What's your opinion on opamp vs discrete? Liquid Spark description made me wondering.

[kongwee]

Absolutely not. Chip op amps are, by definition, made from the same silicon with uniform impurity density throughout. They are more likely to be matched in tolerance than discrete parts and definitely will temperature track better. Surface mount is a packaging technology. The internal components are generally not different from through-hole.

I don't know. I know do resistor, capacitor and inductor have tolerance value. 1% is highest you can easily source. Some amp or speaker marker just make RCI them to bring the tolerance lower. Dunno about the opamp you can get easily.

 

[voodooless]

I don't know. I know do resistor, capacitor and inductor have tolerance value. 1% is highest you can easily source. Some amp or speaker marker just make RCI them to bring the tolerance lower. Dunno about the opamp you can get easily.

Opamp components can be laser trimmed to within 0.01% or better.

 

[solderdude]

I don't know.

Don is absolutely correct, that's something I do know.
It should be noted that on chips you can often also find components that are not available in discrete components. Say a transistor with more than one collector or emitter or some other nifty components and, important, the whole chip is temperature coupled.
op-amps can be realitively simple to very complex designs that would need a huge PCB if made in discrete form.

Funnily enough the Liquid Spark is basically a simple opamp schematic build with not as well matched components as is possible in an op-amp.

Both in discrete circuits as well as op-amps there are poor to great performing ones with various aspects they excel at or are poor at.
A matter of choosing the right component for the job at hand is the most important aspect followed by its implementation and practical build.

 

[Palladium]

All this talk is still ultimately meaningless if you only care about end performance and ($).

 

[MAB]

I don't know. I know do resistor, capacitor and inductor have tolerance value. 1% is highest you can easily source. Some amp or speaker marker just make RCI them to bring the tolerance lower. Dunno about the opamp you can get easily.

You can exceed discrete with monolithic. For instance monolithic op amps have laser trimmed components to eliminate input offsets, differential noise, etc. You are wrong about this, and shouldn't speculate, you just don't know. Perhaps read:

https://www.ti.com/lit/an/sbot037c/sbot037c.pdf?ts=1670385967307&ref_url=https%253A%252F%252Fwww.google.com%252F

 

[Waxx]

Both have their places and uses. Both have advantages and disadvantages. And stories that opamps or discrete circuits are always the best are mostly marketing stories. Discrete can do things opamps can't and reverse. And a good engineer should use the right part on the right place for the right task.

 

[DonR]

Both have their places and uses. Both have advantages and disadvantages. And stories that opamps or discrete circuits are always the best are mostly marketing stories. Discrete can do things opamps can't and reverse. And a good engineer should use the right part on the right place for the right task.

Absolutely but matched JFETs or even dual monolithic JFETs are becoming as useful as selenium rectifiers or germanium transistors.

 

[Waxx]

Absolutely but matched JFETs or even dual monolithic JFETs are becoming as useful as selenium rectifiers or germanium transistors.

I don't think so. JFET discrete transistors are very usefull to add controlled harmonic distortion to a circuit if wished. Much easier than triode tubes can do that. The big problem now is that there are very few good discrete JFET's availeble and that art of tuning JFET's right is something that not many can do right (or, it takes a lot of practice and knowledge to do it right). Nelson Pass his whole emperium is made on that (not always the right way, but often it is). Germanium transistors are just noisy, not in a good way.

If it's your cup of tea, that is an other question... If you want a superclean circuit, then opamps are indeed much easier to deal with. So it's logic most use opamps for that purpose. With JFET it's also possible, and it's done in some microphone preamps that are very popular in the studio's, but it's much more difficult and expensive do to it right.

 

[DonR]

I don't think so. JFET discrete transistors are very usefull to add controlled harmonic distortion to a circuit if wished. Much easier than triode tubes can do that. The big problem now is that there are very few good discrete JFET's availeble and that art of tuning JFET's right is something that not many can do right (or, it takes a lot of practice and knowledge to do it right). Nelson Pass his whole emperium is made on that (not always the right way, but often it is). Germanium transistors are just noisy, not in a good way.

If it's your cup of tea, that is an other question... If you want a superclean circuit, then opamps are indeed much easier to deal with. So it's logic most use opamps for that purpose. With JFET it's also possible, and it's done in some microphone preamps that are very popular in the studio's, but it's much more difficult and expensive do to it right.

I would think a DSP would be better suited to that task as then the end user can apply as much or little as desired. At any rate, this is a niche and singularly unique application which is one reason they are disappearing. Furthermore, the original device being talked about here was the Liquid Spark which makes no such claims but which does make the claim that discrete is low noise (implying lower than monolithic IC) which is, of course, patently false.

