Very interesting!!
So you think settling time is quite vivid?
I have a little story, there are fake OPA627's on the internet these days, a Russian person was curious and opened them up and took detailed pictures of the circuit. He identified it was some kind of Analog Devices chip, then he or someone else looks at the specs and found quite a lot of similarities to the OPA627, I think they had identical settling time.
In my listening, to this fake OPA627, I actually think it sounds pretty similar to the real OPA627, lol......
Thus, I became a little curious if my perceived similarity is correlated to the settling time spec. However, this is my only example for now, perhaps I will make an excel sheet later of op-amp's and perform a kind of blind test.
Have some evidence that it's your design, like a dated forum post or dated internet site or something like that.
I've seen measurements which say that the discrete op-amp's by Burson are identical to the op-amp's by Audio-gd.
Either Audio-gd copied Burson or vice-versa.
I'm just curious, is there a special reason the amp in that other thread looks too complex?
Good luck with your JC2014.
LT1363/LT1223 is a black box for me. I think its complexity equals to a whole amp. I would be hesitate to use it.
If you unveil the internal circuits of the IC, you will find what I am talking about.
Because of wrong style of compensation..Connect C12 right way, from colector Q10 to base Q14, not to emitor (low impedance).. Increase R35, R36 at least by one order.. It has no sense leave this stage to work with 35mA current. And you can ommit C2,R7 completly. Output phase tells quite nothing about phase margin (stability)
R2 should be the same as R20 (DC balance, output ofset), move R2 to parallel with C6, change R6, R8, R34,R13..Look at attachment.
Last edited:
Think about D8 purpose. But if needed it is possible to improve clipping behavior, few componets (R,D).
Agree with you about the settling time, many others also confirm especially when it relates to opamps that the faster settling types have better sound quality all other parameteres being similar.
There is relationship between phase margin and settling time, a phase margin of of around 78 degrees will result in optimum settling time and critical damping. If I remember correctly texas instruments have a paper about this or you could check the following as well.
http://class.ece.iastate.edu/vlsi2/docs/Papers Done/2000-08-MWSCAS-MS RG.pdf.
http://www.circuitsage.com/a2d/optPM.pdf
for some measurement systems "single pole" response is the only accepted option for fast settling - most settling time literature will be focusing on these sorts of measurement applications
in audio there is little evidence that typical small signal "long" settling tails that aren't accompanied by audible frequency response errors can be heard - and some listening test evidence against
on signal content alone it is extremely implausible that microsecond amplifier settling time is important to commercial recorded music signals that have been 4th order low pass filtered at 50 kHz or less (usually much less) by the combination of recording microphones and loudspeaker transducer's basic physics limits even without consideration of electronic processing or recording/playback limitations (phonograph records put 2 more 2nd order mechanical low pass transduction steps in the signal chain with cutting head mechanics and playback phono cartridge stylus/cantilever resonances)
then there is the data from actual psychoacoustic testing where Temporal Masking is well established and strong for forward masking with time constants in milliseconds - most temporal masking models are even single pole as well as using millisecond time constants
you can hand wave about peaking/settling and RF intrusion, IMD folding down - but I'd really like to see some numbers - for competent designs, practical phase, gain margins, good cables/proper shield terminations, typical 200 kHz input filter...
in audio there is little evidence that typical small signal "long" settling tails that aren't accompanied by audible frequency response errors can be heard - and some listening test evidence against
on signal content alone it is extremely implausible that microsecond amplifier settling time is important to commercial recorded music signals that have been 4th order low pass filtered at 50 kHz or less (usually much less) by the combination of recording microphones and loudspeaker transducer's basic physics limits even without consideration of electronic processing or recording/playback limitations (phonograph records put 2 more 2nd order mechanical low pass transduction steps in the signal chain with cutting head mechanics and playback phono cartridge stylus/cantilever resonances)
then there is the data from actual psychoacoustic testing where Temporal Masking is well established and strong for forward masking with time constants in milliseconds - most temporal masking models are even single pole as well as using millisecond time constants
you can hand wave about peaking/settling and RF intrusion, IMD folding down - but I'd really like to see some numbers - for competent designs, practical phase, gain margins, good cables/proper shield terminations, typical 200 kHz input filter...
