Allen Wright said:
A technical response to that statement is not possible. It also leads to the resistive loading of the VAS to "improve" the sound argument. Discussion of these issues is pointless. Simple equivalent part substitution sending people “running out of the room” fall in the category of “wives who could care less noticed while doing the dishes”.
BTW I use a differential passively equalized open-loop JFET preamp myself (I have posted it here, sans values) using the now unavailable Toshiba matched pairs. For phono you don't need anymore than that and it can't oscillate (easily). It's also pretty RF immune. I could have made a hybrid out of an appropriate op-amp, I doubt if I could hear the difference in a true DBT.
CG – This is a composite amp and there a lot of gain in front of the buffer. Walt Jung and Jim Williams have published numerous articles on them, I suppose there’s nothing to learn there either. No one picked-up up on the fact that there’s really maybe 180dB of Aol that’s almost 140dB of NF (making up numbers, one could compute it). This is generally the same technique LT used in that super oscillator app-note to obtain “0” distortion.
Popular HS op-amps frequently are in processes that have poor beta*VA so the simplest folded cascode stages have Aol’s of only a few thousand so they result in very high 3dB points on Aol. So I’m to believe that a little bootstrapping or additional cascoding to improve Aol (with out changing unity gain crossover or input stage transfer function) “degrades” the sound? There’s an old trick on op-amps that had null pins where you could tune the Aol to anything you want.
Time for another round of the confederacy of subjectivists congratulating each other and commenting on how deaf the rest of us are.
scott wurcer said:CG – This is a composite amp and there a lot of gain in front of the buffer. Walt Jung and Jim Williams have published numerous articles on them, I suppose there’s nothing to learn there either. No one picked-up up on the fact that there’s really maybe 180dB of Aol that’s almost 140dB of NF (making up numbers, one could compute it). This is generally the same technique LT used in that super oscillator app-note to obtain “0” distortion.
Popular HS op-amps frequently are in processes that have poor beta*VA so the simplest folded cascode stages have Aol’s of only a few thousand so they result in very high 3dB points on Aol. So I’m to believe that a little bootstrapping or additional cascoding to improve Aol (with out changing unity gain crossover or input stage transfer function) “degrades” the sound? There’s an old trick on op-amps that had null pins where you could tune the Aol to anything you want.
Time for another round of the confederacy of subjectivists congratulating each other and commenting on how deaf the rest of us are.
Scott,
Just *where* are you getting this all from?
Speaking just for myself, I only questioned the very low load impedance attached to the opamp and what effects it would have. It looks to me like the output stage would be headed out of the Class A region into an area where cross-over distortion might be a concern. Isn't that the very reason why you made the special connections in the AD797 design? Are you saying now that was all fluff and unnecessary?
Why the defensive and accusatory posture?
This isn't much fun or very educational. If I wanted this abuse, I'd write for the Globe and read every one of the reply comments...
The composite amp approach is a valid alternative when you are only doing audio. There are plenty of chip amps intended for 4 or 8 ohm loads that are certainly not class A biased. There have been several threads here putting a high quality op-amp in front and compensating the composite, it's a little like a nested feedback approach too.
Yes it uses more gain and feedback to achieve the same result. Whatever crossover distortion residual there is is considerably reduced here probably into the sub-ppm level. A hotter buffer may or may not improve it. Remember the OPA554 was criticised
originaly without regard for output bias level.
Sorry my comments are more aimed at the dissing out of hand any circuit that uses op-amp principles as having a unique bad sound. This is usually offered as a proven fact. It does get irritating.
Never touch the Globe.
Yes it uses more gain and feedback to achieve the same result. Whatever crossover distortion residual there is is considerably reduced here probably into the sub-ppm level. A hotter buffer may or may not improve it. Remember the OPA554 was criticised
originaly without regard for output bias level.
Sorry my comments are more aimed at the dissing out of hand any circuit that uses op-amp principles as having a unique bad sound. This is usually offered as a proven fact. It does get irritating.
Never touch the Globe.
Scott,
>>>This is usually offered as a proven fact. It does get irritating.<<
I was offering it as 30 years of EXPERIENCE, not a proven fact.
What irritates me is engineers not being willing to try my experience gained ideas "because they can't work" - but I'm not aiming this at you.
Regards, Allen (Vacuum State)
>>>This is usually offered as a proven fact. It does get irritating.<<
I was offering it as 30 years of EXPERIENCE, not a proven fact.
What irritates me is engineers not being willing to try my experience gained ideas "because they can't work" - but I'm not aiming this at you.
Regards, Allen (Vacuum State)
As I said I personally have tried both, I love Nelson's stuff and the open-loop JFET stuff too. It's amazing how good the simplest circuits can get at certain points in the signal path. I frankly still find the transducer ends of things to totally dominate. Highish output impedances and real world speakers sort of obviate any listening tests there.
I've listened to a lot of things from both worlds rarely was anyone driven out of the room. The worst highend system was a top Cello setup I almost did run away from that. I've enjoyed the music through Conrad-Johnson as well as Hafler.
