Another Variation on the DAC Question

[oleg87]

This was before my time, but didn't they add preemphasis to some early discs to compensate for the filters on those DACs being absolute crap?

 

[Chrispy]

I've not particularly vigorously compared various optical disc players over the years (having gotten my first one in the mid 80s) as I've not felt there was any difference to flesh out. I might still have an old cd changer from the 90s, and a few dvd players, but mostly I use newer bluray players. I've used their onboard dacs as well as dacs external to the players and just scratch my head with all these who hear differences in the dac particularly. Perhaps people thought they needed to fuss with a dac as much as a vinyl playback system or something? I heard definite improvements in a variety of ways over vinyl altogether, tho.

 

[Mr. Widget]

My ears are human. I claim no exceptional skills.

Glad to know that we share that... both human and both without exceptions skills. (At least as far as hearing goes.)

The constant sense that the needle might leap from its groove in response to some low frequency information certainly added a “liveliness”. Forgive me, but I think your criticisms of digital are more likely a fondness for analogue artifacts or simply imagined. But you could always devise some tests to figure it out. Let us know if you do.

For decades I thought that it was very possible that the wide and deep soundstage that I enjoyed was due to just that... distortions (artifacts) caused in the vinyl playback chain.

What stumps me is that my current digital and analog sources sound very similar today while that has not always been the case.

Yes, I do want to perform a controlled test, because I find this area very interesting. My plan will be to track down a properly operating 80s era CD player and cue up two identical copies of the same CD and perform a blind level matched comparison between that player and its internal DAC and a contemporary transport/DAC.

 

[Mr. Widget]

I heard definite improvements in a variety of ways over vinyl altogether, tho.

No argument there. Vinyl and even 15ips tapes have huge issues.

 

[kemmler3D]

My issue was with the stereophonic imaging and to a lesser extent to the "liveliness" of the sound.

Fair enough, but to me those sound like symptoms of analog defects, hard to imagine how you can impair stereo imaging if both DAC channels are running on the same clock and you have flat FR.

Now, crosstalk or problems with channel matching in the amplification section could cause those.

The progress in DACs you hear over the years is perhaps more easily explained by falling costs of digital chips over time than progress in conversion. As the cost of a competent DAC chip falls, the budget for analog parts increases proportionally...

 

[Mr. Widget]

Fair enough, but to me those sound like symptoms of analog defects, hard to imagine how you can impair stereo imaging if both DAC channels are running on the same clock and you have flat FR.

Now, crosstalk or problems with channel matching in the amplification section could cause those.

When I first entered the digital world with that first CD player I thought that crosstalk or issues caused by not having the cartridge absolutely perfectly mounted could be the cause of this difference, but then as the gulf between analog and digital shrank as my digital sources improved kind of messed with that theory.

The progress in DACs you hear over the years is perhaps more easily explained by falling costs of digital chips over time than progress in conversion. As the cost of a competent DAC chip falls, the budget for analog parts increases proportionally...

In addition to the lowering of costs of digital, in my personal set of circumstances as my income level increased and I became involved in the industry I have gone from absolutely entry level gear to extremely high end prestige gear. (That said, I am not suggesting you need a $20K DAC to hear these changes.)

 

[restorer-john]

That is exactly the point of my post and what I am curious about.

If it turns out that even an early generation digital source is indistinguishable from a modern DAC, then we have 40 years of people fooling themselves... an interesting possibility.

It's very, easy to determine the measureable differences between a 40 year old player and a modern standalone DAC. But once you feed your modern DAC with the same content that a CD player uses (16bit) those 'massive' differences become very small and irrelevant in an audible sense. Most of you listening to digital streams are listening to 16 bit content, whether native or upsampled/remastered whatever. Testing DACs at 24 bit to tease out tiny differences for a SINAD leaderboard is well and good, but it isn't representative of how they are used.

The CD player plays 16 bit content, and they very quickly topped out at the limit of what was theoretically achievable with that content by around 1989 or so. The first few years of CD development were exponential IMO, and little has been achieved with 16 bit content in the four decades since.

My statements are about what, if any, audible differences there are between a 40 year old player vs a modern DAC (playing the same 16 bit content derived from a common CD) with identical level matched content.

Let's compare apples with apples. Take a CD we likely all have, and one that is regarded as technically excellent and pick a track excerpt small enough we won't get in trouble for using it for testing purposes. I have plenty of CD players, thousands of CDs and a few decent interfaces for capturing the analogue outputs.

