World's longest-serving audio DAC and ADC chips

[AnalogSteph]

I thought I'd do some poking around regarding what the oldest converter chips are that are still very much active today (if not necessarily in the same iteration and the same process). Here's what I found:

DACs:

1. AD1856 (1988ish) - the uses for a classic 16-bit R-2R multibit DAC may be limited these days unless perhaps you're an NOS DAC enthusiast, but you can still buy one 35ish years in (limited grade selection and SOIC only, but still)
2. AD1851 (1990ish) - another 16-bit R-2R job that's still active if you need one
3. PCM1728 (1998-06) - the 24/96 DAC, a stripped-down PCM1716, is still alive and well
4. PCM1737/PCM1739 (1999-11) - these early 24/192 DACs are still around as well
5. AD1852 (2000) - this upper-midrange 24/192 DAC is marked NRND (not recommended for new designs) so probably won't be around much longer, but still
6. PCM1748, PCM1742 (2000-12) - the PCM1728 relative with a lesser filter and its 24/192 cousin are still around as well.
7. PCM1738 (2001-01) - this early high-performance 24/192 DAC is still active
8. CS4392 (2001-05) - another classic 24/192 job you can still buy

Honorable mention: CS4334/5/8/9 (1998-2022, what a run for a little consumer 8-pin 24/96 DAC)

And we haven't even gotten to the 2002/3 wave of DACs yet!

ADCs:

1. AD1877 (1994?) - I have no idea who even used this consumer 16-bit delta-sigma ADC back in the day, perhaps this is why it's still active (I guess AD's definition of "active" = "we still have a bunch sitting on the shelf"); can anyone confirm that this really came out in 1994, it seems 1-2 years late?
2. PCM1801 (1999) - another modest consumer 16-bit single speed ADC that's still around
3. PCM1802 (2001-11) - a not quite as modest 24/96 job
4. PCM1804 (2001-11) - an early midrange 24/192 ADC that used to be fairly popular (e.g. EMUs pre-M)
5. AD1871 (2002) - another 24/96 part, NRND now
6. CS5361 (2002) - a classic midrange 24/192 ADC
7. CS5351 (2002) - a consumer ADC, now marked NRND
8. CS5381 (2002-12) - a classic high-end ADC
9. PCM4202 (2003) - another classic high-end ADC

Honorable mentions: AK5393 (1999-01 - late 2018), CS5340/CS5341 (2003 - late 2022), AK5394A (2002 - late 2018, the long-time ADC reference)

Codecs:

• PCM2904/PCM2906 (2002-06) - these classic 16-bit USB codecs are still kicking
• CS4272 (2003-03) - an absolute classic that is found in tons of audio interfaces to this day with no end in sight (more or less a CS5361 + CS4392 combo)

Honorable mention: CS4270 (2005 - late 2022) - yes, the low-end 24/192 codec that's still gracing some USB audio interfaces today (e.g. smaller M-Audio AIR 192, Steinberg UR22C) is headed for retirement, with a last time buy of December this year

 

[McFly]

Interesting list. People say certain DACs are legendary, but if they were, you'd still be able to buy them new today.

The AD1856, perhaps therefore truly legendary? Or does it have a use in some medical machines that are being made serviced in 3rd world countries or something?

 

[r042wal]

FPGA is the way to go because you can program it with code

 

[Blumlein 88]

I think the AD1877 was used in those CDR disc recorders. Also some early devices that let you record music to a hard drive and later into memory chips. The sort of thing a musician might have to record in his home.

 

[MCH]

They are not DACs, but I want to honor here the wm8804/5 i2s/spdif transceivers, released around 2007 I believe and about to be discontinued in 2024 :'-(

 

[dualazmak]

How about Burr-Brown PCM1795?

I still keep and occasionally use ONKYO DAC-1000(S) (launched in December 2010) which has two PCM1795, one for L-Ch, one for R-Ch; still sounds really amazingly nice, up to 192 kHz 32 bit. The USB ASIO driver works fine on Windows 11 Pro.

DAC-1000(S) was/is really almost professional-use-oriented very heavy duty DAC unit with various input modes (ref. here and here.)
If you would be interested, I just found very recent intensive review series article consists of 4 posts; I hope your web browser would properly translate it into English;
https://www.audiostyle.net/archives/51875039.html

 

[dfuller]

• CS4272 (2003-03) - an absolute classic that is found in tons of audio interfaces to this day with no end in sight (more or less a CS5361 + CS4392 combo)

shame about its garbage ultrasonic performance.