 

[syn08]

JFET discrete transistors are very usefull to add controlled harmonic distortion to a circuit if wished.

Can you expand on this, in particular the control method(s)?

Ignoring the subthreshold region (of no practical importance in audio low level applications), the JFET device physics does not reveal anything special, other than the parabolic Id vs. Vgs, which has some advantage over the bipolar transistors (exponential Ic vs. Vbe, potentially generating odd harmonics) but at the price of a much lower transconductance.

 

[Roland68]

Is there any truth to this description?
Btw, I saw the measurements here and there are some flaws to this amp:

Review and Measurements of Monoprice Liquid Spark Headphone Amp

This is a review, detailed measurements and comparison of Monoprice Liquid Spark, Schiit Magni 3 and JDS Labs O2 headphone amplifiers. The Liquid Spark and Magni 3 retail for $99 but only the Monoprice comes with free expedited shipping. The JDS Labs O2 retails for USD $139 on US Amazon with...

www.audiosciencereview.com

The next deja vu... #782
These discussions have been going on for at least 30 years and still no winner in sight
Just look at what the most used technology is in the ASR leaderboard.

 

[DonR]

The next deja vu... #782
These discussions have been going on for at least 30 years and still no winner in sight
Just look at what the most used technology is in the ASR leaderboard.

Only in audiophillia world. The vast majority of EE has moved on to increasing levels of integration. I do bemoan the lack of repairability in modern electronics, however. Everything is now reduced to replacing proprietary ICs and boards.

 

[atmasphere]

OK- here's an advantage of opamps.
When feedback is applied in most amps, its applied to a transistor near the input, typically one of a pair of differential input transistors. The other transistor of the pair usually has the input signal on it. Because they are differential, the feedback signal is thus mixed with the incoming signal so as to do correction.

The problem is that transistor isn't linear. It might be close, but in this case no cigar; the feedback signal gets a bit of distortion added to it as a result. And as a result, bifurcation ensues, resulting in lots of higher ordered harmonics. If you have ever wondered why feedback has a bad reputation, this is why. Because harmonics govern how instruments sound its not rocket science to sort out that when harmonics are added by the amplifier, the instruments aren't going to sound right. Just as the ear assigns tonality to instruments based on their harmonic content, it also assigns tonality to distortion. In this case, it assigns 'bright and harsh' to the higher ordered harmonics and inconveniently, is keenly sensitive to them as they are used to gauge sound pressure by the ear (which has a +120dB range...). If you've ever wondered why those crazy audiophile tube lovers like their tube amps so much, this is at the heart of it! Those added harmonics in the solid state amp makes the amp sound harsh and bright. Tubes take feedback in the same way, but generally speaking the feedback node is a bit more linear and so the feedback does less damage doing its job. Of course with a tube amp you can't apply the sort of feedback you really need to because you'll exceed the phase margin of the design- which is an advantage of solid state, but even there phase margins are still a lurking concern (some amps will go into oscillation with certain capacitive loads on this account; a sure sign that their feedback loops are poorly designed).

However opamps don't take feedback internally. Instead, the feedback resistor is part of a voltage divider network outside the opamp itself. As you might expect, resistors are a lot more linear than either a tube or a transistor. And unsurprisingly, the feedback signal is therefore not distorted nearly as much before it can do its job!

The result is an opamp circuit can be really neutral in a way that discrete circuits cannot (the exception being if the discrete circuit is actually a high performance opamp), as long as you are careful to not ask too much gain out of them (and it should be noted that Gain Bandwidth Product plays an enourmous role here; early opamps from the 1960s and 1970s have a 'sound' because of this and if you are trying to replace such a device in an older guitar pedal or the like you will have to hunt down the older opamps rather than the current fare if you want the musician to not be mad at you for messing up the sound of their pedal when you 'fixed' it...). These days as long as you don't go for more than 20dB with most opamps you're alright.

FWIW the Purifi module takes its feedback in this manner rather than injecting it internally to a non-linear feedback node. Also not surprisingly, its got pretty low distortion.

 

[syn08]

sound pressure by the ear (which has a +120dB range...).

Many inaccuracies in your message (may I suggest The Art of Electronics for a better understanding of what's really happening in feedback circuits) but only the above statement raises a question... 120dB range of sound pressure??? Are you sure???

 

[Stinius]

Op-amps are the best things going, especially for inexpensive high performance audio gear. It is almost impossible for discrete to match a good op-amp. The only problem with feedback for audio circuits is if you don't use enough of it.

No, it is good for inexpencive audio gear.
Nothing beats a well designed discrete circuit.
I dont know why you are mentioning feedback, I agree that you should use as much feedback as possible, but a well designed discrete circuit will be able to use a lot more feedback than an opamp and a lot better compensation.

