Did you mean the ADA4627-1?
The '4627-1 is one of the best sounding JFET op-amps out there, but as you noted doesn't always sound the best in all circuits/applications.
I personally think the Intersil ISL28110/28210 and the OPA1641/1642 sound very similar in VAS stages.
That was a typo. Yes, it's ADA4627-1 which is said to be AD's OPA627. Speaking of op amps on audio buffer, it seems none of the analog characteristic is determinant. I've tried fast (AD8065), low noise (ADA4898-1), low distortion/ high PSRR/CMRR (LME49990) and many other bi-polar/FET op amps. I still got no clue on which characteristic is the key for audio.
If you did have the answer, you'd have a very highly paid job with one of the opamp manufacturers!
There are too many variables to be able to predict audio quality except by trying an opamp in its final circuit.
Where the input comes from, where the output goes to, (source & load impedances) , circuit layout, power supply (& yes, pssrr), internal architecture etc.
Sadly my hearing is no longer acute enough to detect differences between opamps so I'll have to settle for one of the recommended ones. (844 + 627 probably)
I agree. There is no universal solution. The sound depends on too many variables. Last night, I tweaked the DC servo input (add LPF), and the sound changed to certain degree. My circuit is till on a breadboard (no ground plane, long ground trace). Maybe it's too early to make the judgement on op amps.
AD844 with unbuffered Tz pin
Someone commented that without feedback they thought the 844 was no longer an opamp. I think they are right, it can be treated as two distinct blocks:
1) The dac current flows between the inputs, so through the 50R input resistance of the inverting input to ground (since the non inverting input is grounded)
2) The input current is mirrored out of the Tz pin, the current being derived from the +ve supply rail. The voltage gain is the ratio of the input resistance (50r) to the Tz resistor so a 1K resistor will give 4v p-p from a Tda1541 - ie 1.4v rms, albeit, centred about +2v.
Perhaps the Ad844 sounds so good because there is no feedback, its simply a very good current mirror.
Assuming your preamp has a typical input impedance of 47k, there is surely no need for any buffer opamp, just connect it direct to the Tz pin though a capacitor.
I appreciate this will spoil the fun for many who like to compare the sound of opamps! But as one who regards feedback as a bodge and best avoided, adding an opamp when we already have a large enough output from the Tz pin to drive most preamps, seems a step back.
Someone commented that without feedback they thought the 844 was no longer an opamp. I think they are right, it can be treated as two distinct blocks:
1) The dac current flows between the inputs, so through the 50R input resistance of the inverting input to ground (since the non inverting input is grounded)
2) The input current is mirrored out of the Tz pin, the current being derived from the +ve supply rail. The voltage gain is the ratio of the input resistance (50r) to the Tz resistor so a 1K resistor will give 4v p-p from a Tda1541 - ie 1.4v rms, albeit, centred about +2v.
Perhaps the Ad844 sounds so good because there is no feedback, its simply a very good current mirror.
Assuming your preamp has a typical input impedance of 47k, there is surely no need for any buffer opamp, just connect it direct to the Tz pin though a capacitor.
I appreciate this will spoil the fun for many who like to compare the sound of opamps! But as one who regards feedback as a bodge and best avoided, adding an opamp when we already have a large enough output from the Tz pin to drive most preamps, seems a step back.
Yes, a low feedback discrete buffer that can drive my transformer is something I may investigate. Or perhaps a tube based buffer?
SOIC AD844-JR
Do these look OK? Its unlikely anyone would change printing on SOIC's only worth a buck each.
The SOIC legs are a bit fiddly and short so stacking will need help e.g.using a ribbon cable with matching pitch - 40c worth
The SOIC's are only half the height of the DIP-8 and are so affordable they are just asking for a 10 stack per side to allow direct TZ pin output free of an extra buffer stage as per batteryman's inspiration.
Do these look OK? Its unlikely anyone would change printing on SOIC's only worth a buck each.
The SOIC legs are a bit fiddly and short so stacking will need help e.g.using a ribbon cable with matching pitch - 40c worth
The SOIC's are only half the height of the DIP-8 and are so affordable they are just asking for a 10 stack per side to allow direct TZ pin output free of an extra buffer stage as per batteryman's inspiration.
