Hello George,
i find your experiments very interesting, so thank you for sharing your long time efforts!
But reading your experiences, beyond measurements, i could not have a clear understanding of SQ impressions about different opamps (maybe it's too soon for a comprehensive opamps shootout...).
In particular, i can't understand what is the best sounding opamps among 627 family FE-onboard, if ADA4627, OPA 827 or OPA627 himself (no cost limitations, of course). Obviously removing C9, as suggested...
What are your impressions? Is OPA627 the best sounding opamp in My_Ref FE or are its "cousins" more 'magic' sounding? Which one would you suggest for a satisfaction-safe FE mod?
Last edited:
So I'm not the only one who's confused by all these experiments and the conclusions drawn? Can someone organize the various circuit modifications and subjective impressions? My head is spinning.
When C9 is removed and shorted, what happens to R10?
I like the way my amps sound now, and I'm not willing to tear into them until I see something more conclusive. Lots of people have worked on this amp in the past and raved about the results, only to realize after the thrill of modification wore off that the new thing didn't sound any better, or even as good, as the original thing.
I am not being critical of experimentation. In fact, I'm thrilled that people are still thinking about ways to improve this design, and I think it would honor Mauro and make him proud. Please continue, but try to better organize your presentation of results. Perhaps leave discussion of or comparison to MyEvo completely out of it. This is a MyRef thread, after all.
Peace,
Tom E
When C9 is removed and shorted, what happens to R10?
I like the way my amps sound now, and I'm not willing to tear into them until I see something more conclusive. Lots of people have worked on this amp in the past and raved about the results, only to realize after the thrill of modification wore off that the new thing didn't sound any better, or even as good, as the original thing.
I am not being critical of experimentation. In fact, I'm thrilled that people are still thinking about ways to improve this design, and I think it would honor Mauro and make him proud. Please continue, but try to better organize your presentation of results. Perhaps leave discussion of or comparison to MyEvo completely out of it. This is a MyRef thread, after all.
Peace,
Tom E
Tom, i'm not clear who you are addressing this to, and what is confusing you. The changes being made by myself and others are in the same fashion of others done throughout this thread, and are no more or less conclusive.
If you can be specific about what's confusing you, maybe someone can help.
If you can be specific about what's confusing you, maybe someone can help.
Tom, Luca, I hear You..
I can only bring up as an excuse, that initially I definitely did not want people, except for the adventurous minds, to easily replicate these tests, and get into problems..
But as time passing, experience is getting accumulated and I feel more confident in suggesting some variants.
So my suggestion as safe bets would be, presently: ADA4627 & OPA827, as equally well performing parts, and with a way of implementation tested by a slowly growing number of people.
The other chips, as though some of them are also 'okay', like the original OPA627: are less 'inspiring', somehow.
The ADA4898 is a different animal: it is very interesting, but I think it needs some more work, more investigation, because it's bipolar, and so the explored advantages offered by the Fet input counterparts are not so simply applicable. But I feel it's worth the efforts.
And a clear 'wiki' page style reassumption would not be a bad idea..
Ciao, George
I can only bring up as an excuse, that initially I definitely did not want people, except for the adventurous minds, to easily replicate these tests, and get into problems..
But as time passing, experience is getting accumulated and I feel more confident in suggesting some variants.
So my suggestion as safe bets would be, presently: ADA4627 & OPA827, as equally well performing parts, and with a way of implementation tested by a slowly growing number of people.
The other chips, as though some of them are also 'okay', like the original OPA627: are less 'inspiring', somehow.
The ADA4898 is a different animal: it is very interesting, but I think it needs some more work, more investigation, because it's bipolar, and so the explored advantages offered by the Fet input counterparts are not so simply applicable. But I feel it's worth the efforts.
And a clear 'wiki' page style reassumption would not be a bad idea..
Ciao, George
With C9 bypassed, R10 connects to ground. The amp now has a gain at DC of around 30 (12K / 390 + 1), so it's critical that the amp has no DC on it's input.
Yes, and I suppose this is what AndrewT had been hinting at. Bluntly..
But he is right, about being always careful.
So, with the C9 shorted, you have an amp with full gain at DC. In contrast with a design still using C9, or servo-controlled applications, where the gain at DC level is set back to 1.
It obviously means that care should be taken what regards the DC component at the input of the amp. C13 is there for this reason, to block any input DC from getting amplified, and doing harm on the output.
Just an example: how many Gaincard, Gainclone style amplifiers had been constructed in the world in your opinion? They all work in the very same way, they have full DC gain
and need an input DC blocking cap so as to prevent DC from the source propagating through. They usually are not even including DC output protection.
Then, loads of designs apply DC-servo. What is not clear for the users that DC-servo is not able to block dangerous levels of DC from propagating through.
