Fully discrete circuit vs non-discrete

Hey

First of all, I am not sure if there is a specific term for a non-discrete circuit. Doing some googling, most people refer to it as "integrated", but I never heard the term used for DAC's.

Is a fully discrete circuit always better than a hybrid or non-discrete one? As far as I understood, in DAC's non-discrete circuits, they generally use OP-Amps.

For instance, if one dac uses a non-discrete output stage and the other a fully discrete one, with everything else exactly the same, would the discrete one sound better?

Or is a full discrete circuit just a means to an end, a way for designers to squeeze out the maximum performance of a DAC chip? Or a necessity for some DAC designs?

Would love to hear your opinions.

 

I am pretty sure that ICs get a bad rap because they often appear in lower-end products with minimal attention paid to the power supply. Discrete designs can be awful if poorly designed and amazing in the hands of someone competent. Same thing with ICs.

We probably don't see any TOTL products with ICs due to audiophile bias. I know Mark Levinson had a TOTL preamp a while back that used a ton of ICs - but then some guy from Stereophile wrote an article...

I just finished a DIY preamp using Neurochrome Universal Buffers (with opamps) powered by an AMB sigma 22 power supply. Its the best preamp I have ever had discrete or otherwise. I've got some all discrete AMB alpha 20 circuits on the way to compare to the Neurochrome so we will see.

 

Somewhere in the head-fi Schiit story megathread there's a good chapter on the move away from discrete to ICs. Some of the reasons is that there is more development on ICs with newer and better devices. Why spend money researching and developing "big" components? Everything is moving to smaller form factors, and ultimately smaller becomes cheaper when the decreased cost of materials (both from size and fewer components) outpaces the miniaturization and complexity cost. A packaged system also presumably has optimized layouts inside, and are all shielded or so small that the interference/noise/gremlin pickup is negligible.

You go discrete for specific applications, like higher power/voltage/whatever that can't be handled by a small package, or a config that doesn't quite exist in an integrated circuit, but chances are you'd use some pieces of integrated circuits within the body of the discrete module anyways. And unless you've got a whole buttload of testing and analysis equipment, you're probably not getting the best out of those components that are slapped together according to some factory datasheet. A few mm of spacing on a lead could mean the difference between optimal behaviour or a deathly oscillation that cooks your circuit.

 

Out of curiosity, can the Neurochrome Universal Buffer directly drive IEMs? The spec says it has an output impedance less than 0.1 ohm, and the current requirements of an IEM are reasonably low.

Also, you said buffers - you have one on the input and one on the output?

 

Except in certain circumstances there’s no reason for discrete to be better on a 1:1 basis. By which I mean if we had the manufacturing capability to make custom integrated chips as easily and cheaply as we could do PCBs, no one would be doing discrete design. This also means if you get a datasheet for any op-amp and try to duplicate the circuit from the schematic discretely, it will not be better and often is worse.

The advantage discrete has is making certain designs that aren’t available from the options being manufactured. You can kind of see this with DAC chips as R2R chips were widely available at some point but as they stopped making them, DAC manufacturers started making their own discrete versions.

Other reasons have to do with thermal management. If you are trying to doing anything with Class A amplification you’re going to have issues with heat which can affect how your components behave. In an IC this means lots of heat sinking and fans however in a discrete circuit it often is fine air-cooled. I mean just look at vacuum tubes which are discrete but don’t need heatsinks.

One thing I don’t entirely understand is Class D as Bruno Putzey has said that discrete designs are better than chip based. Not sure if it has to do with minimizing high frequency switching noise

 

If we talk audio signal amplification, there aren't too many options on how to implement IC based circuits. You deal with very high gain devices and are forced to employ deep negative feedback to pull the gain down and get magical performance numbers. With discrete components you can optimize for open loop performance and use low to moderate NFB to linearize further.

With that said, nowadays the average opamp amplifier is pretty decent. There are nice sounding opamps for almost every application out there and plenty of people who know how to use them.

I think it has to do with high frequency circuit design where PCB design geometry starts to play a role.

 

I'm not really an IEM guy so this is not something I have thought about. If you send Tom at Neurochrome a question through his website, he answers all questions in seemingly seconds.

I have balanced Khozmo ladder attenuator boards feeding two separate Neurochrome UBs. The UBs are the output stage of my preamp.

 

I just remembered that PS Audio uses FET's in the output stage of their flagship BHK amplifier. According to Paul McGowan it's the closest thing to tube sound. He loves the tube sound but he is not a fan of tubes in the output stage because of the high impedance which would need get lowered with a transformer among other reasons like heat, size and wear.

Moral of the story, if he used FET's it's not a fully discrete design and he did that to get as close as possible to what in his opinion is the best sonic signature. So I guess, fully discrete isn't always better, depending what you are trying ro achieve.

I thank for instance if you are trying to run a DAC with 2 ESS9038PRO chips in dual mono and in currente mode full class A, I don't think it's possible to make an output stage that isn't at least hybrid. I don't think OP amps alone can meet these demands.

 

Discrete or IC better? Sorry, that's wayyyyyyyy toooooo complicated to answer.

• In what application?
• At what voltage level?
• Driving what load?
• At what efficiency?
• What level of complexity is tolerable?
• How much size you got available?
• What kinda heat can you dissipate?
• How good of an IC are you talking?
• What's the budget?
• How "discrete?" Do you allow servos? Or coupling caps?
  
• What kind of "discrete?" Tubes? Sand?
• How good of a discrete designer?

Simplistic reductions like "FET = tube" are not helpful, or true. They are very, very different parts.

Generalizations without understanding the engineering needs of a specific application are not helpful, either. The best I/V converter for a current-out DAC remains a discrete current-feedback stage, designed for inherently low input impedance, not relying on a billion dB of feedback to create the virtual ground, due to its inherent settling time advantage and minimal summing error.

Until you can look at an application like this in detailed engineering terms, weighing the constraints of size, cost, heat, type of devices available, possible effects of supply constraints, required power supply, filtering, and support circuitry, I'm afraid this question can't really be answered--and even then only for that particular application.

That said, oversimplifying:

• Discrete can still be tailored to a specific application in a much more customized way...but it is much, much more difficult than IC design, and requires more time and effort.
• ICs are better, in some cases much better, than they have been...but designing with them is kinda boring, because it becomes a game of "everyone grab the latest and greatest" and then everything is the same.

Of course, both of the above are my opinions, which are colored by my experience, and may be wrong. I am frequently wrong.

I should do a chapter on this.