Jan, I believe we both agree that simple VAS load by a resistor is no cure, though it provides higher OLG wo corner.
There are those here that have stated simply loading the VAS on an AD844 greatly improves the sound. Easy to try, does nothing for me. Neither does putting my cables up on Jenga blocks, yes I have tried it and heard nothing what more do you want? Self policed DBT no less 😀
Last edited:
Many of us, audio circuit designers, have our beliefs. Me, I am no exception. I agree with John that good discrete designs almost always sound 'better' even than best opamps.
My 'belief' is that reasons are in:
1) input stage with preferably not exponential transfer function. JFET is better than BJT, it is more linear and less susceptible to EMI/RFI interference.
2) input stage to be operated at high idle current and low noise.
3) high slew rate design, with symmetrical slewing, is better than lower slew rate design with non-symmetrical slewing
4) output stage to be operated at high idle current.
Opamps are usually compromised in all four points mentioned.
During listening tests, I have found no advantage of low-feedback or 'no-feedback' design against higher feedback design, in case they both were fast, with high slew rate and high idle currents. As opposite to John's opinion, no feedback designs were less accurate and unable to reproduce complex classical music as good as higher feedback designs.
My 'belief' is that reasons are in:
1) input stage with preferably not exponential transfer function. JFET is better than BJT, it is more linear and less susceptible to EMI/RFI interference.
2) input stage to be operated at high idle current and low noise.
3) high slew rate design, with symmetrical slewing, is better than lower slew rate design with non-symmetrical slewing
4) output stage to be operated at high idle current.
Opamps are usually compromised in all four points mentioned.
During listening tests, I have found no advantage of low-feedback or 'no-feedback' design against higher feedback design, in case they both were fast, with high slew rate and high idle currents. As opposite to John's opinion, no feedback designs were less accurate and unable to reproduce complex classical music as good as higher feedback designs.
If you want less PIM at a certain freq and signal level, there's a few things you can do. Remember that the cause (in this context) is Gm modulation due to high error signals. The error signal depends on both OL gain and feedback factor at the point of interest. What DOES help to keep Gm modulation, and thus PIM, low, is OL gain at the freq of interest, and that is NOT helped by squashing the gain at lower freqs. Trouble is, it is all interlocked: Gm, OL gain, feedback factor, compensation cap. The take-home here is that simple solutions like squashing the LF gain has no effect where you need it, at best, and may even make things worse.
As to the discrete vs integrated issue, it's dangerous to make sweeping general statements. I know a LOT op modern, extremely linear and wideband opamps, and I know a LOT of crummy and highly non-linear and distorting discrete 'creations'. YMMV.
jan
But, complementary-differential topology with JFET input as discussed here may need NO or very small compensating caps. I have several designs of this kind and hundreds of units built. They do not suffer from integrator compensation cap issue. And they do not suffer from Gm modulation due to high error signals.
Last edited:
If you want less PIM at a certain freq and signal level, there's a few things you can do. Remember that the cause (in this context) is Gm modulation due to high error signals. The error signal depends on both OL gain and feedback factor at the point of interest. What DOES help to keep Gm modulation, and thus PIM, low, is OL gain at the freq of interest, and that is NOT helped by squashing the gain at lower freqs. Trouble is, it is all interlocked: Gm, OL gain, feedback factor, compensation cap. The take-home here is that simple solutions like squashing the LF gain has no effect where you need it, at best, and may even make things worse.
As to the discrete vs integrated issue, it's dangerous to make sweeping general statements. I know a LOT op modern, extremely linear and wideband opamps, and I know a LOT of crummy and highly non-linear and distorting discrete 'creations'. YMMV.
jan
No GNFB in a preamp makes sense to me if you use a VGT volume control, like CH is doing.
In my opinion that’s the best way to make a preamp.
BTW: nice new avatar Pavel.
S
In my opinion that’s the best way to make a preamp.
BTW: nice new avatar Pavel.
S
Ehhh. OK. I suppose there is some reason for that? It sounds better to you, right?
jan
For the record, Walt first alerted me to the AD817 and later, the AD825. Yes, Scott, I had AD825 samples before you gave me some, a couple of years ago, but I will thank you again for giving me some as well. As Scott can tell you, some people, somewhere, seem to think the AD825 makes a wonderful line amp. Do you have any reasons as to why?
Yes, it sounds a lot better to me, but do I have a clue about sound?
Your opinion is of course NO.
S
I have never tried it as a line amp, BUT I designed it into a really advanced power amp, and it does a GREAT JOB. It serves as the 'reference' amp for the power amp.
Maybe because it has open loop bandwidth of ~10kHz?
Walt Jung mentions it in his paper.
A little bit of low frequency offset error fed back and the OLBW goes to 10Hz. Some old op-amps let you do this from the null pins. Difference in sound nada.
What is interesting to me is the mysterious existence of a COS term in the transfer function in Barrie Gilbert's op amp example. This apparently comes from the finite open-loop bandwidth. Any quibbles? Any nadas? Anything?
I would like to talk about something else for a moment, for completeness. Why did I move from a 'seemingly adequate' pre-preamp input stage that could be easy to make and quiet, to a more complex pre-preamp input stage that was just as quiet, but had 10 times more output for a given level of 3rd harmonic distortion, or 1/100th the 3'rd harmonic distortion at a given output level? The answer is: CCIR(F) high frequency intermodulation distortion. While both units worked OK at normal levels and measured OK with SMPTE or a harmonic distortion sweep, CLOSE two tone distortion, such as: 19,900 and 20,000 Hz would measure BADLY in the original JC-1 or the Sota head amp, AFTER post RIAA EQ in the phono stage. Think it through, you can get almost 100 times more distortion this way, because of the EXTERNAL base boost of the difference components. This made making the most LINEAR INPUT STAGE I could possibly make, necessary for best audio reproduction, and the Vendetta was born.
The AD825 was recommended and used in audio many years ago before you started to speak about it. There were small companies producing opamp replacement modules with the AD825, and some 8 years ago there was a real web rush about this circuit. Many people liked it, some not. The higher OLG corner and lower Ao is probably only one reason. The other reasons are JFET input stage, very high slew rate AND symmetrical slewing. There is a disadvantage of this circuit, high voltage noise. In case you need gain, the AD825 is useless. And, some find it less detailed compared to other excelllent opamps.
http://www.octave-electronics.com/lcaudio/ad825.shtml
http://www.soundlabsgroup.com.au/p/IC-Module-AD825/IC+Module+AD825
http://www.referenceaudiomods.com/M...reen=CTGY&Store_Code=RAM&Category_Code=OPAMPS
http://www.positive-feedback.com/Issue30/diy.htm
http://www.skratchlounge.com/index.php?/topic/4416-opamp-updatead8599-better-than-ad825-read-on/
http://www.head-fi.org/forum/thread/293845/sonic-comparason-ad825-vs-744
http://www.edaboard.com/thread1708.html
The last link shows discussion dated year 2002. But what is 9 years compared to 35 years?
Last edited:
- Status
- Not open for further replies.