Discrete sounds better than integrated? a possible reason

[Winston OBoogie]

"... Without getting too esoteric, but warning what follows is pretty esoteric ... "

I found it interesting John.
I suppose, were we building a phono pre today (less likely than 35+ years ago of course), or operating an op-amp at high noise gain in a summing bus (not the best way to do it as you know ), then a better device would be one where the compensation is brought out externally.
Naturally this isn't as 'plug-and-play' as just wrapping feedback around a unity-gain stable device and demands some more knowledge on the part of the 'designer' ...  But a two-pole, or output transitional compensation scheme that's way above your highest RIAA time constant, or keeps phase at a healthy distance (say 12dB) from instability with the high noise gain summing would be good.

Perfect application for discrete in my book .  But having access to, and implementing a more suitable compensation scheme in an I.C.  would do it. 
No?

 

[L´Andratté]

Still brooding over if or why discrete sounds better than integrated... ;D

 

[abbey road d enfer]

Still brooding over if or why discrete sounds better than integrated... ;D

The answer is it depends...

 

[ruffrecords]

Still brooding over if or why discrete sounds better than integrated... ;D

To summarize and simplify what I said in the very first post.

Some discrete amplifiers from long ago used a technique whereby the amount of NFB was more or  less constant no matter what value the gain was set at. Apart from considerably simplifying managing the stability of the amplifier it meant that you could then have the same amount of NFB at all frequencies in the audio spectrum so any distortion the circuit produced was the same at all audio frequencies and all gains.

OP amps do not use this technique. Instead they take a brute for approach to stability. The result of this is that the amount of NFB declines by 6dB per octave with increasing frequency in the audio band and it also decreases as gain increases. Less NFB means more distortion so op amp distortion increases at 6dB per octave with increasing frequency in the audio band and it also increases as gain increases.

I am suggesting these differences may explain the difference in tone between them.

Cheers

Ian

 

[matriachamplification]

Others have suggested that the transformers are a major contributor

Like most all things this took me some time to absorb and digest prior to commenting.

Without going into the pshycological or biological imprints sound can triggers... I am all about the placebo effect. If my subconcious jives with it, I'm good.


On the subject of transformers; although my exprience deals with tube amps, we can officially say there is a huge sonic impact transformers have on tone. My neigbour Ted (passed away) was a Canadian pilot in WWII and a amp designer. I remember him explaining to me that transformers were speedbumps, they almost always leave a impact on sound. We did some testing with a Sunn Model T and Marshall Superbass that was quite fun.

That said, even in the case of Helios where little transformers were used, this "should" impact sonic properties to some degree. Outside of that... its not Vacuum Tube!

Thanks for taking the time to read my thoughts.

Be Well

 

[abbey road d enfer]

On the subject of transformers; although my exprience deals with tube amps, we can officially say there is a huge sonic impact transformers have on tone.

Indeed. Nobody denies that. The ways transformers impact the electrical response are well known.
The B-H curve, the Barkhausen effect, the resonances due to capacitance and leakage inductance, the eddy current losses, all this is extremely well documented.
What is still somewhat a mystery is why all these faults result in something that is musically pleasant.
It can partially be explained by the fact that most of these effects tend to soften transients.
Another factor is acquired taste, but it can't explain everything.

 

[matriachamplification]

Indeed. Nobody denies that. The ways transformers impact the electrical response are well known.
The B-H curve, the Barkhausen effect, the resonances due to capacitance and leakage inductance, the eddy current losses, all this is extremely well documented.
What is still somewhat a mystery is why all these faults result in something that is musically pleasant.
It can partially be explained by the fact that most of these effects tend to soften transients.
Another factor is acquired taste, but it can't explain everything.

Thanks for the input about transformers.

My grasp of NFB and looped fixed gain is weak at best. When I am better versed maybe I will stumble upon some insight.

I do like the idea of imperfections creating perfections. Sounds poetic. 

Be Well

 

[JohnRoberts]

I found it interesting John.
I suppose, were we building a phono pre today (less likely than 35+ years ago of course), or operating an op-amp at high noise gain in a summing bus (not the best way to do it as you know ), then a better device would be one where the compensation is brought out externally.

I killed way too many brain cells thinking about this stuff.

Way back when, all opamps required external compensation or stipulated non-zero minimum closed loop gain. For example the ever popular ne5534 was only stable down to closed loop gains higher than +10dB without additional compensation.  The much disrespected LM741 was the first internally compensated unity gain stable integrated circuit op amp (and slower than dirt). 