Yes, I do want to perform a controlled test, because I find this area very interesting. My plan will be to track down a properly operating 80s era CD player and cue up two identical copies of the same CD and perform a blind level matched comparison. between that player and its internal DAC and a contemporary transport/DAC.

I've done this so many times myself. With old players vs new, DACs vs CD players etc. Pretty much all my goto 'test' CDs I have multiple copies just for that purpose. I honestly wish there were audible differences because that would be a nice justification for all the expensive stuff I bought!

 

[Blumlein 88]

My first CD player was one of the 2nd gen Magnavox 1040 units which were also sold as Philips 104 units. I basically said when CD players drop to $200 I was buying one. This unit is from 1984 or so. There was a Memorial Day sale for $199. It had been something like $450 ($599 MSRP) and of course it basically became $199 after that permanently.

Specs for it all quoted as 20hz-20,000 hz. Measurements from mags showed it met or slightly exceeded these specs.
FR was +/- .3 db Measurements by magazines showed a tiny droop at 20 khz it was basically perfectly flat from a few hertz to 15,000 hz.
Crosstalk was 86 db.
SNR was 90 db.
THD was -86 db and IMD was also -86 db.
Output max was 2 volts RMS.

This player btw used 14 bit DACs running at 4x oversampling. And low level linearity wasn't all that great.

There were some early players or those just after with better specs than this.

Was even this lowest end Magnavox audibly less than perfect? I don't know, but it couldn't have been very far from inaudible. People began to talk about jitter and low level linearity, ringing transients so on and so forth. Yet in retrospect I don't know if any of that is a problem to human listeners.

It also seems this is the time when high end subjectivist audio really took off. It was some small barely known niche previously. I mean ham radio folks outnumbered it by probably 100 to 1. I've thought when we finally had a full fidelity source, the only way to differentiate the sound was by subjective comparisons which are mostly self deception by how your brain works and myths and fantasy that results from that situation.

To fill in what Restorer John is saying. Specs for Philips 960 which was using 16 bit DACs still with 4x oversampling around 1987 to 1989.
FR is 2-20,000 hz +/- .1 db
Crosstalk is 100 db.
SNR is 101 db.
THD and IMD was -96 db.
Output max 2 volts RMS.
Basically everything you can get from 16 bits.

 

[Mr. Widget]

It's very, easy to determine...

Thanks for the post. Very interesting. I'll still need to set up my own test but it seems apparent what the outcome will be.

My first CD player was one of the 2nd gen Magnavox 1040 units which were also sold as Philips 104 units. I basically said when CD players drop to $200 I was buying one. This unit is from 1984 or so. There was a Memorial Day sale for $199. It had been something like $450 ($599 MSRP) and of course it basically became $199 after that permanently.

That's why I picked 1985 and $200 CD player... that's when I stepped up and started buying CDs as well. And we were not alone. At that time and as CD production increased, record stores in a mater of months went from mostly vinyl to mostly CDs. It was surprising to watch such a quantum shift.

Initially CD prices were supposed to eventually be lower than vinyl records but of course the record companies never lowered their CD prices... and we know what happened to them a couple of decades later.

 

[Blumlein 88]

Thanks for the post. Very interesting. I'll still need to set up my own test but it seems apparent what the outcome will be.

That's why I picked 1985 and $200 CD player... that's when I stepped up and started buying CDs as well. And we were not alone. At that time and as CD production increased, record stores in a mater of months went from mostly vinyl to mostly CDs. It was surprising to watch such a quantum shift.

Initially CD prices were supposed to eventually be lower than vinyl records but of course the record companies never lowered their CD prices... and we know what happened to them a couple of decades later.

Yes, albums were about $7, and CDs were $15 or so. They could hardly make enough for several years so price never came down even though cost to produce them was well less than $1.

 

[Jim Shaw]

This whole thread rather sounds like, "Is a silk purse really better than a purse made of a sow's ear?"