The PCB aims to accomodate two pairs of TO3 metal can power transistors, but also compatible with TO264. Metal cans are said to have the least IM at radio frequency. I am not sure whether that is true. But I believe they have the best thermal stability
Any idea what power transistors are best for the output of this amp?
Any idea what power transistors are best for the output of this amp?
I appreciate your reality check technical concerns but in an op-amp like the OPA2111KP the sound is just too different to accept that it's not there.
In the spec sheet all I can see is an unusually low slew rate and a settling time at around 10 microseconds, but I'm open-minded to that it sounds different for another reason which I'm not aware of.
http://www.ti.com/lit/ds/symlink/opa2111.pdf
I'm pretty sure a transducer is capable of IR significantly lower than 10 microseconds and it should be able to convolve with the chaos of 44,100 samples per second but not extract more detail from there.
Last edited:
Just passing by, but I noticed the intriguing apparent symmetry of the VAS stage. (I'm referring to the schematic on Page 6)
In fact the top half is a Baxandall current-source and the bottom half an EF-VAS with the EF inverted (see p195 of Audio Power Amplifier Design 6th Edn)
Most artistic! I like it.
I should however say that I don't think the high output impedance of a Baxandall current-source is going to make any percptible difference.
<NITPICK>I assume the current-mirror degeneration resistors are mean to be 680R. Seems rather high- is there enough voltage for the mirror to work properly?</NITPICK>
In fact the top half is a Baxandall current-source and the bottom half an EF-VAS with the EF inverted (see p195 of Audio Power Amplifier Design 6th Edn)
Most artistic! I like it.
I should however say that I don't think the high output impedance of a Baxandall current-source is going to make any percptible difference.
<NITPICK>I assume the current-mirror degeneration resistors are mean to be 680R. Seems rather high- is there enough voltage for the mirror to work properly?</NITPICK>
Jcx,
1 second divided by 44,100 = 22.675736 microseconds per sample.
The shape of the 10 microsecond settling time should convolute the sinusoidal time flow from sample to sample and within samples to a reasonable extent the way I see it.
It's not like human hearing is only Hertz cycles and hair cells, divided up like keys on a piano, with the highest note on the piano at 50 microseconds, thus the alleged fastest time we can hear is 50 microseconds, or 20 kHz, as if they were precisely identical.
It's written that a tonal character under 4 ms is perceived as an atonal click, but we are superimposing a 10 microsecond settling time shape here unto everything and everywhere, so the convolution becomes perceived in seconds, since it's everywhere.
We're not claiming superhuman hearing of 500 nanosecond sounds, we're claiming hearing that 500 nanosecond sound multiplied by N million, thus affecting the total shape.
1 second divided by 44,100 = 22.675736 microseconds per sample.
The shape of the 10 microsecond settling time should convolute the sinusoidal time flow from sample to sample and within samples to a reasonable extent the way I see it.
It's not like human hearing is only Hertz cycles and hair cells, divided up like keys on a piano, with the highest note on the piano at 50 microseconds, thus the alleged fastest time we can hear is 50 microseconds, or 20 kHz, as if they were precisely identical.
It's written that a tonal character under 4 ms is perceived as an atonal click, but we are superimposing a 10 microsecond settling time shape here unto everything and everywhere, so the convolution becomes perceived in seconds, since it's everywhere.
We're not claiming superhuman hearing of 500 nanosecond sounds, we're claiming hearing that 500 nanosecond sound multiplied by N million, thus affecting the total shape.
Last edited:
Jcx if all op-amp's sound the same with perfectly flat FR and less than 0.001% THD then you must think users in audio forums are a rather insane lot.