Naturally this isn't as 'plug-and-play' as just wrapping feedback around a unity-gain stable device

Just about is, modern premium op amps could probably beat that pants off of many well regarded legacy designs.

and demands some more knowledge on the part of the 'designer' ...  But a two-pole, or output transitional compensation scheme that's way above your highest RIAA time constant, or keeps phase at a healthy distance (say 12dB) from instability with the high noise gain summing would be good.

Good luck finding an off the shelf op amp with flat open loop gain to above audio bandpass, but technically the highest RIAA time constant is a pole at 75 uSec (so only roughly 2kHz).

Perfect application for discrete in my book .  But having access to, and implementing a more suitable compensation scheme in an I.C.  would do it. 
No?

I am less confident in my ability to design a discrete op amp better than I can buy off the shelf. The last time I rolled my own DOA was back in the late 70s/early 80s for use inside a console sum amp because the LM394 super match pair made a much lower input noise op amp than I could purchase. (I need to mention our own Sam Groner who has published some very respectable DOA designs.)

For large bus structures there are better ways and I have been writing about them since the 80s, like Paul Buff's "Transamp" which is a variant on Cohen's topology where open loop gain increases with closed loop gain for constant loop gain margin, and my current source summing that delivers modest (low) noise gain.

JR 

 

[analag]

Older is not better because its older. The older I get the more certain I am about that.

JR

PS: RIP for Rupert

I guess you have never heard about...wine?

All my custom gear are discrete electronics. They do sound better if they are built right!

 

[Whoops]

I guess you have never heard about...wine?

Older Wine can become vinegar

All my custom gear are discrete electronics. They do sound better if they are built right!

Like any other piece of gear discrete or not, if they are designed and built right they will sound "better"

 

[clintrubber]

To summarize and simplify what I said in the very first post.

Some discrete amplifiers from long ago used a technique whereby the amount of NFB was more or  less constant no matter what value the gain was set at. Apart from considerably simplifying managing the stability of the amplifier it meant that you could then have the same amount of NFB at all frequencies in the audio spectrum so any distortion the circuit produced was the same at all audio frequencies and all gains.

OP amps do not use this technique. Instead they take a brute for approach to stability. The result of this is that the amount of NFB declines by 6dB per octave with increasing frequency in the audio band and it also decreases as gain increases. Less NFB means more distortion so op amp distortion increases at 6dB per octave with increasing frequency in the audio band and it also increases as gain increases.

Hi,

Nice thread! Didn't manage to identify those dual setting sections in the design-examples you mentioned, do you have any direct links to schematics where both open & closed loops gains are changed in tandem?

Would be interesting and possible to experiment with this (one day...);
if certain circuits don't have the open loop setting possibility, it could be added.

For 'opamp'-open loop gain, the biasing could be changed (say CA3080, LM13600/13700), or the degeneration resistor values of the first diff pair could be changed (/added).

Bye 

 

[ruffrecords]

Hi,

Nice thread! Didn't manage to identify those dual setting sections in the design-examples you mentioned, do you have any direct links to schematics where both open & closed loops gains are changed in tandem?

Would be interesting and possible to experiment with this (one day...);
if certain circuits don't have the open loop setting possibility, it could be added.

For 'opamp'-open loop gain, the biasing could be changed (say CA3080, LM13600/13700), or the degeneration resistor values of the first diff pair could be changed (/added).

Bye

Here is a link to the three transistor Neve input stage.

http://www.jlmaudio.com/neve_ba283.htm

The gain of the first stage (TR1) and hence the open loop gain is determined by the ratio of R4 to the parallel combination of R10, R11 and any resistor connected between pins T and V. At the same time, the closed loop gain is determined by the ratio of R9 and the same parallel combination of R10, R11 and any resistor between T and V. To change the gain you add a resistor across pins T and V which both incresaes the open loop gain and the close loop gain.

Here is a link to the Helios three transistor mic pre scehmatic:

https://groupdiy.com/index.php?topic=58393.msg742685#msg742685

The gain of the first stage (VT1) and the open loop gain is set by the ratio of R5 to the series combination of R3 and any resistor connected from C5 to ground. At the same time, the closed loop gain is set by the ratio of R4 to the value of any resistor connected from C5 to ground. To increase the gain you reduce the resistance from C5 to ground which increases both the open loop and closed loop gains.