The technology of digital to analog conversion has come a long way since 1985-1990. One's inability to hear a difference may have much to do with the listener's ear training. I have, for example, an early Philips DVD player that sounds relatively awful playing either CD or DVD audio through its analog outputs. Yet, take its S/PDIF coax output and send it to a modern DAC -- in my case a Modi3 -- and it sounds accurate and splendid. You must recognize that there have been improvements, not only in decoding but in the analog audio circuits in optical players over time. And improvements in the analog transfer function between the player and preamp. Just because old and new players both have RCA connectors doesn't mean the analog transfer is the same. The driving-point and input-point transfer functions can matter. Analog signal components in the chain matter. And, analog signal treatments have changed, generally for the better, in 30-40 years.

The analog portions of a CD player are in there -- and they matter -- a lot. Meanwhile, we wring our hands over the digital filtering and decoding techniques... but decoding has been reduced to practice in that period also.

I'm watching, with great curiosity, what Schiit is doing with its upcoming CD player. Schiit, for example, knows something of both digital conversion and analog circuits.

Just one man's view.

 

[restorer-john]

This player btw used 14 bit DACs running at 4x oversampling. And low level linearity wasn't all that great.

It's the low level linearity (primarily zero cross spikes) with ladder D/As where you can 'hear' differences. But, you need -60dBFS pure sine tracks and elevate the playback level to silly volumes to 'hear' it.

The standard setup for CD players with high level bit adjustment (MSB etc) is a -60dB sine, viewed on a scope and adjusted for the least zero-cross error in the waveform. The most expensive players had up to four high level bit adjustments to make low level linearity a non-issue.

 

[Mr. Widget]

This whole thread rather sounds like, "Is a silk purse really better than a purse made of a sow's ear?"

The technology of digital to analog conversion has come a long way since 1985-1990. One's inability to hear a difference may have much to do with the listener's ear training. I have, for example, an early Philips DVD player that sounds relatively awful playing either CD or DVD audio through its analog outputs. Yet, take its S/PDIF coax output and send it to a modern DAC -- in my case a Modi3 -- and it sounds accurate and splendid. You must recognize that there have been improvements, not only in decoding but in the analog audio circuits in optical players over time. And improvements in the analog transfer function between the player and preamp. Just because old and new players both have RCA connectors doesn't mean the analog transfer is the same. The driving-point and input-point transfer functions can matter. Analog signal components in the chain matter. And, analog signal treatments have changed, generally for the better, in 30-40 years.

The analog portions of a CD player are in there -- and they matter -- a lot. Meanwhile, we wring our hands over the digital filtering and decoding techniques... but decoding has been reduced to practice in that period also.

I'm watching, with great curiosity, what Schiit is doing with its upcoming CD player. Schiit, for example, knows something of both digital conversion and analog circuits.

Just one man's view.

Ok, another point of view... a data point in my view as I place more importance on anecdotal evidence than the binary zero that many here would rate it.

 

[Blumlein 88]

It's the low level linearity (primarily zero cross spikes) with ladder D/As where you can 'hear' differences. But, you need -60dBFS pure sine tracks and elevate the playback level to silly volumes to 'hear' it.

The standard setup for CD players with high level bit adjustment (MSB etc) is a -60dB sine, viewed on a scope and adjusted for the least error in the waveform. The most expensive players had up to four high level bit adjustments to make low level linearity a non-issue.

Yes, some CD test discs had fade out signals starting at -60 db or similar. With headphones listening at high levels you could hear a difference.

 

[kemmler3D]

Yes, some CD test discs had fade out signals starting at -60 db or similar. With headphones listening at high levels you could hear a difference.

Indeed, it's not hard to hear the limitations of 16 bit if you want to. Just take some audio in Audacity, lower the gain by 85dB or so, save it as 16-bit, load it again, and normalize the file. You'll hear some stuff.

In the process, think about whether any playback use case is even within an order of magnitude of this process... and you'll realize why worrying about a modern DAC is silly.


(e: I just tried this, and it works ... but what's amazing is how voice is still pretty intelligible and normal-sounding underneath all the horrible noise... even -85dB isn't enough to get down to 1- or 2-bit nastiness for the signal... you can still make out speech after a -95dB pass...)

I think the better way to hear quantization is to amplify an existing signal that far below peak to full scale.


I wanted to +1 @Jim Shaw 's points about the analog circuitry around DACs. That's where the action is, if there is any to be had in a DAC or CD player. The digital components get a lot of attention and blame for any perceived fault, but the process of turning numbers into voltages is really not where we should be spending our energy.

 

[restorer-john]

You must recognize that there have been improvements, not only in decoding but in the analog audio circuits in optical players over time. And improvements in the analog transfer function between the player and preamp. Just because old and new players both have RCA connectors doesn't mean the analog transfer is the same.

