Marantz CD67 mk2 measurements

[NTTY]

The Asus Xonar ST used the part and posted loopback measurements like this. Also, as good as the LM4562 is, it does have its quirks, including an RF allergy that @pma pointed out.

I still have the little Asus Xonar STU somewhere. Designed to provide audiophiles with what they wanted:

I played the random AOP swap with the AOP in the LPF slot. Here below are the best and worst cases:

More distorsion, that’s all. I did not keep all results, but here were the main changes (bold = stock AOP), ordered by THD:

1. LM49720 : THD = 0.00046%, THD+N = 0.00090%, SNR = 102.2dB, ENOB = 17.0bits
2. RC4560 : THD = 0.00049%, THD+N = 0.00092%, SNR = 102.2dB, ENOB = 17.0bits
3. OPA1612 : THD = 0.00050%, THD+N = 0.00092%, SNR = 102.2dB, ENOB = 17.0bits
4. OPA1622 : THD = 0.00050%, THD+N = 0.00092%, SNR = 102.2dB, ENOB = 17.0bits
5. MUSES8820 : THD = 0.00050%, THD+N = 0.00092%, SNR = 102.2dB, ENOB = 17.0bits
6. LM4562 : THD = 0.00051%, THD+N = 0.00093%, SNR = 102.2dB, ENOB = 17.0bits
7. OPA828 : THD = 0.00051%, THD+N = 0.00093%, SNR = 102.1dB, ENOB = 17.0bits
8. ORACLE02 : THD = 0.00053%, THD+N = 0.00094%, SNR = 102.2dB, ENOB = 17.0bits
9. MUSES02 : THD = 0.00084%, THD+N = 0.0012%, SNR = 101.8dB, ENOB = 17.0bits
10. TL072 : THD = 0.00097%, THD+N = 0.0013%, SNR = 101.4dB, ENOB = 16.5bits
11. TL082 : THD = 0.001%, THD+N = 0.0013%, SNR = 101.2dB, ENOB = 16.4bits

Cheers.

 

[AnalogSteph]

LOL - RC4560, the little opamp that could. That's not even a particularly low-noise type or anything. Did it fall over in SMPTE/CCIF IMD at least? Maybe there's a reason why it costs slightly more than the RC4580 after all (which wasn't all that great in Samuel Groner's testing), even though absolutely speaking it's still dirt cheap.

By contrast, JRC's fancified NJM4580, the MUSES02, did not do all that well. Beating a TL072 isn't saying much. And boy is the thing expensive.

 

[NTTY]

LOL - RC4560, the little opamp that could. That's not even a particularly low-noise type or anything. Did it fall over in SMPTE/CCIF IMD at least? Maybe there's a reason why it costs slightly more than the RC4580 after all (which wasn't all that great in Samuel Groner's testing), even though absolutely speaking it's still dirt cheap.
.

I did not record the IMD tests, but I could run them again (one day).
Both RC4560 and RC4580 were unhappy in this test

EDIT: See following messages below, I did an error with the LM4562, and I wrongly reported a variance in the above table. The true table is therefore the one below:

 

[johny_2000]

Tell me why I like this 90's board design...
Something that brings back memories of that time...

 

[horias2000]

I still have the little Asus Xonar STU somewhere. Designed to provide audiophiles with what they wanted:

View attachment 399387

I played the random AOP swap with the AOP in the LPF slot. Here below are the best and worst cases:

View attachment 399388

More distorsion, that’s all. I did not keep all results, but here were the main changes (bold = stock AOP), ordered by THD:

1. LM49720 : THD = 0.00046%, THD+N = 0.00090%, SNR = 102.2dB, ENOB = 17.0bits
2. RC4560 : THD = 0.00049%, THD+N = 0.00092%, SNR = 102.2dB, ENOB = 17.0bits
3. OPA1612 : THD = 0.00050%, THD+N = 0.00092%, SNR = 102.2dB, ENOB = 17.0bits
4. OPA1622 : THD = 0.00050%, THD+N = 0.00092%, SNR = 102.2dB, ENOB = 17.0bits
5. MUSES8820 : THD = 0.00050%, THD+N = 0.00092%, SNR = 102.2dB, ENOB = 17.0bits
6. LM4562 : THD = 0.00051%, THD+N = 0.00093%, SNR = 102.2dB, ENOB = 17.0bits
7. OPA828 : THD = 0.00051%, THD+N = 0.00093%, SNR = 102.1dB, ENOB = 17.0bits
8. ORACLE02 : THD = 0.00053%, THD+N = 0.00094%, SNR = 102.2dB, ENOB = 17.0bits
9. MUSES02 : THD = 0.00084%, THD+N = 0.0012%, SNR = 101.8dB, ENOB = 17.0bits
10. TL072 : THD = 0.00097%, THD+N = 0.0013%, SNR = 101.4dB, ENOB = 16.5bits
11. TL082 : THD = 0.001%, THD+N = 0.0013%, SNR = 101.2dB, ENOB = 16.4bits

Cheers.

It's very interesting that here is a difference between the LM49720 and the LM4562 as these are identical parts. The datasheets are identical and there was somebody from TI on a forum (can't remember which one) that said that the two parts are the same, just the name changed when TI bought National as the LM4562 was a National part. I use the LM4562 and the LME49720 in many of my projects and I'm really happy with it. I now have a few pieces of the OPA1612 and I want to see how I can take advantage of the better noise performance in my projects.

 

[johny_2000]

I now have a few pieces of the OPA1612 and I want to see how I can take advantag

I have the LME49710/49720 as well as other Burr Brown OPAs, and recently acquired some OPA1612As to test the differences in my designs.

 

[horias2000]

Tell me why I like this 90's board design...
Something that brings back memories of that time...

View attachment 399447

It does look good I have to admit. I think it's because the layout is not crowded. There are plenty of large THT components and space was not a problem when designing the PCB. It's a dream for a PCB layout engineer to have this kind of projects, where board space is not an issue and you can place the parts according to their purposes.

 

[Scytales]

It's very interesting that here is a difference between the LM49720 and the LM4562 as these are identical parts.

To be fair, the metrics you referred to are very close to each other. The difference between 0.0009 and 0.00093% THD+N is less than 0.3 dB. I would expect greater item to item variation inside the same production batch of a particular device.

What is more interesting to me is that the LM4562 under-performed severely compared to the LME49720 when used after the DAC I/V stage (what is called Preamp in NTTY's table in message #23 above), which (if there were nothing unexpected going on) strongly points out to some undocumented difference(s) between the two chips despite their common datasheet on every possible parameters.

Perhaps there are subtle difference(s) in the die implementation of this Op-amp schematics or a different manufacturing process or routing of the internal bonding wires between the pins of the chip case and the die. In any case, we can strongly suspect that the LM4562 has some susceptibility to HF interference as AnalogSteph rightly stressed, citing PMA's measurements: if the post-DAC I/V stage has sufficient bandwidth, a great deal of HF coming out of the DAC chip would be present at the output of this stage.

By the way, I would be very curious to see the schematics of the Asus Xonar STU in order to try to understand why some Op-amps under-performed compared to others in NTTY's experiments.

Incidentally, should OPA2640 be read OPA2604 in the table in message #23?

 

[NTTY]

It's very interesting that here is a difference between the LM49720 and the LM4562 as these are identical parts. The datasheets are identical and there was somebody from TI on a forum (can't remember which one) that said that the two parts are the same, just the name changed when TI bought National as the LM4562 was a National part. I use the LM4562 and the LME49720 in many of my projects and I'm really happy with it. I now have a few pieces of the OPA1612 and I want to see how I can take advantage of the better noise performance in my projects.

I don't know the history of these, but that's less than a dB difference of THD. And even if I average 32 times, it might be the type of difference we get. Waiting enough for the right moment to print screen, I could probably get the same measurement for the two. Also, if they come from different batches/different factories, it would be enough to explain.

Cheers.