Most of the solid state PA threads evolve around one of a few high power op-amp structures, I don't see the point in telling these people they are wasting their time on "dead end" technology. The project to put together a round robin listening test on an op-amp line stage seems to have fallen apart, too bad.
I've listened to a lot of things from both worlds rarely was anyone driven out of the room. The worst highend system was a top Cello setup I almost did run away from that. I've enjoyed the music through Conrad-Johnson as well as Hafler.
Most of the solid state PA threads evolve around one of a few high power op-amp structures, I don't see the point in telling these people they are wasting their time on "dead end" technology. The project to put together a round robin listening test on an op-amp line stage seems to have fallen apart, too bad.
Allen Wright said:
There are many ways to skin the cat, as John Roberts (former Peavey console designer) uses to say. Some follow beliefs and fashions, some don't. Here on the forum once was told that it was arrogant to post my schematic near a legendary one, because mine was much simpler giving excellent results. They tried to explain why my schematic is bad, but failed. Something was wrong. What was wrong? Cheating. Like, jumping over the labyrinth drawn on the floor instead of searching for the path is cheating. But why all that complexities are needed, if the end result is to come to the point B from the point A? Like, I hurt feelings of people who believe that the solution can't be so simple.
Syn08 did the job of skinning the cat. He used a triple cascode to convert voltages to currents; how accurate is this conversion, is the question. To address the issue he converted currents to voltages back, by an amp with low output resistance, in order to sweep that errors of the voltage to current conversion under the feedback rug of very low feedback resistance to add less noises. Other approaches would be valid as well, but it is his child, and made exactly how he wanted to make it. It works as intended, it sounds pleasing his ears, bravo.
Allen Wright said:
Allen your phono stages look great. The split RIAA one is closest to what I have used for years except I went all JFET and 15V supplies. I run the input pair "naked" no source resistors or snubbers but at 5mA a side. I have no idea why the input stage would be unacceptible with the National parts the snubbers should take care of any RF. I don't know what process the MAT02's are in but the LM's are possibly much older technology. Ever tried the THAT parts?
The comment about the high feed back op-amp "sound" doesn't have any wiggle room, and you're not the only one that says it. I just don't see thirty years of process advances doing nothing. There is happiness for some with the Pacific RIAA and others with paralleled 6DJ8'S. We can live in harmony
I have found that 1:1 transference of tube optimized circuits to solid state is usually less good than the early tube circuits, except for noise. This is because of the added non-linear capacitance of bipolar devices and fet's, and the less linear transfer function of the solid state devices.
Scott,
We used to use the 2SK147 duals in the bottom of the cascode, at more like 10mA each, and of course no source R's to get enough gain for MC carts.
But it was never enough gain so I thought, bipolars have a huge amount more gM than jfets if run at decent currents, so gave the LM394 a try - also at circa 10mA.
My woman did run from the room when asked to listen to it - saying that I would ruin any reputation I had if I played it to people! I put the jfets back and she said "that's how it was and that's what I like".
Not to be defeated, I tried several bipolars and the MAT02 worked really well, and Johanna liked it as as much as the jfets. I have asked several people at AD about the MAT02/SSM2210 process but no one can give me any info.
In the NS literature, the LM394 is claimed to be a large number of paralleled devices. I suspect the MAT02 is as well, but if anyone knows anything more - I'd love to hear it.
regards, Allen (Vacuum State)
We used to use the 2SK147 duals in the bottom of the cascode, at more like 10mA each, and of course no source R's to get enough gain for MC carts.
But it was never enough gain so I thought, bipolars have a huge amount more gM than jfets if run at decent currents, so gave the LM394 a try - also at circa 10mA.
My woman did run from the room when asked to listen to it - saying that I would ruin any reputation I had if I played it to people! I put the jfets back and she said "that's how it was and that's what I like".
Not to be defeated, I tried several bipolars and the MAT02 worked really well, and Johanna liked it as as much as the jfets. I have asked several people at AD about the MAT02/SSM2210 process but no one can give me any info.
In the NS literature, the LM394 is claimed to be a large number of paralleled devices. I suspect the MAT02 is as well, but if anyone knows anything more - I'd love to hear it.
regards, Allen (Vacuum State)
Allen Wright said:
Read the datasheet and view the chiplayout picture and you will see how it is made.
http://pdf1.alldatasheet.co.kr/datasheet-pdf/view/48991/AD/MAT02.html
Allen Wright said:
That I can find out. I know there is a folklore joke around about the LM394 but I think they all are now optimized cross-quads of multi-stripe devices. Philips gave a paper deriving the optimum size from the sheet-rho of your metal/diffusions. They built a .25nV diff-pair and made a crude op-amp as a demonstration vehicle. The feedback network was 14-7 Ohm base resistors 7 in series, seven in parallel for G = 49. The input pair ran at 25mA a side IIRC.
BTW I think THAT uses an advanced SOI process on their stuff.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.