Dc -servo circuits saturate usually at already very low levels -- after that they are not able to compensate any more, and just let through the rest of the high DC input signal. At that point, the circuit amplifies it with nominal gain, ~30dB generally.
So, circuits with DC servo are prone almost the same way to possible distruction.
The only effective measure is Dc-output protection, like the one in these Penasa/ FE amps.
Ciao, George
But he is right, about being always careful.
So, with the C9 shorted, you have an amp with full gain at DC. In contrast with a design still using C9, or servo-controlled applications, where the gain at DC level is set back to 1.
It obviously means that care should be taken what regards the DC component at the input of the amp. C13 is there for this reason, to block any input DC from getting amplified, and doing harm on the output.
Just an example: how many Gaincard, Gainclone style amplifiers had been constructed in the world in your opinion? They all work in the very same way, they have full DC gain
and need an input DC blocking cap so as to prevent DC from the source propagating through. They usually are not even including DC output protection.
Then, loads of designs apply DC-servo. What is not clear for the users that DC-servo is not able to block dangerous levels of DC from propagating through.
Dc -servo circuits saturate usually at already very low levels -- after that they are not able to compensate any more, and just let through the rest of the high DC input signal. At that point, the circuit amplifies it with nominal gain, ~30dB generally.
So, circuits with DC servo are prone almost the same way to possible distruction.
The only effective measure is Dc-output protection, like the one in these Penasa/ FE amps.
Ciao, George
Last edited:
Thanks for clarifying. Still not sure I want to start swapping opamps. C9 has always been one of the "weak links" in this design. I use the Blackgate std (actually, I was the first to try it in the old MyRef C boards), and it sounds very good. I understand its function, and that the opamp must be changed first. I also now realize that R10 sets the gain and must remain, so dumb question there. However, with C9 out of circuit, are there any restraints on changing R10 value to obtain different gain?
I have a pair of new OPA627's on hand. Is it worthwhile to try them and short C9, or does the amp sound as good with LM318 and BG std C9 in place? Are there any other mods that must be done at the same time?
Peace,
Tom E
I have a pair of new OPA627's on hand. Is it worthwhile to try them and short C9, or does the amp sound as good with LM318 and BG std C9 in place? Are there any other mods that must be done at the same time?
Peace,
Tom E
no cap sounds much better than any cap...
yes. Stability and performance. This is a very sophisticated and delicate nested feedback circuit. It's not a circuit which anyone can play with: only a competent and expert professional EE such as George (Joseph K) can do so.
Please let me make it very clear, before someone make disasters:
In this circuit it is NOT possible to simply swap OpAmps!
It is possible to use different OpAmps, BUT some other circuit details (mainly compensation) will likely have to be modified accordingly.
Dropping in whatever OpAmp you have at hand without proper circuit tuning will likely result in wild self-oscillations. Which might destroy your amplifier, your speakers or even set your house in fire. You've been warned.
If you want to improve circuit performance without making damage, you have to follow exactly all of George's indications about circuit modifications and components. And double check you did it all right.
As George clearly stated in previous posts, while the OPA627 can be made to work somehow, for various reasons it is NOT recommended to use it. That is: DON'T TRY IT at home.
To date, the only recommended OpAmps are ADA4627 or OPA827. Make sure to follow detailed instructions before swapping.
And yes, the modified amplifier sounds way much better than the original version. Yet another big advantage is that using those OpAmps you'll no longer need a buffer or preamplifier to get the most out of this nice amplifier. You can simply place a volume potentiometer in front of it.
Last edited:
Agreed. For me the opamp substitution was a much bigger step forward than removing C9.
I'm not following, why is this?
Increasing gain by reducing R10 will make the amp more stable, and vice-versa. Don't increase R10 unless you are willing/able to adjust compensation.
Ideally, OpAmps should "see" exactly the same impedance (to ground) from both inputs (both for DC & AC). Failing to do so may negatively affect performances (and SQ) as well as increase DC offset on output. Placing a volume control potentiometer on the amplifier input will prevent the possibility to guarantee such a balance (for every position of the pot). But JFET-input OpAmps (such as ADA4627 and OPA827) are much less sensitive to impedance unbalance than BJT-input ones (such as the original LM318).
In fact, long ago we had found out that just adding a potentiometer in front of an (unmodified) my_Ref or myEvo to turn it into an "integrated" amplifier was detrimental in terms of SQ WRT driving it with a suitable buffer or preamp. Lately (after the mods) we tried again, and this time results have reversed: they sounds better when used with just a potentiometer. Less is more, as it should be.