 

[NTTY]

By the way, I would be very curious to see the schematics of the Asus Xonar STU in order to try to understand why some Op-amp under-performed compared to others in NTTY's experiments.

Note that #23 message here was not about the Asus but another low value Aiyima DAC-A5 pro.

Incidentally, should OPA2640 be read OPA2604 in the table in message #23?

Indeed

 

[AnalogSteph]

I did not record the IMD tests, but I could run them again (one day).
Both RC4560 and RC4580 were unhappy in this test

You must have had some socket contact issues on one of the supply pins there, and opamp legs with varying levels of oxidation. (Maybe a bad solder joint on the socket, even.) There is no reason why a genuine LM4562 should be a complete bust when the basically identical LME49720 turns in some of the best results. I would suggest arming yourself with contact cleaner + IPA and retesting any part with results worse than 0.01% at the very least.

 

[NTTY]

You must have had some socket contact issues on one of the supply pins there, and opamp legs with varying levels of oxidation. (Maybe a bad solder joint on the socket, even.) There is no reason why a genuine LM4562 should be a complete bust when the basically identical LME49720 turns in some of the best results. I would suggest arming yourself with contact cleaner + IPA and retesting any part with results worse than 0.01% at the very least.

It could be socket contact issue indeed, I did not think about that. I bought the device new. I can test again, I just need to find where I've put it

EDIT: I don't even need to do that, you're right. I probably wrongly inserted it because looking back at my file with all recorded measurements, it had the issue only once:

OpAmp in PRE AMP slot	OpAmp in LPF slot	Crosstalk @1kHz	THD 1kHz @0dBFS	THD+N @1kHz
LME49720	LM4562	-137dB	0.0028%	0.0041%
LM4562	LM4562	-137dB	0.0028%	0.0041%
NE5532*	LM4562	-114dB	0.47%	0.47%

Lesson learned: I need to be more careful at what I do and analysing the results!

And I updated the table I published before with the correct values.

 

[Scytales]

I should have thought harder of the obvious. That damned electrical contacts. The real plague of electronics.

 

[AnalogSteph]

It could be socket contact issue indeed, I did not think about that. I bought the device new. I can test again, I just need to find where I've put it

EDIT: I don't even need to do that, you're right. I probably wrongly inserted it because looking back at my file with all recorded measurements, it had the issue only once:

OpAmp in PRE AMP slot	OpAmp in LPF slot	Crosstalk @1kHz	THD 1kHz @0dBFS	THD+N @1kHz
LME49720	LM4562	-137dB	0.0028%	0.0041%
LM4562	LM4562	-137dB	0.0028%	0.0041%
NE5532*	LM4562	-114dB	0.47%	0.47%

Lesson learned: I need to be more careful at what I do and analysing the results!

And I updated the table I published before with the correct values.

So far so good. What about the poor little RC4560 though? I'd want to recheck every single part that has both distortion and chsep in orange or red. The M5218 (somewhere between a 4558 and 4560 but closer to the latter) and even the TL072 did just fine after all. It might mean that the table turns out looking quite boring, but that's a good thing in my book. You're only looking for -91 dB THD and -88 dB THD+N, that's not exactly cutting edge performance.

Any vintage home computer enthusiast will be able to tell yout that DIP sockets exist in varying levels of quality. You ideally want the dual wipe variety, not just cheap and crappy single wipe ones.

BTW, as I wrote here,

Note that TI now sells two LM833 variations, one being the original NatSemi part (called "LM833-N", LM833N in DIP), the other their take on the MC33078 as discussed below (LM833P in DIP).

These are two entirely different parts and should be distinguished as such.

 

[NTTY]

So far so good. What about the poor little RC4560 though? I'd want to recheck every single part that has both distortion and chsep in orange or red. The M5218 (somewhere between a 4558 and 4560 but closer to the latter) and even the TL072 did just fine after all. It might mean that the table turns out looking quite boring, but that's a good thing in my book.

