Open loop bandwidth in op-amp.

[lumanauw]

I've got a question. I compared LM4562 with AD827.
AD827 has OL gain less than 80dB, flat until 10khz.
LM4562 has OL gain about 130dB, no flat section, just descending.

But hearing test shows LM4562 sounds better than AD827.

Looking at OL gain, OL bandwith, graph, will not tell you how an opamp will sound?

 

[DSP_Geek]

The AD827 has higher distortion, with 2nd harmonic less than 0.01% at 3V out and 1k load. However, the distortion isn't characterized with regard to signal level, so there might be more distortion at lower outputs. This isn't impossible: some older video opamps underbiased the output stage to keep the current stable with respect to temperature. This wouldn't mess up video signals too badly, but it'd play hell with audio.

The two parts really differ in power supply rejection ratio (PSRR). The 827 has only about 75 dB, whereas the 4562 has about 110 to 120 dB up to 10 kHz. That could be an issue if you have crud floating around on the supply lines: some applications still use 5532s, which have good PSRR, for that very reason.

 

[unclejed613]

yup..... it's called "gain error"...... and a short description about it is here........ http://www.elexp.com/t_gain.htm

did you measure the distortion and noise in both circuits?

you really can't hear anything that can't also be measured......

do a phase comparison too......

the ad827 has a noise figure of 15nv/sq rt hz, a Zin of 300k, and a THD of -90db, while the same measurements for the lm4562 are a noise figure of 2.7nv/sq rt hz, Zin of 1000M, and a THD of -100db.

also, according to the data sheets, in-band (1hz-100khz) phase margin of the ad827 begins at over 100degrees and settles at 90 degrees(with a sharp shift between 5khz and 10 khz), while the lm4562 is constant at 80 degrees, which means that within the audio bandwidth, signals through the ad827 have varying phase relationships for different frequencies, while through the lm4562, the phase relationships remain constant. while the ear is not a very good "phase meter", music with jumbled phase relationships just doesn't sound as good as music with constant phase relationships.

interpreting data sheets is more than just looking at open loop bandwidth, etc.... there's a lot of information to be digested, and it helps to have a general "quick glance" set of specs you are looking for. then after you have narrowed down a few choices, you can weed through the remaining data sheets in more detail. after that, if you still have multiple choices of a device, you can build a comparative prototype and do listening tests. a ZIF (Zero Insertion Force, an IC socket with a lever to open and close the pin contacts) socket can be helpful if you build a listening test prototype for several devices with the same pinout, and you can have a friend help you with a double-blind test. the best way to do this is to shave the numbers off the devices and put "A" and "B" on them and do an ABX test and have your friend do the chip swapping and record keeping. with a ZIF socket, all that will change is the chip, and all of the rest of the circuit components will remain the same, with no chance for overheating the chip or cold solder joints between tests.

 

[jcx]

The two parts really differ in power supply rejection ratio (PSRR). The 827 has only about 75 dB, whereas the 4562 has about 110 to 120 dB up to 10 kHz...

I see the LM4562 data sheet curves only sometimes supports this

But I also believe in the predictions of "A General Relationship Between Amplifier Parameters, And Its Application to PSRR Improvement"

http://citeseer.ist.psu.edu/cache/p...hip-between.pdf

which gives:

1/(CMRR) + 1/(PSRR+) + 1/(PSRR-) = (1/Adiff) * Zload/(Zload+Zout)

I choose to believe that the LM4562 data sheet graphs were somehow mixed up and the above relation between open loop diff gain, psrr, and cmrr still has to hold

At least the LM4562 datasheet +/-12 Vs +psrr curve goes through 80 dB @ 10 KHz which seems consistent with the diff gain curve, satisfying the above relation – that one of the “parasitic” gains should be expected to be of the same magnitude as the diff gain

 

[lumanauw]

While the title of this thread "Open loop bandwith in opamp", it seems only seeing this one alone doesn't corelate tightly to good or bad sounding. There are other more important data that corelates better to good sounding.

