Replacing NJM2068DD in Yamaha Amplifier

[MilanAudio]

Are you sure you want to bother with the old TL072? It has rather poor common-mode linearity, and in term of feedback resistor values you're looking at 10k/4k7 at the very least if you do not want to annoy its output stage overly much, perhaps even 22k/10k. (At this point it can't hurt to have maybe 10 pF in parallel with the feedback resistor, to eliminate input capacitance related gain peaking.) On the pro side, you can quite easily stack two to offset their pathetic driving abilities, or more - mind you, their input capacitance adds up as well.

AD826 is usually depicted with 1k in feedback but I imagine the existing values should work decently as well. Its linearity in general isn't that spectacular but better than the TL072 in common mode at least. (It's a video opamp, so rather designed to drive ~1 Vrms into 75 ohms than 6 Vrms into a few kOhms.) I wouldn't try keeping my finger on this guy though, it'll be at around 400 mW in idle.

With LM4562/LME49720, I'd just leave resistor values as-is, they should work fine.

Really, the input amp as-is isn't bad. I may be inclined to go up to 6k8/2k2 for reduced output stage loading at worst-case levels, but that's about it. Yamaha has done much worse, like torturing a poor little OP275 with 390R/220R at the same spot in the A-S700 (hint: it's an OK-ish load driver but well short of an NE5532). If you don't bother its output stage too much, the NJM2068 isn't a bad performer at all (NwAvGuy measurements - Samuel Groner unfortunately skipped this part).

The question though is, do you even need or want 9.2 dB of gain there in this day and age? CD levels have gone up quite a bit since the early days, and 150 mV of input sensitivity on a CD input quite arguably was generous even when this amp was new. (Where do you typically have the volume pot set, 9 o'clock or even lower?) So if you are running a standalone CD player into this and predominantly listen to releases from the '90s and newer, you might as well take out the 2k2s, replace them with the 1ks (the part even is commonly used with direct feedback) and call it a day. The NJM2068 makes a perfectly fine unity gain buffer at 2 Vrms.

Now if you are running substantially less than CD player level into it, e.g. because you have a modern 24-bit DAC with more than enough dynamic range and you like to have your collection ReplayGained with settings adequate to cover just about everything that has ever come out on CD (like me with an extra -3.2 dB, which means a total of -9 to -13 dB on modern pop releases more often than not)... then you pretty much have no reason to even be bothering with this modding business at all. As NwAvGuy showed, the NJM2068 will happily work in a gain of 3 circuit with 1k5/750R at 2 Vrms all day long with negligible distortion, and that's a somewhat heavier load than what you have there.

But people just like tinkering, I guess.

In this day and age, I would consider "CD input" a bit of a misnomer... it makes a splendid choice for a computer or mobile player though.

I think I can give a try to many op amps and (perhaps) I will hear the difference. Just checked my of items and some DIP8 sockets are around me. Also I found in my box NJM4580DD so testing can start.

The reason (as I mentioned) why I want change NJM2068 is that I am missing heights in the sound. For a few days I had Yamaha A-S500. Comparing it to Yamaha AX-10 which is 20 yo I am quite impressed with with AX-10 and its low and mid frequencies in the sound. I thought OK, I will buy A-S500... But why? AX-10 is small, cute, really powerful amp (tested on floor speakers), just the heights are not that present over "PURE DIRECT".

And if you are talking about CD input level, well, I would be happy if my amp could be quieter. And yes, I am listening to CDs from CD player and FLAC from computer. So if changing R251/R256 should be changde from 220 ohm to 300Ohm, I will be happy to do it And I will see/hear.

 

[indianajo]

If you're missing highs from the amp, perhaps the VAS or driver transistors in the amp portion are dull & lifeless. My AX6 build had dull highs as I built it, with 3 mhz Ft drivers. I had a survivor 1966 design board in the stereo chassis to compare it to, which sounded really good as modified by djoffe. I changed AX6 to 30 mhz Ft drivers last month and the highs sound like they are supposed to. The new drivers were MJE15028/29. BD139/140 are supposed to make good VAS transistors. Post the amp portion of your schematic and let the team look at it. MJE15028/29 used to cost $4.05 each, now they are $1.20. So processes have improved. The TIP31C/32C I replaced were $.35. A bargain, but not with blah sound.
I have to say specs on NJM2068 look really good. I put similar ST33078 instead of 4558, and that improved my RIAA cartridge preamp a lot. They oscillated though, I had to put local +-v bypass caps and bypass the feedback resistor with a 33 pf ceramic disk cap. I've got 4580 in a Peavey CS800s amp and it sounds really good.