Happy to discuss this further.

The vintage TOTL players pretty much all used NE-553x opamps in their 'analog' stages, for LPF, IV and buffer stage duties. Of those, the lowest noise versions were selected (AN) Some used Analog Devices ICs for LPF duties and later, some top machines even built bespoke discrete FET input mini 'buffer' stages.

Their output impedances across the audible bandwidth were extremely low. We are talking from 50 to a few hundred ohms, so modifications to the transfer function (FR) are a complete non-issue. My reference machine here has a Zout of 200R on the RCAs and a genuine 600R on the XLRs due to the implementation of proper 600:600R balanced transformers.

 

[Jim Shaw]

Happy to discuss this further.

The vintage TOTL players pretty much all used NE-553x opamps in their 'analog' stages, for LPF, IV and buffer stage duties. Of those, the lowest noise versions were selected (AN) Some used Analog Devices ICs for LPF duties and later, some top machines even built bespoke discrete FET input mini 'buffer' stages.

Their output impedances across the audible bandwidth were extremely low. We are talking from 50 to a few hundred ohms, so modifications to the transfer function (FR) are a complete non-issue. My reference machine here has a Zout of 200R on the RCAs and a genuine 600R on the XLRs due to the implementation of proper 600:600R balanced transformers.

This isn't the space, and I'm not the guy to tutor on transfer functions, transient analysis, steady-state testing, or the limitations of Fourier transform analysis. There are engineering college textbooks that specialize in that. The one I studied, "Information transmission, signals, modulation, and noise," is out of print. But science has learned a lot more since then, anyhow. Because it is mathematically intensive doesn't mean it's superhuman to understand. Just be aware that it's a fourth-year engineering course, normally. But, it will start to unfold the effects of transfer functions on signal paths -- and mostly the unlikelihood of perfect transfer.

IF that's jibberish to tinkerers, so be it. But just know that engineers should know a lot more today, and designers have better building blocks than they had in 1985. And they use them...

 

[Mr. Widget]

This isn't the space, and I'm not the guy to tutor on transfer functions, transient analysis, steady-state testing, or the limitations of Fourier transform analysis. There are engineering college textbooks that specialize in that. The one I studied, "Information transmission, signals, modulation, and noise," is out of print. But science has learned a lot more since then, anyhow. Because it is mathematically intensive doesn't mean it's superhuman to understand. Just be aware that it's a fourth-year engineering course, normally. But, it will start to unfold the effects of transfer functions on signal paths -- and mostly the unlikelihood of perfect transfer.

IF that's jibberish to tinkerers, so be it. But just know that engineers should know a lot more today, and designers have better building blocks than they had in 1985. And they use them...

Thanks for the input.

I don't think most of us debate the notion that competent engineering has improved in the past 40 or so years as is shown in SINAD numbers, but for me at least, the question isn't is a modern DAC superior, but is it audibly superior. My untested notion had been that indeed we had made audible improvements, but now this is up for debate and I at least want to test this idea.

 

[Blumlein 88]

Here are some basic measurements of an early 14 bit Philips DAC by mansr. Pretty much meets spec.

Philips CD150 measurements

Produced 1985-1987, the Philips CD150 is basically a cost-reduced version of the original CD100. While the CD mechanism and chassis are cheaper, the DAC section remains the same. The digital audio data from the CD is oversampled 4x by the SAA7030 chip before a pair of TDA1540P chips (one for...

www.audiosciencereview.com

 

[Mr. Widget]

Here are some basic measurements of an early 14 bit Philips DAC by mansr. Pretty much meets spec.

Philips CD150 measurements

Produced 1985-1987, the Philips CD150 is basically a cost-reduced version of the original CD100. While the CD mechanism and chassis are cheaper, the DAC section remains the same. The digital audio data from the CD is oversampled 4x by the SAA7030 chip before a pair of TDA1540P chips (one for...

www.audiosciencereview.com

Sure, but my recollection of my Magnavox branded Phillips player was not that there was any noticeable noise, distortion, or frequency emphasis/deficiency... it was simply that it sounded different from what I expected.

As has been suggested here and elsewhere at ASR, either the decades of analog playback trained me (and probably others) to expect a certain sonic character which was missing, or there is an as yet unmeasured characteristic that would explain this unexpected and undesired difference.