(YMMV).Ideally, OpAmps should "see" exactly the same impedance (to ground) from both inputs (both for DC & AC). Failing to do so may negatively affect performances (and SQ) as well as increase DC offset on output. Placing a volume control potentiometer on the amplifier input will prevent the possibility to guarantee such a balance (for every position of the pot). But JFET-input OpAmps (such as ADA4627 and OPA827) are much less sensitive to impedance unbalance than BJT-input ones (such as the original LM318).
In fact, long ago we had found out that just adding a potentiometer in front of an (unmodified) my_Ref or myEvo to turn it into an "integrated" amplifier was detrimental in terms of SQ WRT driving it with a suitable buffer or preamp. Lately (after the mods) we tried again, and this time results have reversed: they sounds better when used with just a potentiometer. Less is more, as it should be.
It is good to see the mods make the amps friendly to installation of a passive pot. I saw George's Full version of Evo Rev A this week. The pot is right at the RCA jacks.
I like using 10k pots. But with not need one right now. My two favorite dacs have four channel volume controls for the Evo and a sub.
Tom,
I'm afraid Paolo is generating some confusion; I have never said do not use the OPA627; I have stated several times it can be used, with the same guidelines like those for the ADA4627 given;
Your question regarding the value of this test -- hm, good question.
As I stated, I did not find anything especially wrong with this chip, it works very well in the FE design, it measures like a beast, and it sounds ok..
Is it better than the original, the LM318? Good question. For sure the OPA627 brings all the discussed advantages of the Fet inputs; you can try with C9 shortened; you can place a potentiometer direct on the input; it is less noisy than the LM318, because of the missing input current noise; you can use with higher impedances. It's not as perfect in opressing common mode input distortion as the OPA827, though.
And in some days I might have again an another amp in it's original format.
Then I can compare that to the FE with OPA627's in it, directly.
If you would decide to 'move ahead', I can help You.
Ciao, George
I'm afraid Paolo is generating some confusion; I have never said do not use the OPA627; I have stated several times it can be used, with the same guidelines like those for the ADA4627 given;
Your question regarding the value of this test -- hm, good question.
As I stated, I did not find anything especially wrong with this chip, it works very well in the FE design, it measures like a beast, and it sounds ok..
Is it better than the original, the LM318? Good question. For sure the OPA627 brings all the discussed advantages of the Fet inputs; you can try with C9 shortened; you can place a potentiometer direct on the input; it is less noisy than the LM318, because of the missing input current noise; you can use with higher impedances. It's not as perfect in opressing common mode input distortion as the OPA827, though.
And in some days I might have again an another amp in it's original format.
Then I can compare that to the FE with OPA627's in it, directly.
If you would decide to 'move ahead', I can help You.
Ciao, George
George,
Again, it was very nice to be able to chat, my pleasure!
And I hope we had just curbed up your curiosity enough, so You might wish to launch in experimenting..
Ciao, George
Hi all
I can share my experience
I've fitted Opa827 , C9 is returned in my capacitor bag and Rin now is 330K, for all this mods i have to thanks Joseph. For who use Cin, now the C can go down to 330nF.
The result is wonderful, especially with my LDR, detail, dinamic, punch and soundstage are first class, the sound is free of compression, smooth and gorgeous ...and like Unixman said, no more needs of Pre or Buffer, just a volume regulator.
This mods raise FE ( IMHO) in the olimpus of Amplifier...I tried FE versus Gryphon, McIntosh and other amps, the FE was preferred ( by owners of Gryphon, Mc etcetera, not just by me )
For who have speakers with low sensibility and low impedance, I can suggest to try 4 FE , 2 for High and 2 for low frequency ( if speakers have double connection )
With this configuration i have drove Penaudio Chara Charisma, Merlin MMx without problems and with a stunning results.
Ciao
Marco
I can share my experience
I've fitted Opa827 , C9 is returned in my capacitor bag and Rin now is 330K, for all this mods i have to thanks Joseph. For who use Cin, now the C can go down to 330nF.
The result is wonderful, especially with my LDR, detail, dinamic, punch and soundstage are first class, the sound is free of compression, smooth and gorgeous ...and like Unixman said, no more needs of Pre or Buffer, just a volume regulator.
This mods raise FE ( IMHO) in the olimpus of Amplifier...I tried FE versus Gryphon, McIntosh and other amps, the FE was preferred ( by owners of Gryphon, Mc etcetera, not just by me )
For who have speakers with low sensibility and low impedance, I can suggest to try 4 FE , 2 for High and 2 for low frequency ( if speakers have double connection )
With this configuration i have drove Penaudio Chara Charisma, Merlin MMx without problems and with a stunning results.
Ciao
Marco
But the inputs are unbalanced with or without a buffer or pot. In fact, adding a pot will bring the inputs closer in balance, regardless of whether a bipolar or FET input is used.
I'm not denying that it may sound better without a buffer, I just don't see how the FET input cares less about a buffer being present. George?