PCM1732 (HDCD inside)

[thatsright]

This DAC chip, first appearing in 1999, kind of reminds me of the classic "Intel Inside" logo.

PCM1732 PDF

Part #: PCM1732. Description: 24-Bit, 96kHz, Stereo Audio DIGITAL-TO-ANALOG CONVERTER With HDCD Decoder. File Size: 176.0 Kbytes. Manufacturer: Burr-Brown (TI).

pdf1.alldatasheet.com

It has been used in several disk players (see Dutch list), and has built-in HDCD encoding. The only non-Pacific Microsonics device to do so. The PMD100/200 is often said -- especially in the DIY community -- to be the "best-sounding" stand-alone DF. If you're in doubt about the desirability of these chips, check out eBay and Ali prices on the PMD100/200 sometimes!
But could Burr-Brown actually shrink down that whole PMD100 chip -- in its entirety -- onto even a smaller wafer along with "ENHANCED MULTI-LEVEL ∆Σ DAC" and LPF?
The PMD100 was avail in TQFP44 and DIP-28. The PMD200 was 100-pin TQFP.

 

[voodooless]

But could Burr-Brown actually shrink down that whole PMD100 chip -- in its entirety -- onto even a smaller wafer along with "ENHANCED MULTI-LEVEL ∆Σ DAC" and LPF?

Sure they could, but who’s going to buy it?

Also I’m not sure what bits of the PDM actually made it into the PCM1732. I think it may only be the DHCD decoder.

 

[thatsright]

Sure they could, but who’s going to buy it?

I noted the Dutch list in my OP

https://www.dutchaudioclassics.nl/the_complete_d_a_dac_converter_list

I count 42. There may be more not on that list.
Also I bought a handful of those ICs, for DIY, maybe 2008/9 . And there are Sellers and Buyers on Ali and eBay.

 

[voodooless]

Also I bought a handful of those ICs, for DIY, maybe 2008/9 . And there are Sellers and Buyers on Ali and eBay.

That’s hardly enough to justify redesigning the chip for a smaller process node. Especially the analog bits will cost significantly.

 

[thatsright]

That’s hardly enough to justify redesigning the chip for a smaller process node. Especially the analog bits will cost significantly.

I don't understand what these sentences mean. What do they mean?

 

[MRC01]

Years ago I had a Rotel RCD-1070 which used the PCM-1732. It was nice but nothing special. Whether I was hearing the chip or the player's implementation, who knows? Why would people get excited about that chip?

OTOH, my Tascam DA-3000 recorder uses the PCM-1795 and it does sound slightly different from 2 other DACs I have handy (which sound much closer to each other, one using WM8741 the other using ESS9028). The Tascam sounds slightly warmer, less edgy. For no obvious reason in simple measurements like FR and distortion. Maybe it's because the Tascam has pro level balanced outputs at about 16 V or about 12 dB louder than consumer 4 V? That means matching volume uses less gain, which means all else is not equal.

 

[voodooless]

I don't understand what these sentences mean. What do they mean?

Well, you want the stuff in a smaller package. To do that you can’t simply shrink the design for a smaller process node. Significant parts of the chip will need to be redesigned for the smaller size and new manufacturing processes. This takes copious amounts of time and money. If only a handful of people buy these chips, that’s not something BB would likely invest in.

 

[thatsright]

Well, you want the stuff in a smaller package. To do that you can’t simply shrink the design for a smaller process node. Significant parts of the chip will need to be redesigned for the smaller size and new manufacturing processes. This takes copious amounts of time and money. If only a handful of people buy these chips, that’s not something BB would likely invest in.

Yup ... and I think others have implemented HDCD in DSP (fpga). And Windows can do this natively.
Rather, I was hoping that Burr-Brown had somehow managed to shrink down the whole topology of the orig. PMD100. It was the non-HDCD part of that PMD design that was part of its goodeness. There are some designers that think they can SPICE the orig design and replicate it in firmware or DSP. Wham bam thank you ma'am. Forgetting about glass gaps and conductor thickness and parasitics.

About the 1732 ... look closely at that Dutch list. The CDPs using it are not cheapies.

 

[voodooless]

There are some designers that think they can SPICE the orig design and replicate it in firmware or DSP

That should be simple: capture the impulse response and use it to build the filter. Any modern DSP can do it.

I doubt though that it will be any better than what is already inside any SOTA DAC.

 

[MAB]

But could Burr-Brown actually shrink down that whole PMD100 chip -- in its entirety -- onto even a smaller wafer along with "ENHANCED MULTI-LEVEL ∆Σ DAC" and LPF?

