ASR "Myth" regarding "discrete" circuitry...

[barrows]

I see it mentioned over and over on these forums, that the reason for poor performance of a given DAC on certain measurements is blamed on the use of "discrete" circuitry as opposed to using an IC design. I see this mentioned both in the case of DACs which may have a discrete converter architecture, and those which may have a discrete (or hybrid IC/discrete) output stage architecture.
I would suggest that anyone espousing this position go and read Amir's measurements of the Mola Mola Tambaqui DAC. The Tambaqui DAC uses both a discrete converter stage (moving average FIR filter, I believe) and a discrete output stage (no ICs). The Tambaqui also achieved the best measurements so far at this site. It is clear from these measurements that the use of discrete circuitry is not a cause of worse measuring performance, and it annoys me to read people believing such here. While it may be easier for a less accomplished Design Engineer to achieve good performance via IC based designs, it is clear, that if the Engineer has the skills, a discrete based approach can perform as well, or perhaps even better than, an IC based design.
For examples of other DACs with discrete design and good measurements one need look no further than Chord and dCS, both companies use discrete converters and output stages as well.
So guys, let's stop talking about how discrete circuits are a cause of worse measurements! Worse measurements are caused by designers who are either not good enough, or are not concerned so much with problems at < -120 dB.

 

[solderdude]

Tambaqui = $ 10k
Chord: Dave $ 10k
dCS > 10k

For that kind of money one should expect top quality, discrete or off the shelf chips shouldn't really matter at these prices.
BUT one also can buy crap for even more $.

Most of the dislike is for cheap gear that uses discrete DACs for the sake of it or uses discrete circuitry for the sake of it and don't mind the poor performance.
In general, in the lower price class off the shelf DAC chips and opamps offer better performance at the same price point.

 

[mansr]

It is clear from the Moolah Moolah and other measurements that achieving top performance with discrete circuitry is only possible at great expense. That is all.

 

[Joachim Herbert]

Veri made the point before.

 

[Juhazi]

https://audioxpress.com/news/akm-pr...lution-for-high-end-audio-product-development

"Asahi Kasei Microdevices Corporation (AKM) has expanded its lineup of premium audio Digital-to-Analog converters (DACs) with the development of a two-chip solution that separates the digital aspect from the analog. While conventional DACs use a single chip to convert signals from digital to analog, the new AKM approach splits the conversion process between the AK4191 delta-sigma modulator and the AK4498 DAC. This results in an improvement in sound quality that is easily perceptible while listening to music, the company states. "

"By using two ICs, AKM has minimized the effects of digital noise within the analog output, resulting in a perceived improvement of the ratio of signal to noise. This improvement, while not easily quantifiable via traditional measurement techniques, is easily experienced during controlled listening tests," the company states.
---
I can smell some money burning...

 

[barrows]

It is clear from the Moolah Moolah and other measurements that achieving top performance with discrete circuitry is only possible at great expense. That is all.

I know you do not really mean that! Chord Mojo and Qutest do quite well using discrete circuitry for much less money. But of course when a company has to actually pay a talented Design Engineer who can actually design a good DAC discretely, the cost of the product is going to reflect the increased development cost.

My only point in this thread is an attempt to stop folks from blaming "discrete circuitry" for poor measurements, as clearly, using discrete circuitry is not a problem in and of itself. The implementation is what matters.

 

[maverickronin]

Very few companies with the engineering chops to design a discrete DAC with performance commensurate with modern IC DACs have the desire to do so.

Very few companies with the desire to create a discrete DAC have the engineering chops to give it performance commensurate with modern IC DACs.

It's much easier to design around an existing DAC chip rather then design everything from the ground up so the problem is that (from the standpoint of objective measurements) they chose the wrong architecture for the job and that in theory putting the same amount of effort into a design based on an off the shelf DAC chip should yield better results. This is why discrete circuitry is "blamed".

In practice though they're likely to be basing what should be engineering decisions on belief in audiophiles myths or marketing to those who believe in audiophile myths and are quite likely to get poor to mediocre performance out of even the best chips with a sloppy implementation.

 

[barrows]

https://audioxpress.com/news/akm-pr...lution-for-high-end-audio-product-development

"Asahi Kasei Microdevices Corporation (AKM) has expanded its lineup of premium audio Digital-to-Analog converters (DACs) with the development of a two-chip solution that separates the digital aspect from the analog. While conventional DACs use a single chip to convert signals from digital to analog, the new AKM approach splits the conversion process between the AK4191 delta-sigma modulator and the AK4498 DAC. This results in an improvement in sound quality that is easily perceptible while listening to music, the company states. "

"By using two ICs, AKM has minimized the effects of digital noise within the analog output, resulting in a perceived improvement of the ratio of signal to noise. This improvement, while not easily quantifiable via traditional measurement techniques, is easily experienced during controlled listening tests," the company states.
---
I can smell some money burning...