 

[MilanAudio]

If you're missing highs from the amp, perhaps the VAS or driver transistors in the amp portion are dull & lifeless. My AX6 build had dull highs as I built it, with 3 mhz Ft drivers. I had a survivor 1966 design board in the stereo chassis to compare it to, which sounded really good as modified by djoffe. I changed AX6 to 30 mhz Ft drivers last month and the highs sound like they are supposed to. The new drivers were MJE15028/29. BD139/140 are supposed to make good VAS transistors. Post the amp portion of your schematic and let the team look at it. MJE15028/29 used to cost $4.05 each, now they are $1.20. So processes have improved. The TIP31C/32C I replaced were $.35. A bargain, but not with blah sound.
I have to say specs on NJM2068 look really good. I put similar ST33078 instead of 4558, and that improved my RIAA cartridge preamp a lot. They oscillated though, I had to put local +-v bypass caps and bypass the feedback resistor with a 33 pf ceramic disk cap. I've got 4580 in a Peavey CS800s amp and it sounds really good.

The service manual with schematics is here: ax-10.pdf

When I did re-cap I was looking also for replacing some other components, but many of them are out of stock (another frustrating thing for me).

The amp has lots of carbon resistors, so it was another big 'chapter' for reading and getting 'educated'. Somewhere is written that carbon resistors are getting old are more noisy than metal oxide and metal film resistors.

Well, if accident happen and I will burn my AX-10, then I have another two AX-10 (I would rather to sell them and keep just one.)

 

[sgrossklass]

The amp has lots of carbon resistors, so it was another big 'chapter' for reading and getting 'educated'. Somewhere is written that carbon resistors are getting old are more noisy than metal oxide and metal film resistors.

You're right, the thing is littered with carbon film everywhere, even the phono stage. Blame the bean counters. The only concession to linearity seems to have been parts with higher wattage rating (P=10 instead of P=5 for unmarked ones - less self-heating, hence better linearity).

I'd imagine the most critical ones would have to be the ones in the power amp feedback loop (beware: wattage rating for the 33k probably is not too luxurious and multiple parts may be needed if all you have is 1/4W MF) and CD input amp feedback.

If you want to know whether the CD input amp is contributing to "dull" sound: simply use one of the other high-level inputs.

If I had to guess as to the probable causes of "dull" sound -
1. Condition (contact resistance) of speaker protection relay
2. Your mind fooling you with the absence of any audible noise
3. Apples-and-oranges comparisons without proper level matching
4. Power amp feedback resistors

 

[Mark Tillotson]

That circuit appears to lack any HF decoupling for the opamps too, having ludicrously large 1000uF electrolytics on the 15V rails for some reason - and they are 16V rated which is pushing it somewhat. 25V caps might be wiser.



Many modern opamps need HF decoupling, 100nF 50V ceramic between V+ and V- right next to the package.

 

[MilanAudio]

... having ludicrously large 1000uF electrolytics on the 15V rails for some reason - and they are 16V rated which is pushing it somewhat. 25V caps might be wiser. ...

I think that two caps are there just for filtering/buffering/smoothing that 15V for feeding op amps.

In my basked are already 25V versions.

 

[12Bass]

In my experience, LM4562/LME49720 is a little more open sounding than the NJM2068. I also measured lower THD and IMD with RMAA in analog loopback when I replaced the original NJM2068s used for the 1/2 inputs and outputs on my Echo Gina24 sound card with LM4562s. The card still has six output channels with NJM2068 which can be used for direct comparison. However, I wouldn't expect that a new op amp would clear up an otherwise dull sound, as that seems like it may be caused by something else.

 

[MilanAudio]

In my experience, LM4562/LME49720 is a little more open sounding than the NJM2068. I also measured lower THD and IMD with RMAA in analog loopback when I replaced the original NJM2068s used for the 1/2 inputs and outputs on my Echo Gina24 sound card with LM4562s. The card still has six output channels with NJM2068 which can be used for direct comparison. However, I wouldn't expect that a new op amp would clear up an otherwise dull sound, as that seems like it may be caused by something else.

Yes, I believe that things can be better. Right now I do order with Mouser and I think having an extra soldering fun with adding new relays, few resistors and replacing polypropylene capacitor C252/C253 with polystyrene one could be another little improvement.

I hope I will have enough time to check it also with oscilloscope (and find out how to use it )