I could (re)try indeed, but that will be far in my list of priorities

You're only looking for -91 dB THD and -88 dB THD+N, that's not exactly cutting edge performance.

Yep, that was the idea at the time. I wanted to start from what some people told me, that they can hear an obvious difference between OPA, but they didn't tell me that I had to use super high performing devices for the testing At the end of the day, in blind tests, I could not hear a difference. A friend of mine reliably spotted an issue with the RC4560, though, but not me.

Any vintage home computer enthusiast will be able to tell yout that DIP sockets exist in varying levels of quality. You ideally want the dual wipe variety, not just cheap and crappy single wipe ones.

Indeed.

BTW, as I wrote here,

Nice work!

 

[AnalogSteph]

At the end of the day, in blind tests, I could not hear a difference. A friend of mine reliably spotted an issue with the RC4560, though, but not me.

It was probably working fine for the most part but clipping on peaks since it was running short on voltage on one side. I bet if you had just played a near-0dBFS sine you would 100% have heard that.

This is an important lesson in a way: Factors you had not thought of can ruin your tests, and you may hear or measure a difference but it might not be because of the aspect you were focusing on (the opamp type). This is why repeatability is considered so critical in science. You should probably update the respective thread with a note, lest someone be temped to draw invalid conclusions from corrupted data.

I had not previously considered that poor contact in sockets could be such a major issue either. Imagine people are merrily opamp rolling and actually hearing what they think is a difference between types, while actually it's just the bloody sockets that are setting them up. It's obviously known that they're not making things any better, what with parasitic inductance and all, but having to write off 1/3 of your results because of poor contact is substantial.

 

[NTTY]

It was probably working fine for the most part but clipping on peaks since it was running short on voltage on one side. I bet if you had just played a near-0dBFS sine you would 100% have heard that.

Probably, but the idea was to listen through music.

This is an important lesson in a way: Factors you had not thought of can ruin your tests, and you may hear or measure a difference but it might not be because of the aspect you were focusing on (the opamp type). This is why repeatability is considered so critical in science. You should probably update the respective thread with a note, lest someone be temped to draw invalid conclusions from corrupted data.

That’s a good lesson indeed. If I remember correctly, I tested 89 combinations and did much more swapping. Now it could indeed be a problem with the socket on some of them. I can repeat the test for those that did not shine, it would not be so long.

I had not previously considered that poor contact in sockets could be such a major issue either. Imagine people are merrily opamp rolling and actually hearing what they think is a difference between types, while actually it's just the bloody sockets that are setting them up. It's obviously known that they're not making things any better, what with parasitic inductance and all, but having to write off 1/3 of your results because of poor contact is substantial.

Good lesson learned (again) and one not much documented indeed. I’ll go back to my excel since it also includes the sequence, maybe the socket got worn. It would show and I can test again to confirm.
I added a note to the initial post as a warning

 

[Mart68]

I have a CD67SE in my collection. Never liked it. Was hoping there would be something in these measurements to correlate to that, but seemingly not.

Construction is very plasticky, which I hate, so probably just cognitive bias born of that.

 

[ReDFoX]

I own an original CD67 and as long as I remember, I had a feeling that highs were rolled off a little bit, cause it sounded different from my Pioneer DV565...
Do you have any tips for burning a test CD? I can't quite get REW to do offline analysis properly

 

[restorer-john]

I have a CD67SE in my collection. Never liked it. Was hoping there would be something in these measurements to correlate to that, but seemingly not.

Construction is very plasticky, which I hate, so probably just cognitive bias born of that.

The "SE" in that particular case was a cynical exploitation of early audiophile nonsense, subsequently perpetuated by Marantz and the sadly departed Ken Ishiwata. Screwing on a steel base plate and gluing on a inner steel/bitumen pad to make the unit heavier and sound more expensive when you rapped your knuckles on the casework. They threw some Elna Silmics at it, and the HDAM- I think it got a deluxe copper plated box didn't it?

Still didn't change the truly horrible loading mechanism and cheap transport Philips were famous for.

I had one, sold it and never regretted it.