Indeed, AKM is on to something here. But this is nothing new, separate DAC and DSP/oversampling architectures have been with us for a long time, like PCM 1704 and DF 1704, and many others. And many computer audiophiles are going one step further than this, and doing all their DSP/oversmapling in software, and just allowing the actual DAC to do the conversion only, think HQPlayer connected to a DAC which allows for direct DSD conversion, like the ADI 2, or a Holo Audio in NOS mode, etc.

 

[mansr]

I know you do not really mean that! Chord Mojo and Qutest do quite well using discrete circuitry for much less money. But of course when a company has to actually pay a talented Design Engineer who can actually design a good DAC discretely, the cost of the product is going to reflect the increased development cost.

My only point in this thread is an attempt to stop folks from blaming "discrete circuitry" for poor measurements, as clearly, using discrete circuitry is not a problem in and of itself. The implementation is what matters.

Foolishly using discretes in a misguided belief that ICs are inherently bad is often a cause of poor measurements. If you're not willing (or able) to put in insane amounts of development effort, you'll get much better results using ICs where someone else has already done the hard work.

 

[Fluffy]

In the Schiit Magni 3 review, it is shown that the IC version is better than the discrete version, both at the same price. Doesn't get much more apples to apples than that.

 

[barrows]

In practice though they're likely to be basing what should be engineering decisions on belief in audiophiles myths or marketing to those who believe in audiophile myths and are quite likely to get poor to mediocre performance out of even the best chips with a sloppy implementation.

The above is a massive assumption based on absolutely no actual facts or information. I actually work in the high end industry, and have never heard any talk about engineers or sales folks basing design decisions on "audiophile myths" or marketing in product development meetings.

 

[BDWoody]

My only point in this thread is an attempt to stop folks from blaming "discrete circuitry" for poor measurements, as clearly, using discrete circuitry is not a problem in and of itself. The implementation is what matters.

What if the attempt at utilizing discrete circuitry actually is the problem? Couldn't the 'discrete circuitry' be fairly blamed?

Have there been any 'all DAC's using discrete circuitry are bad' statements?
I'm guessing you have one of these...?

 

[barrows]

In the Schiit Magni 3 review, it is shown that the IC version is better than the discrete version, both at the same price. Doesn't get much more apples to apples than that.

All this proves is that the engineering of the discrete version is worse, not that discrete circuits are worse. Again, see the Tambaqui measurements.

 

[mansr]

My only point in this thread

... was to start an argument.

 

[Fluffy]

All this proves is that the engineering of the discrete version is worse, not that discrete circuits are worse. Again, see the Tambaqui measurements.

If at the same price one technology gets better performance, that technology is better. Don’t see a point paying extra for what you can get for less, just to boast that the design is discrete.

 

[barrows]

I'm guessing you have one of these...?

I would advise that you refrain from guessing... i have a few different DACs here, a Buffalo PRO (DIY with ESS 9038 and an IC based I/V stage OPA 1632), older version of the Buffalo with a discrete (only 4 MOSFETS total for a balanced DAC!) I/V stage. A DSC-2 based design which features a discrete converter stage (similar to what the Tambaqui uses) combined with an IC based output/gain stage (OPA 1612 and OPA 1632). And a Chord Mojo which I use for travel.
I am NOT saying: "discrete is better" or "IC is better", ALL I am saying is that using discrete circuitry is never a cause of bad measured performance. One can make chips measure poorly pretty easily as well, especially DAC chips with their high sensitivity to power issues on their Vref pin(s).

 

[BDWoody]

I would advise that you refrain from guessing

Gosh, thanks.
You've been just so...helpful.

A little too troll-ey for me...but I do hope you feel better now that you've straightened everyone out.

Have a nice day.

 

[mansr]

One can make chips measure poorly

Yes, but it is surprisingly difficult. Even the Ciunas monstrosity doesn't perform all that badly.

 

[amirm]

It is very common for discrete implementations to underperform IC ones. For this reason, it is safe to assume Discrete anything means it is less performant unless shown via measurements to be otherwise. After all, if a company is proud of what they have accomplished in discrete manner, they show it off with measurements.

As such, there is no myth here. It is just the probability based on data so far. And tendency of companies thinking lower performing audio gear with words "discrete" sell better.

 

[infinitesymphony]

I doubt most people here believe that any design topology is the One True Way. More often than not you see people rooting for designers to overcome typical problems in the parts they're using, ex. whether or not a DAC using an ESS chip overcame the "ESS hump." Shortcomings are acknowledged here.