 

[AnalogSteph]

They are not DACs, but I want to honor here the wm8804/5 i2s/spdif transceivers, released around 2007 I believe and about to be discontinued in 2024 :'-(

Actually they may be even older than that, I first found them listed in Wolfson's 2005 product selection guide. It doesn't look there was any public-facing datasheet until 2007 though, and the AES paper is from 2006. These may have inspired ESS to incorporate a DPLL into their DACs.

How about Burr-Brown PCM1795?

That's almost to be filed under "sprightly youth", it only came out in 2009! Mind you, it only was a 32-bit update of the classic (24-bit) late-2003 PCM1796 that was part of the 2002/3 "wave" I mentioned (AD1955, PCM4104, CS4398, PCM1792/94, PCM1796/98).

shame about its garbage ultrasonic performance.

Well yeah, it probably is a modulator with few bits but high-order noise shaping (like 5th order), more on the traditional / economy side of things. Being shared with the CS4392 from 2001, it's a fairly old design. Classic Cirrus/Crystal datasheets of the mid-'90s would routinely discuss out-of-band noise and its mitigation (e.g. CS4329/27, CS4303 <-- if you think the CS4272 DAC is tough, you haven't seen that one), unfortunately they stopped doing so around 1999.

 

[muslhead]

FPGA is the way to go because you can program it with code

Even if you did have the "code" to program it, an fpga is a digital device while everything above would require at least some analog functionality. As such, unless the fpga is either analog or a mixed mode digital device, it is not the way to go. In addition, from a cost standpoint, fpgas are very, very cost ineffective way to implement anything that is made in volume (there is software that will allow you to optimize the design but it will never compete.

 

[r042wal]

Even if you did have the "code" to program it, an fpga is a digital device while everything above would require at least some analog functionality. As such, unless the fpga is either analog or a mixed mode digital device, it is not the way to go. In addition, from a cost standpoint, fpgas are very, very cost ineffective way to implement anything that is made in volume (there is software that will allow you to optimize the design but it will never compete.

I suggested that becaise my PS Audio DAC is FPGA and I noticed several other companies going that direction. It looks like you have more flexability but your point is well taken

 

[AnalogSteph]

As long as other DACs using boring off-the-shelf chips are running circles around the results in the performance department (and they do), it's all a bit of a moot exercise. That's what you get when boutique gear is trying to be different for the sake of being different. Not like something like the Pacific Microsonics Model Two, where all the effort was actually needed to achieve its superior performance.

 

[Prana Ferox]

Interesting list. People say certain DACs are legendary, but if they were, you'd still be able to buy them new today.

The AD1856, perhaps therefore truly legendary? Or does it have a use in some medical machines that are being made serviced in 3rd world countries or something?

You generally can't die-shrink an analog circuit and expect it to maintain the same electrical characteristics or be drop-in interchangeable, so even if it's the same basic circuit the manufacturers would have new part numbers. A lot of 'legendary' chips were made on wildly obsolete processes requiring fabrication tools that flat out don't exist any more. Analog signal chips tend to stick around longer (and keep wildly obsolete fabs going longer) but you have to sell absolute mountains of them for it to be worthwhile, especially considering many of these chips go for pennies on the taperoll, legendary or not.

 

[AnalogSteph]

I suspect that the likes of CS4272 and CS5361 have been ported to a new process at least once. The guys at AKM will definitely have some experience with that as well after the fire fiasco.

The ADC design that I suspect to be the CS5361's "occupies 5.62-mm2 active area in a 0.35-µm double-poly, three-metal CMOS process" - yes, 350 nm! For perspective, later 3.3 V Pentiums (P54C) and Pentium MMX were made on a 0.35 micron process node. It would have been only a few years old (if still outdated) back in 2003, but qualify as downright geriatric nowadays.

 

[DWPress]

Looking to add another Roon zone for my pressroom in-wall speakers I dug out an ancient Burr-Brown Japan PCM2702 DAC from the closet. Happy to report that it is still recognized under Mac OS 14.2 Sonoma. I don't care how well it measures, the Sonace in-wall speakers are just ok and often have the sound of machinery to compete with.