S

 

[dfuller]

Discrete has one advantage that monolithic op amps can have trouble with, and that is noise performance. Very few monolithic op-amps can match the noise performance of a John hardy 990C.

But otherwise, discrete is more expensive, harder to design for low distortion, and mostly marketing for line-level signals. High power (i.e. speaker amps) is another matter entirely, where I would wager discrete output transistors is a better option.

 

[atmasphere]

Many inaccuracies in your message (may I suggest The Art of Electronics for a better understanding of what's really happening in feedback circuits) but only the above statement raises a question... 120dB range of sound pressure??? Are you sure???

Yes I am. Not saying that 120dB won't be painful!

My statements are correct although a bit generalized (which often leads to inaccuracies) so those without an EE can follow along. Are you saying that you can distort a feedback signal and it will still do its job properly?? -or am I just reading that into your comment (and if so I apologize)? Norman Crowhurst wrote about this problem in feedback circuits 65 years ago so I expected nothing controversial here!

 

[Stinius]

OK- here's an advantage of opamps.
When feedback is applied in most amps, its applied to a transistor near the input, typically one of a pair of differential input transistors. The other transistor of the pair usually has the input signal on it. Because they are differential, the feedback signal is thus mixed with the incoming signal so as to do correction.

The problem is that transistor isn't linear. It might be close, but in this case no cigar; the feedback signal gets a bit of distortion added to it as a result. And as a result, bifurcation ensues, resulting in lots of higher ordered harmonics. If you have ever wondered why feedback has a bad reputation, this is why. Because harmonics govern how instruments sound its not rocket science to sort out that when harmonics are added by the amplifier, the instruments aren't going to sound right. Just as the ear assigns tonality to instruments based on their harmonic content, it also assigns tonality to distortion. In this case, it assigns 'bright and harsh' to the higher ordered harmonics and inconveniently, is keenly sensitive to them as they are used to gauge sound pressure by the ear (which has a +120dB range...). If you've ever wondered why those crazy audiophile tube lovers like their tube amps so much, this is at the heart of it! Those added harmonics in the solid state amp makes the amp sound harsh and bright. Tubes take feedback in the same way, but generally speaking the feedback node is a bit more linear and so the feedback does less damage doing its job. Of course with a tube amp you can't apply the sort of feedback you really need to because you'll exceed the phase margin of the design- which is an advantage of solid state, but even there phase margins are still a lurking concern (some amps will go into oscillation with certain capacitive loads on this account; a sure sign that their feedback loops are poorly designed).

However opamps don't take feedback internally. Instead, the feedback resistor is part of a voltage divider network outside the opamp itself. As you might expect, resistors are a lot more linear than either a tube or a transistor. And unsurprisingly, the feedback signal is therefore not distorted nearly as much before it can do its job!

The result is an opamp circuit can be really neutral in a way that discrete circuits cannot (the exception being if the discrete circuit is actually a high performance opamp), as long as you are careful to not ask too much gain out of them (and it should be noted that Gain Bandwidth Product plays an enourmous role here; early opamps from the 1960s and 1970s have a 'sound' because of this and if you are trying to replace such a device in an older guitar pedal or the like you will have to hunt down the older opamps rather than the current fare if you want the musician to not be mad at you for messing up the sound of their pedal when you 'fixed' it...). These days as long as you don't go for more than 20dB with most opamps you're alright.

FWIW the Purifi module takes its feedback in this manner rather than injecting it internally to a non-linear feedback node. Also not surprisingly, its got pretty low distortion.

This is far out and loaded with meaningless nonsense. What are you smoking?

 

[DonR]

Discrete has one advantage that monolithic op amps can have trouble with, and that is noise performance. Very few monolithic op-amps can match the noise performance of a John hardy 990C.

But otherwise, discrete is more expensive, harder to design for low distortion, and mostly marketing for line-level signals. High power (i.e. speaker amps) is another matter entirely, where I would wager discrete output transistors is a better option.

At those levels, noise is not audible. Even the NE5532 has inaudible noise at 5 nV/sqrt(Hz) which is less than 1/2 an order of magnitude more but has an order of magnitude less THD. It sells for pennies.

 

[dfuller]

At those levels, noise is not audible. Even the NE5532 has inaudible noise at 5 nV/sqrt(Hz) which is less than 1/2 an order of magnitude more but has an order of magnitude less THD. It sells for pennies.

It matters much more when there's lots of gain going on, IME. Mic preamps for example where there is a truckload of gain (55dB on the low end) are much more sensitive to noise.