Hope that helps.

Cheers

Ian

 

[L´Andratté]

What elegance! Very obvious now that you point it out ( :) but totally escaped me before, thanks for pointing that out! And thanks clintrubber for asking ;D

 

[abbey road d enfer]

When transistors were expensive, many preamps used just two of them, and benefitted from the same advantages.
When transistors became cheaper, many used differential stages, which traded the disadvantage of having varying gain margin against the convenience of eliminating coupling caps.
AFAIK I've never seen a transimpedance amp that eliminates coupling caps in the signal path.

 

[JohnRoberts]

Here is a link to the three transistor Neve input stage.

http://www.jlmaudio.com/neve_ba283.htm

The gain of the first stage (TR1) and hence the open loop gain is determined by the ratio of R4 to the parallel combination of R10, R11 and any resistor connected between pins T and V. At the same time, the closed loop gain is determined by the ratio of R9 and the same parallel combination of R10, R11 and any resistor between T and V. To change the gain you add a resistor across pins T and V which both incresaes the open loop gain and the close loop gain.

Here is a link to the Helios three transistor mic pre scehmatic:

https://groupdiy.com/index.php?topic=58393.msg742685#msg742685

The gain of the first stage (VT1) and the open loop gain is set by the ratio of R5 to the series combination of R3 and any resistor connected from C5 to ground. At the same time, the closed loop gain is set by the ratio of R4 to the value of any resistor connected from C5 to ground. To increase the gain you reduce the resistance from C5 to ground which increases both the open loop and closed loop gains.

Hope that helps.

Cheers

Ian

Open loop gain that tracks with closed loop gain is a well explored characteristic of the "cohen" topology (also used inside modern mic preamp ICs).

I mentioned this in passing in my 1980 console performance article (when describing Transamp used in a summing amp).

JR

 

[ruffrecords]

Open loop gain that tracks with closed loop gain is a well explored characteristic of the "cohen" topology (also used inside modern mic preamp ICs).

You gotta spill the beans on which modern mic preamp ICs use this technology.

I mentioned this in passing in my 1980 console performance article (when describing Transamp used in a summing amp).

JR

Yes, the Transamp I think was the first 'op amp' to employ this technique but I was not aware it was now embedded in ICs.

Cheers

Ian

 

[abbey road d enfer]

You gotta spill the beans on which modern mic preamp ICs use this technology.

the SSM2015 was launched in 1983!

 

[JohnRoberts]

You gotta spill the beans on which modern mic preamp ICs use this technology.Yes, the Transamp I think was the first 'op amp' to employ this technique but I was not aware it was now embedded in ICs.

Cheers

Ian

The transamp is not an op amp but hybrid combination of discreet bipolar transistors and IIRC a TL072 IC op amp, all inside a potted module.

I don't feel like making a list but pretty much any high performance IC mic preamp uses that topology (Cohen), for the obvious benefit.

Not a secret...

JR 

 

[clintrubber]

Here is a link to the three transistor Neve input stage.

...

Here is a link to the Helios three transistor mic pre scehmatic:

...

Cheers

Ian

Hi Ian,

Thanks for those links, hadn't realized before there's indeed a pretty elegant 'tandem action' going on in those topologies.

If this is indeed the main reason for things 'sounding better' then that'd be a nice (re-)discovery, begging to implement it 'for integrated' as well (saw the Cohen example mentioned, need to dig it up & have a look again).

How do DOAs fit in here? They'd be suffering from the non-tracking open loop gain consequences as well, although I expect DOAs don't reach the same open loop gain levels as integrated opamps do.

Bye

 

[abbey road d enfer]

If this is indeed the main reason for things 'sounding better' then that'd be a nice (re-)discovery, begging to implement it 'for integrated' as well (saw the Cohen example mentioned, need to dig it up & have a look again).

This topology is in use since ages (1980's) in most british mixers (Soundcraft, Amek...) and many others, where it makes a difference, i.e. in stages with a high noise gain, that is, typically, mic preamps and summing amps. Do they "sound better"?

How do DOAs fit in here? They'd be suffering from the non-tracking open loop gain consequences as well, although I expect DOAs don't reach the same open loop gain levels as integrated opamps do.

DOA's have their own raison d'être, but in this respect tend to perform less well than recent opamps with GHz response. They can be used in a transamp topology, though.