Hello Douglas ,
My oppinion is that with the Baxandal topology the main advantage is the small temperature coefficient dependency of the output current .
But they ARE...! 😛
That little detail aside, it is perfectly possible to make opamps misbehave by mistreating them in all kinds of ways or stumbling into some shortcomings. If you build a single-9V cMoy with a TL072 and try to drive 32 ohm headphones with that, it will most probably sound like poo (and measure like that, too). Internal output impedance of this part is high, current capability and GBW modest, and they don't like capacitive loading much (of which a headphone cable offers aplenty) - and that's without even running into phase inversion or common-mode input limits in general. Yet, take that same part, run it on +/-15 V and build a line-level inverter with it (10k/10k, output iso 470R), and it may be totally inconspicuous.
As it happens, I have even written about this kind of thing recently. In a nutshell, many things (including opamps) just do not perform the same under any random circumstances. Implementation matters just as much as the part itself. Making sweeping assumptions from single use cases easily produces contradictions and confusion - and there sure is plenty of that among audiophile folk.
Last edited:
but how do we hear the difference between LME49710 and OPA627 when both are implemented correctly and running in highest performance?!
And, that is the question. After reading a very enlightening paper a few weeks ago written by a well-versed young engineer, I've kind of humbled myself and my past views of FR and HD being the end all. He worked in electronics and speaker design. Most companies/brands don't outright credit their product designers, so we don't usually hear from these people. The name rang a bell because he designed speakers for a couple bigger brands in the industry and custom commissions. I'll find the name/link later. Following his work put harmonic and frequency response FFT measurements into a sort of perspective- components to the larger machine of distortions, not nessecarily the main event. For most cases they are pretty descriptive. But I'm glad that he shared how under other design criteria 2 amps could measure pretty close in, say, two or three measurements by FFT/common methods and sound different because they measure different elsewhere. Very interesting work on harmonic phase. It seems he took it to the next tier. This part sticks in my memory... if the phase of the harmonic components are altered, 3rd order can sonically resemble 2nd order through temporal context. When those distortions are out of the hearing threshold, others still have to be dealt with. Very sensible, and its funny that we tend to forget this with time and skip to our cake.
Last edited:
Kouiky yes please find the link if you can, it sounds noteworthy.
Hi Bimo, just basic tests like I select the op-amp in a small box without looking, then I insert it without looking, then I listen and identify what it is, then I check.
Not scientific but satisfying enough for me.
I'm thinking about making two versions of "The Wire" later, one with LME49990 and one with something else, like you say with both checked to the same SPL, then I can run a test with someone elses help, for example they just insert the cable into X or Y rolling a dice at odd or even numbers and then I identify, repeat 20 times.
Hi Bimo, just basic tests like I select the op-amp in a small box without looking, then I insert it without looking, then I listen and identify what it is, then I check.
Not scientific but satisfying enough for me.
I'm thinking about making two versions of "The Wire" later, one with LME49990 and one with something else, like you say with both checked to the same SPL, then I can run a test with someone elses help, for example they just insert the cable into X or Y rolling a dice at odd or even numbers and then I identify, repeat 20 times.
Hi, it is a great pleasure to see your comments😀
The simulation tells there is a 1.119mA current through the mirror collectors, base current is about 2.5167 for the left and 2.4238 for the right.Voltage over the 690R risistor is 0.774V.
What is your idea how to leverage the high output impedance of a Baxandall current-source?
Cheers
Jazz
My pleasure.
I have no real idea how the high output impedance of the Baxandall current-source could be put to advantage. Here the VAS is loaded by the output stage, and it so it's hard to see how an unusually high current-source output impedance could do any good.
If the VAS is buffered from the output stage, then the high output impedance might increase the o/l gain at low frequencies where it is not controlled by the Miller capacitor. This might improve the PSRR, but not the THD.
- Status
- Not open for further replies.