opamp designs vs. discrete designs

[kelly]

This topic has come up over in the HeadRoom forum in which one user requested that HeadRoom build an amp similar to Kevin Gilmore's all discrete solid state headphone amp.

I've decided to stop participating in the HeadRoom forum for various reasons so I'll not be responding to that thread and I think the topics that spooled from that thread are more general than HeadRoom specific anyway: namely those related to opamp designs vs. discrete designs.

My favorite headphone amplifier currently is Meier Audio's Corda Prehead-1 which is an opamp design. In direct AB comparison with Antness' Gilmore amp I very slightly preferred the Prehead. But what surprised me was that the sound was so similar anyway. Given that I credit both Gilmore and Meier as being excellent designers, my assumption (as a layman) was that the Gilmore would blow away the Corda, but it didn't turn out that way. Instead, it was simply *very* close.

I'd appreciate it if any of the more technical guys could shed light on this. I hope Jude will have some patience if this involves people who design and sell amps employing either philosophy. I'd really like to see this debated further and hope those of you who are technical and inclined to do so will be patient with folks like me who may need a little more fundamental explanation.

Anyone care to comment?

 

[HD-5000]

I'm in no ways techinical, but I know that an opamp is merely a bunch of transistors crammed into a small chip. I wouldn't be surpised if an opamp based headphone amp outperforms an all discreet design.

 

[AndreYew]

Kelly,

It depends.

Generally, higher speed requires smaller geometry, hence opamps. Devices are better matched, and are on the same piece of material so thermal effects are the same. But you're stuck with whatever design Burr-Brown, etc. give you. Note that HeadRoom doesn't just do opamps --- at least on a simplified schematic of an older HeadRoom module I saw, discrete transistors (probably for current gain) drive the headphones. The opamp was there for voltage gain and buffering the crossfeed.

Discrete is generally slower, but simpler, and can be customized to what you need. Gilmore wanted to drive low-Z phones like Grados, and build something to fulfill his audio philosophy, so he did discrete. You can certainly get more power out of discrete designs, but we're talking speaker amplification here, not headphones. Discrete transistors also come in IC packages with more than one transistor packed into one package for better matching and thermal stability.

Both can work very well, but both require attention to detail to do so. I think it's oversimplifying to pit designs against each other on the basis of discrete vs. op-amp.

--Andre

 

[Zanth]

Andre: that cleared up a lot for me! Thanks

 

[kelly]

Andre
I didn't at all mean to oversimplify and have people pick a winner. Your post is exactly the kind I was hoping to see and I much appreciate it.

 

[AndreYew]

Kelly,

Sorry --- I didn't mean to imply that you specifically were oversimplifying. I was just speaking to the habit in discussions (as exemplified in the HeadRoom Gilmore amp thread) of polarizing complex issues.

--Andre

 

[kelly]

Andre
I understand but I'm hoping to bring out the other kind of discussion. I thought some excellent points were made in that thread on both sides. I'd like to see that continue but the HeadRoom forum is a limited place to discuss that. They sell their own amps and some folks might not speak their mind consider it's their forum or in the least would be viewed differently if they did. I'd just like to see more elaboration on the reasoning of one philosophy versus the other. The less simple the better for this discussion.

 

[jopi]

Something along the same lines:
I remember reading about an interview with Tim de Paravacini from EAR about differences between tube and solid state equipment. He stated that design is much more important than components and a good designer can make tube equpiment sound like solid state and vice versa.

Tim prefered tube equipment for marketing reasons.

 

[morsel]

Traditionally, discrete designs are better because of the inherent flexibility in designing from the ground up, however, opamps are cheap, easy, and can be made to perform admirably.

 

[skippy]

well, opamps don't have to be on a single chip. you can make discrete opamps. you can even make opamps out of tubes. an opamp is just a dc coupled differential amplifier, usually with a really high gain. with all designs there are tradeoffs.

one such example would be with an opamp, because of the high open loop gain, you have to use a lot of feedback which has it's advantages and disadvantages. it lowers thd, but creates distortion of other sorts. blah blah blah...

 

[Gergor]

Quote:

Originally posted by skippy ... but creates distortion of other sorts. blah blah blah...

What sort of distortion? With a really high loop gain system it may have stability issue, but creating distortion... how does it create distortion?

 

[antness]

Kevin Gilmore goes a bit more in depth as to exactly what kind of distortion the high open loop gain of an opamp creates. The info is available in the HeadWize Project Library

 

[Gergor]

I took a quick look at http://headwize2.powerpill.org/projects/showproj.php?file=gilmore3_prj.htm]headwize[/URL], it didn't explain much, but it mentions "transient intermodulation products are much higher than they should be" in a high loop gain system.

Well, I don't exactly know what transient intermodulation product is, but since the word "transient" is used, I suspect it has to do with the op-amp being slew limited. I suppose in a high loop gain system, one probably needs to use large compensation capacitor to make the loop stable. If the bias current (in case of class A) is not beefy enough, then the amp will be slew limited when a large amplitude fast signal is encountered.

Well, I guess it makes some sense. I'm not a designer. I'm just trying to understand the issue based on my limited textbook knowledge.

 

[skippy]

since you're feeding the signal from the output back to the input, the output signal isn't exactly the same as the input signal, it's a little delayed in time.

 

[AndreYew]

Skippy,

Quote:

since you're feeding the signal from the output back to the input, the output signal isn't exactly the same as the input signal, it's a little delayed in time.

This is a common argument against negative feedback, but the feedback equations do take this into account, and properly designed feedback loops do behave the way the feedback equations predict they do.

TIM is really not a big deal since it doesn't exist in well-designed feedback schemes.

--Andre