I'll take them off your hand, if/when you feel gifting coming on! Ditto if you have TDA1547 and TDA1307. PM me.
BTW:
About Taobao ... almost impossible to use them OUTSIDE of China (even tho' they have options for UK/USA/etc, but no-go unless you're in PRC or HK)
Well, maybe the very early generation models were. John Atkinson's tests in Stereophile usually reported good jitter Measurements from the PLM series onward.
Shifting gears ...
I can only report and claim that my third-gen Philips TDA1540-based CD player -- purch'd early 1986, completely un-modded -- sounded more exciting and dynamic than my Dual CS-505/Ortofon OM-30 combo. I'm sure that CDP had poor jitter.
I still have a couple of unmodded vintage CDPs from late 80s, safely stored in my closet. And I bring them out every once in a while for a memory refresh. Yup ... that same exciting, dynamic sound that I don't think modern DS or R2R (discrete or IC) dacs can match.
Dreams and magic!
Not "zero-effort" if you're driving with road music coming out of your Land Rover's high-end streaming system.
Not "zero-effort" at home if the Tidal streamer is playing while you're efforting the bottle of ale to your lips.
I have an original Myryad MC100 which uses the 1bit CXD2565M before a later revision switched to a 24bit DAC.
On the circuitboard, the CD data stream goes to the CXD2565M and to the driver/pulse transformer for SPDIF.
On the circuitboard, the CD data stream goes to the CXD2565M and to the driver/pulse transformer for SPDIF.
Near zero. No discs to clean or finite life unobtainable laser to burn up. Just a mouse click or two or a screen to tap.
Hey, some some of the old dac chips like TDA1541 had a whole lot going for them in a perceptual sense, at least if well implemented. The old won't keep up with progress forever though. Eventually clearly better sounding stuff is going to come along. The question is when? I think we are there but the best is not cheap, which is the biggest remaining problem. Getting past the high cost factor would require us to move past current measurement standards. If we had measurements that better correlate with subjective sound quality, then engineers would presumably be more motivated to figure out a way to compete on those better correlating measurements. Problem is figuring out and agreeing on an updated measurement standard. Its not such a simple thing to do.
It's surprising that no other company has tried to FPGA dynamic-element-matching or even "chip it". Maybe DEM needs a lot of real-estate -- all the trace space of DIP form factor. It may be some of patent or copyright issue Philips had on DEM.
I don't think Philips ever released anything smaller than DIP size for 1541 ??? Even with the single-bit 1547 and the mid-1990s TDA1307 filter chip (both Philips' high-end digital audio devices) , Phliips stuck with DIP while their CC dacs were avail in SOIC (SMD).
Hmmm ... come to think of it, well into the age of SOIC/SMD (early 90s, henceforth), manufs with ICs like PMD100 and AD1862 stuck with thru-hole DIP.
ESS certainly chipped it, just they're doing it in software rather than hardware.
http://ispg.ucsd.edu/wordpress/wp-c...on-Why-Dynamic-Element-Matching-DACs-Work.pdf
Perhaps there is an opportunity there. You could pitch to the Dragons a plan to recreate the TDA1541A using the 5nm process. They will, of course, need much bigger tables for the piles of cash.
Some form of dynamic element matching is used in all multibit sigma-delta modulators, but it is quite different from the original dynamic element matching Philips invented long ago.
As a multibit sigma-delta DAC usually only has a limited number of different levels (say 16 for a four-bit DAC), they are usually made of that number plus one equally weighted unit elements. Some clever algorithm then selects the elements such that mismatch mainly cause ultrasonic errors. Data-weighted averaging was one of the first (and worst) such algorithms.
As a multibit sigma-delta DAC usually only has a limited number of different levels (say 16 for a four-bit DAC), they are usually made of that number plus one equally weighted unit elements. Some clever algorithm then selects the elements such that mismatch mainly cause ultrasonic errors. Data-weighted averaging was one of the first (and worst) such algorithms.
The reasons why no one ever uses the original version anymore is not difficult to understand: it requires lots of external capacitors that all cost money, the wire loops to and from those capacitors can all pick up interference, it requires a relatively large supply voltage that many modern IC processes can't handle and even with DEM for the higher bits, the matching requirements are quite tough.
Before insisting on updated measurement standards you should first prove that the existing ones are not adequate. One way to prove this is through controlled listening tests. Once there is surmounting evidence from such listening tests I'm sure the industry would be more eager to study the issue.
Do you think this hasn't been proved?
It is proven beyond doubt that some 2nd order harmonics under some conditions are hard to hear and that the higher orders are easier to hear, which significance is further obscured by the fact that both spl and harmonic profile of the distortion matter.
Given none of these things are ever considered as a whole, it seems to me that adequacy can be questioned.
Given I guess we don't have enough data, it is hard to implement and we rather have raw, blind and dumb numbers (lowest THD beats higher THD etc.) this is where we are.
But I'd say there certainly is proof.
Btw Gedlee introduced the Gedlee Metric because of this.
It is proven beyond doubt that some 2nd order harmonics under some conditions are hard to hear and that the higher orders are easier to hear, which significance is further obscured by the fact that both spl and harmonic profile of the distortion matter.
Given none of these things are ever considered as a whole, it seems to me that adequacy can be questioned.
Given I guess we don't have enough data, it is hard to implement and we rather have raw, blind and dumb numbers (lowest THD beats higher THD etc.) this is where we are.
But I'd say there certainly is proof.
Btw Gedlee introduced the Gedlee Metric because of this.
Can you point to links of controlled listening tests which prove perceivable differences between well regarded and reasonably well measuring DACs?
Have you tried Gedlee metric on any dac?
Edit: I have implemented several DACs with various technologies (DS, Multibit, OS, Non-OS). Some very well measuring, some reasonably well measuring. IME once the levels are matched to within 0.1dB and sighted observation eliminated the differences become very small in AB testing. That is why I take all subjective sighted listening results with a large grain of salt.
Have you tried Gedlee metric on any dac?
Edit: I have implemented several DACs with various technologies (DS, Multibit, OS, Non-OS). Some very well measuring, some reasonably well measuring. IME once the levels are matched to within 0.1dB and sighted observation eliminated the differences become very small in AB testing. That is why I take all subjective sighted listening results with a large grain of salt.
Last edited:
Interesting, Philips continuous calibration already used n + 1 equally-weighted unit elements (for a different reason, but still).
Last edited:
Sure.
TDA1541A (and AD1862) was one of the best ancient DAC chips.
It was so good, as TDA1540 and TDA1543 were terrible....
IME it means the limiting factor is the rest of your system besides the dacs. For example, are you using ESL speakers for the listening tests?