My First Chip! PCM5102 112dB Stereo DAC

[Rochey]

Folks,

just wanted to share something I'm very very proud of. The first DAC (Digital to Analog Converter) that I defined and worked on at TI.

http://focus.ti.com/docs/prod/folders/print/pcm5102.html

I'm beaming with pride -- It's a 112dB Stereo DAC, with an integrated negative rail charge pump and line driver, so you don't need any opamps after it (a simple RC Low pass filter and your off to the connector).
In addition, there's a fancy PLL involved that'll autodetect your I2S rate, configure the device, and generate it's own internal master clock - so no need for 24.576MHz clocks in your product!

There's some other smarts in there too - it uses a next generation architecture based on the PCM1792 (The flagship 132dB DAC that TI has)... and more...

I've done multiple, multi-hour listening tests in Japan with customers on this one... and it sounds incredible

Sorry -- little over enthusiastic... just REALLY excited.  

Don't be suprised if Expat Audio has a DAC card soon

[Rob Flinn]

congratulations Rochey, very cool !!!

[V9977]

Holy moley!

Heart-felt good wishes.
Hope it gets everywhere on the globe.

[zayance]

 

[Rochey]

thank you gents... certainly very proud here.

[pucho812]

  that is awesome. Congrads. I believe a round of guinness is in order.

[Skylar]

Here you are, sir:

[jackies]

Congrats, that's a feat for sure!
Now, a special group buy is in order?!?

[radiance]

Congrats! 

May I ask, what a negative rail charge pump is?

[Rochey]

Sure -- the device takes +3V3, then generates, using a charge pump, a -3V3 -- which is then used by the line driver.

That way - you get a ground biased output, set to 2.1Vrms -- or roughly 5.6Vpp.

/R

By the way - thanks everyone for the great feedback! 

[JohnRoberts]

Kool...  congrats..

Not sure I follow the full benefit of the charge pump inside. Sure it saves DC blocking caps on output, but you need to use caps for the charge pump so that's kind of a wash (albeit not the same kind of caps). Having 6dB more output swing is always useful. 

I've been looking at some of cellphone codec parts for interesting medium Fi applications.

JR

[abechap024]

That is awesome! Congrats

[Rochey]

John, transient and muting pop and click is one of the thing the ch.pump addresses.

No need for muting transistors following the opamp.

[Kingston]

Congratulations for taking a fresh new step in the evolution of high end DACs. Now let's have some incredibly small scale but affordable project utilizing this.

[abechap024]

I would also second a DAC DIY project with this I've needed to upgrade my stereo DAC for a while.

[sodderboy]

Thats slick!  Do you have a pic of the first wafer or sumpin?
Mike
PS: Spoken as a shareholder.

[JohnRoberts]

John, transient and muting pop and click is one of the thing the ch.pump addresses.

No need for muting transistors following the opamp.

Sweet... when they came into Peavey for their regular dog and pony shows I tried to tell them that the balanced output drivers they were flogging needed turn on/off mutes and drive capability to the full unregulated rails.

Glad they finally listened to somebody, at least about turn on/off thumps..

JR

[PRR]

> benefit of the charge pump inside. Sure it saves DC blocking caps on output

The customary level is still -10dbV on VU meter; taking 16db headroom gives 2V rms or 5.6V peak-peak. Unbalanced. It may be time for a smaller standard, but who will be first to offer a "wimpy" output?

System designers insist on silly supplies like 3.3V.

The least-distasteful way to get 5.6Vpp from 3.3V supply is a rail-doubler.
______________________________________________________________

http://focus.ti.com/lit/ds/symlink/pcm5102.pdf

Page 1: "Compared with existing DAC technology, the PCM510x family offers up to 20dB lower out-of-band noise, reducing EMI and aliasing in downstream amplifiers/ADCs. (from traditional 100kHz OBN measurements all the way to 3MHz)"

Punctuation typo: period after "ADCs" should enclose the parenthetical clause.

PDF glitch: when trying to select the right column (1st page) I get the left column also.

Pg 25: "...reproduce virtually all frequencies through to it’s maximum sampling rate of 384kHz."

No apostrophe in the possesive of "it". (I no u don't no, but someone in TI ought two.)

And "virtually" is a red-flag weasel-word. What do I really get? I bet it is -1db at 75KHz -3db at 150KHz. Really, not virtually.

Pg 23: "The XSMT input pins voltage range is from -0.3V to DVDD + 0.3V. The ratio of external resistors must be considered within this input range."

English: This is really awkward phrasing. "considered within"??

Engineering: 20% over "ideal" causes damage?? Houses, bridges, meters, and gauges are designed for 2X to 3X over nominal stress without damage. Stuff happens. You can't put a 100K resistor before the input diodes? Then change the threshold to some lower voltage.

[MeToo2]

Congratulations Rochey. Cool. 

Kool...  congrats..

Not sure I follow the full benefit of the charge pump inside. Sure it saves DC blocking caps on output, but you need to use caps for the charge pump so that's kind of a wash (albeit not the same kind of caps). Having 6dB more output swing is always useful.  

I've been looking at some of cellphone codec parts for interesting medium Fi applications.

JR

The charge pump cap on the power supply is not passing any audio.

Other alternative of course would be to use ±3,3V supplies, but these days mono rail supplies are becoming the norm.

> benefit of the charge pump inside. Sure it saves DC blocking caps on output

The customary level is still -10dbV on VU meter; taking 16db headroom gives 2V rms or 5.6V peak-peak. Unbalanced. It may be time for a smaller standard, but who will be first to offer a "wimpy" output?

System designers insist on silly supplies like 3.3V.

The least-distasteful way to get 5.6Vpp from 3.3V supply is a rail-doubler.
______________________________________________________________

You could be right about the need for a new standard interface level.

AFAIK the choice of 3.3V supply is driven more by miniaturization, and thus cost and reliability and power efficiency, than anything else.

I don't think there's any real choice but to reduce the power supply voltage to these things, or else risk dielectric breakdown, or frying the chips at high switching frequencies, or having to use more / bigger/ less efficient batteries than a single Li-ion.

Swings and roundabouts. Things may end up with a lower noise floor, but then less headroom at the top end. This is a trend that has been going on for many years (since the move from tubes to transistors to IC's). I quite like the sound of ~200V DC supplies personally. But don't try running many transistors or IC's on that sort of supply

Generally the lower noise floor at the bottom end has delivered a greater benefit than the loss of headroom at the top end.

[JohnRoberts]

Congratulations Rochey. Cool.

Kool...  congrats..

Not sure I follow the full benefit of the charge pump inside. Sure it saves DC blocking caps on output, but you need to use caps for the charge pump so that's kind of a wash (albeit not the same kind of caps). Having 6dB more output swing is always useful.  

I've been looking at some of cellphone codec parts for interesting medium Fi applications.

JR

The charge pump cap on the power supply is not passing any audio.

Other alternative of course would be to use ±3,3V supplies, but these days mono rail supplies are becoming the norm.

Thinking about this some more, I am warming to the concept. My first impression was that the chip would be interfacing with more circuitry, that would require bipolar supplies and output treatment anyhow, so the charge pump would be redundant, but upon reflection a bunch of these applications may be generating actual audio line level outputs for the outside world, so the negative supply has utility.

I withdraw my complaint.


JR