 

[Zapper]

I suspect that the likes of CS4272 and CS5361 have been ported to a new process at least once. The guys at AKM will definitely have some experience with that as well after the fire fiasco.

The ADC design that I suspect to be the CS5361's "occupies 5.62-mm2 active area in a 0.35-µm double-poly, three-metal CMOS process" - yes, 350 nm! For perspective, later 3.3 V Pentiums (P54C) and Pentium MMX were made on a 0.35 micron process node. It would have been only a few years old (if still outdated) back in 2003, but qualify as downright geriatric nowadays.

Yeah but that's a 5V chip. To operate at 5V you need channel lengths of 0.6um or so. There is no point in going to more advanced lithography if you want typical audio voltage levels.

These "geriatric" processes are actually still used a lot, and new chips are being designed in these processes. The 0.35um BCD processes aren't much like the digital processes used for the Pentium. Some have a rich analog tool kit - precision low tempco SiCr resistors, MOS, MIM, and MOM caps, high performance NPN and PNP transistors, zener diodes, high voltage LDMOS FETs and Schottky diodes.

Another aspect of "geriatric" analog processes is that analog doesn't scale well. Precision requires area, in things like FETs, bipolars, resistors, and capacitors. As the demand for precision goes up, these devices have to get bigger, not smaller.

The most modern analog technologies are now a superset of several technologies, to integrate the analog capabilities with more advanced logic. They may have dense-ish 1.2V logic from a 45nm or 65nm node, 1.8V devices from 180nm, 5V devices dating to 0.6um, and a range higher voltage LDMOS devices, some capable of withstanding 120V or more for various power applications. This is the type of process you want if you're going to make an single chip Class-D amp with integrated DSP, for example.

 

[Zapper]

AD1856 (1988ish) - the uses for a classic 16-bit R-2R multibit DAC may be limited these days unless perhaps you're an NOS DAC enthusiast,

Are there any modern measurements of these old DACs? My web searches yielded nothing. I'm kind of skeptical that there has been much meaningful (i.e. audible) improvement in DACs since these were new. Sure, the specs are more impressive: more bits, higher data rate, higher SiNAD, more gilding the lillies. But I haven't heard a noticeable improvement in DACs since good quality 16bit 4x oversampling DACs like the AD1856 were introduced.

 

[Zapper]

especially considering many of these chips go for pennies on the taperoll, legendary or not.

Some of them go for a lot more than that. I just saw LF156 JFET op-amps on Digikey for $38.77 each in a TO-99 metal can package. Probably for military and aerospace in limited quantities. The old LT1785 RS485 chip goes for $3.32 each for 2500. That's still made with something like a 2um bipolar-only process, and it still sells in high quantity. I'm sure I could find a lot of other examples. A lot of old analog chips meet a need, and the need isn't changing, so they keep selling.

 

[JSmith]

Are there any modern measurements of these old DACs?

Datasheet, pages 10 & 15;

https://www.analog.com/media/en/technical-documentation/data-sheets/ad1856.pdf

JSmith

 

[Zapper]

Datasheet, pages 10 & 15;

Thanks. That distortion is fine for me.

I was thinking of all the measurements one can do with an audio analyzer - IMD and multitone and fft spectra, etc.

 

[AnalogSteph]

The most modern analog technologies are now a superset of several technologies, to integrate the analog capabilities with more advanced logic. They may have dense-ish 1.2V logic from a 45nm or 65nm node, 1.8V devices from 180nm, 5V devices dating to 0.6um, and a range higher voltage LDMOS devices, some capable of withstanding 120V or more for various power applications. This is the type of process you want if you're going to make an single chip Class-D amp with integrated DSP, for example.

It all makes sense, thanks. I assume squeezing all of these processes with their different requirements onto a single chip isn't necessarily easy or widespread among fabs, hence why AKM would have opted for the AK4191 + AK4499EX split. It's the first case of separate digital filter and DAC I can think of in over 20 years... the last ones were pretty much killed off by high-performance (120dB class) single-chip DACs at the end of the '90s. Anyone know what happened to NPC? Well, this, apparently:

History | Seiko NPC Corporation

You can see the history of Seiko NPC Corporation.

www.npc.co.jp