Sparkos Discrete op amp
Discrete Op Amps - Sparkos Labs.com Here is a link to the web page. I am experimenting with the SS3601. It was necessary to convert my JFET based buffer. It was originally designed for the OPA627 and works well with the OPA1641. Original buffer had over 1 Megohm input impedance. A gain set at 2. I made the following changes. Input impedance set to 100K ohm. Gain at 2 actually saturated the Lundahl LL1690's in the deep bass. So this was changed to Gain =1. No saturation now. The difference isn't one bit subtle. I am taking the PCM1704 dac to my cousins on Monday. After getting a second opinion I will post my listening impressions.
Discrete Op Amps - Sparkos Labs.com Here is a link to the web page. I am experimenting with the SS3601. It was necessary to convert my JFET based buffer. It was originally designed for the OPA627 and works well with the OPA1641. Original buffer had over 1 Megohm input impedance. A gain set at 2. I made the following changes. Input impedance set to 100K ohm. Gain at 2 actually saturated the Lundahl LL1690's in the deep bass. So this was changed to Gain =1. No saturation now. The difference isn't one bit subtle. I am taking the PCM1704 dac to my cousins on Monday. After getting a second opinion I will post my listening impressions.
Original buffer had over 1 Megohm input impedance. A gain set at 2. I made the following changes. Input impedance set to 100K ohm. Gain at 2 actually saturated the Lundahl LL1690's in the deep bass. So this was changed to Gain =1. No saturation now. The difference isn't one bit subtle. I am taking the PCM1704 dac to my cousins on Monday. After getting a second opinion I will post my listening impressions.
Bit confused, are the Lundahl LL1690's that were saturated by the buffer on the output of the dac? If so why not just take the output from the buffer? Or is it that you wanted a balanced output with the Lundahl LL1690's?
Cheers George
Do these look OK? Its unlikely anyone would change printing on SOIC's only worth a buck each.
View attachment 600359
The SOIC legs are a bit fiddly and short so stacking will need help e.g.using a ribbon cable with matching pitch - 40c worth
The SOIC's are only half the height of the DIP-8 and are so affordable they are just asking for a 10 stack per side to allow direct TZ pin output free of an extra buffer stage as per batteryman's inspiration.
Rather you than me soldering ribbon cable to 1.27mm pitch pins!
Alternatively you could solder each one to a soic to dil adaptor, then stack these. Much easier as some 1mm dia Cu wire can be run through all 10 holes in a vertical stack, and only for the 5 holes that needed linking.
Also, the current requirement of the stack is over 100ma so its better if the power pins were fed from a central point and not via a daisy chain of 1/2 pitch ribbon cable.
Thats smart. Thanks for your help.
An externally hosted image should be here but it was not working when we last tested it.
Output filter
DAC to DDNF to Buffer to LL1690 reconstruction filter. Gets rid of stuff above the audio band. To much high frequency rubbish without it. So the Sparkos have so much deep bass they saturated the LL1690 when used with gain =2. No issue with unity gain buffer configuration.
DAC to DDNF to Buffer to LL1690 reconstruction filter. Gets rid of stuff above the audio band. To much high frequency rubbish without it. So the Sparkos have so much deep bass they saturated the LL1690 when used with gain =2. No issue with unity gain buffer configuration.
DAC to DDNF to Buffer to LL1690 reconstruction filter. Gets rid of stuff above the audio band. To much high frequency rubbish without it. So the Sparkos have so much deep bass they saturated the LL1690 when used with gain =2. No issue with unity gain buffer configuration.
But wouldn't the Lundahl's be just low order LP filter, that could be done, somewhere up the line on the buffer seeing it's got a feedback loop, then no need for the Lundahl to be incorporated.
Cheers George
Filtering and line transformers
True. With the first pole on the DDNF, the Lundahl is technically a second order filter at about 21 Khz. An important distinction is that the line transformers are actually extended response meaning they really start to roll off faster above their passband. So I'd estimate the response is closer to 5th order. However I have not attempted to measure it. So it's kind of a guess. I may look at that filtered buffer in the future. I hate the Sallen Key filters so keep that in mind. 
True. With the first pole on the DDNF, the Lundahl is technically a second order filter at about 21 Khz. An important distinction is that the line transformers are actually extended response meaning they really start to roll off faster above their passband. So I'd estimate the response is closer to 5th order. However I have not attempted to measure it. So it's kind of a guess.
I may look at that filtered buffer in the future. I hate the Sallen Key filters so keep that in mind.
Line transformers...
I tend to prefer the transformer approach over the active filter one. Adding additional active devices in my experience does more harm then a quality transformer does. My experience might not be typical.
I tend to prefer the transformer approach over the active filter one. Adding additional active devices in my experience does more harm then a quality transformer does. My experience might not be typical.