with two LM3886 ICs and sells many for$4000.00 each. They are called a Gain-Clone Amplifier in Google.
Gee audio, you really don't like a fair fight, do you? you made this outrageous jump from what a hobbyist could conveniently do to what some jackal manages to con ill-informed audiophiles out of in your desperation to discredit class d amps...

And you are comparing that to 30W class AB mono at $7 each... seems to be a bit of a discrepancy there ... #### audioguru ##### Well-Known Member Most Helpful Member had you read the spec, it is 125W into 8 ohms unclipped. I got 100W at 0.03% looking at the graph. But the graph showing BTL distortion clearly shows clipping beginning at 90W when the supply is 50V. At 142W the distortion is 2% to 5%. The graph of "Unclipped Output power Vs Supply Voltage" shows 142W with a 50V supply. And you are comparing that to 30W class AB mono at$7 each.
The datasheet for the LM3886 says that its output into 8 ohms is typically 50W at 0.1% max distortion with a 70V supply. But the graphs of THD vs Output Power show 60W at 1kHz with only 0.002% distortion and show 60W at 20kHz with only 0.01% distortion.

The LM3886 has only 11 pins in two rows and has a metal tab that is bolted to a heatsink. The web is full of optional circuits and pcb designs.