I think you mean more advanced process node, not smaller wafer.
Also, analog circuits don't benefit from process node shrink like digital circuits.
As you mention, FPGA can emulate. No way it will be cheap, but many orders of magnitude cheaper than getting a contract foundry to manufacture an antique chip on a modern process. Plus, it will take a costly design effort to port the old design to the new process. And, older process nodes are often more optimal for analog... with precision passive components, transistors optimized for analog, and analog-friendly design tools and collateral.
Why you would want to do this is another story. The absurd prices of these old chips on eBay kind of reminds me of the high prices of oddly shaped Cheetos, or old bicycles. People buy them to have them. And if they did actually DIY one, not going to sound different unless they mess up the implementation.

 

[AnalogSteph]

About the 1732 ... look closely at that Dutch list. The CDPs using it are not cheapies.

Same generally goes for the CS4397, the other consumer-level DAC with a HDCD decoder.

These days it's all moot anyway, use foo_hdcd and be done with it.

 

[thatsright]

Same generally goes for the CS4397, the other consumer-level DAC with a HDCD decoder.

The CS4397 does not have an integrated HDCD decoder. Only the PCM1732, as non-PMD monolithic device, and per the subject of this thread, has that feature. Correct me as needed.

 

[thatsright]

I think you mean more advanced process node, not smaller wafer.
Also, analog circuits don't benefit from process node shrink like digital circuits.
As you mention, FPGA can emulate. No way it will be cheap, but many orders of magnitude cheaper than getting a contract foundry to manufacture an antique chip on a modern process. Plus, it will take a costly design effort to port the old design to the new process. And, older process nodes are often more optimal for analog... with precision passive components, transistors optimized for analog, and analog-friendly design tools and collateral.
Why you would want to do this is another story. The absurd prices of these old chips on eBay kind of reminds me of the high prices of oddly shaped Cheetos, or old bicycles. People buy them to have them. And if they did actually DIY one, not going to sound different unless they mess up the implementation.

Yes, technology node, https://en.wikichip.org/wiki/technology_node as needed. Unless the orig PMD100 footprint was small enough to "saw off" in its entirety, onto the 1732 architecture.
I generally prefer discrete analog architecture (discrete tube or transistor vs. opamps) -- after the DAC for output buffering and LPF.
Discrete digital can be a bit weird. I don't care for the sonics of many discrete R2Rs I have heard (e.g, Denafrips). All those ladder components spaced so far apart might give that architecture characteristics some find desirable. But not me-- in any I've heard to date. Curious about MSB and Weiss , however.
I might also say the same for analog processing inside the DAC chip. E.g., typical voltage out of modern Delta-Sigma means they have opa's inside--possible a compromise due to parasitics and puny trace dimensions . Also not sure about all that DSPing (DFing and SRCing) inside the same DAC chip, as many modern all-in-ones offer.

I like external ASIC upsamplers (SRCs), 74-logic based re-clocking, and asic DFs. I'm also very open to "ancient" devices: well done-parallel input implementations (PCM53-55) and external S/H (PCM53-56). Yes, I think 16-bit Red Book sound may have peaked in 2nd and 3rd gen CD players. See my recent NEC thread elsewhere on ASR.

 

[thatsright]

That should be simple: capture the impulse response and use it to build the filter. Any modern DSP can do it.

I doubt though that it will be any better than what is already inside any SOTA DAC.

I came from the manuf. industry. Where Bosch, Siemens and Allen Bradley could afford to DSP or ASIC . Robots, factory automation, etc. And even for the cheapest, lowest-cost devices, ASIC was usually best as it was much faster and more reliable.
TI, AD, Cirrus/Crystal, Philips, NEC, et al, put a lot of R&D and $$ into ASICs.
And now johnny-come-latelys , like current hottie one-man-show Rob Watts -- and the in-toe Chord fanboyz -- claim a $12K million-tap fpga/dsp device is the cat's meow? Well maybe. At $12k.
EDIT:
About Watts ... his earlier work:
Rob Watts started with DPA -- Deltec Precision Audio -- and early models from that outfit used Philips chipsets. And then later DPA, moved on to PA. See:

http://www.acoustica.org.uk/DPA/deltec.html

I have no idea what the output of DPA DACs were like. Methinks: discrete solid state ???
Around the same time, another high-end manuf was also using Philips Bitstream chipsets: Jadis. But the $12k (1994) Jadis JS1 used tubes. It measured poorly, but got good reviews. There's gold in them that hills.