 

[DSP_Geek]

At least the LM4562 datasheet +/-12 Vs +psrr curve goes through 80 dB @ 10 KHz which seems consistent with the diff gain curve, satisfying the above relation – that one of the “parasitic” gains should be expected to be of the same magnitude as the diff gain

This device seems to be happiest at 15 volts, since a lot of the parameters worsen considerably at 12 V. As for the equation, I'll need to sit down with the paper and a cup of coffee.

 

[Chucko]

In fact, in 1974, I asked Bob Widlar, the designer of the ua702, 709, LM301 (741 is derived from the 301) etc, etc, and the greatest designer of the IC op amp of all time. He said that IC's up to that time were really not made or optimized for audio, so we should build our own circuits.

Widlar was indeed a master of the linear IC technology of the time. It was not feasible to make fast, high beta PNP transistors in his era. One of Widlar's talents was finding clever ways to use the crummy PNPs so as to minimize their deficiencies.

Today high-quality complementary bipolar transistors are a given in linear IC design. But given the vast numbers of op amps used in audio, from low-end consumer to high-end studio applications, I'm surprised at how little progress has been made on chip design specifically for audio.

TIM or SID, which are the same thing, are not directly related to open loop bandwidth in most designs, and that is all that we could measure that was different than our already precision IM or THD measurements. However, Otala persisted in recommending 20KHz open loop bandwidth, but realized that his original hypothesis regarding TIM being the only factor was incomplete, and other distortions must be present that are audible to the trained ear. We still think this today, but we don't have a standardized test to measure PIM or any other distortion that might be important.

Are you aware of any tests for measuring PIM? I'm not up to speed on the literature.

What concerns me on this thread is the amount of 'guessing' that many people do about why circuits are designed the way they are.

A lot of us are amateurs. And some of us are just picking this stuff up for the first time. I've been interested in audio circuit design since the early '70s, but I don't have any formal training in electronics.

 

[lineup]

Yes Sir. I would like to ask if you intend to offer finished products or kits ?
Thanks and regards,

beppe

beppe
Thanks very much for asking Dear Sir.

Lineup Audio Lab
... will never have anything to do with money buy/sell.

We are pure DIY Audio Amplifier designers.
So, whatever we will publish is to be 100% free.
And you will have to get the stuff yourself.
If you want to try to build our from schematics and instructions.

Acually, Not much different from here at www.diyaudio.com
... and here 97.3% of everything is still free


Regards
lineup

 

[andy_c]

Are you aware of any tests for measuring PIM? I'm not up to speed on the literature.

Bob Cordell built an instrument for PIM measurement back in the '80s. His article, with a block diagram of the instrument, is here

 

[peranders]

We are pure DIY Audio Amplifier designers.
So, whatever we will publish is to be 100% free.
And you will have to get the stuff yourself.
If you want to try to build our from schematics and instructions.

1 Who are "we"? No info about who you are, or we for that matter.

2 Where are the stuff you are talking about? The homepage is empty.

 

[beppe61]

beppe
Thanks very much for asking Dear Sir.
Lineup Audio Lab
... will never have anything to do with money buy/sell.
We are pure DIY Audio Amplifier designers.
So, whatever we will publish is to be 100% free.
And you will have to get the stuff yourself.
If you want to try to build our from schematics and instructions.
Acually, Not much different from here at www.diyaudio.com
... and here 97.3% of everything is still free
Regards
lineup

Well, thank you very much for your kind reply and advice.
I will follow the evolution of the mentioned project with great interest.
Kind regards,

beppe

 

[jcx]

I see the LM4562 data sheet curves only sometimes supports this

But I also believe in the predictions of "A General Relationship Between Amplifier Parameters, And Its Application to PSRR Improvement"

http://citeseer.ist.psu.edu/cache/p...hip-between.pdf

which gives:

1/(CMRR) + 1/(PSRR+) + 1/(PSRR-) = (1/Adiff) * Zload/(Zload+Zout)

I choose to believe that the LM4562 data sheet graphs were somehow mixed up and the above relation between open loop diff gain, psrr, and cmrr still has to hold

At least the LM4562 datasheet +/-12 Vs +psrr curve goes through 80 dB @ 10 KHz which seems consistent with the diff gain curve, satisfying the above relation – that one of the “parasitic” gains should be expected to be of the same magnitude as the diff gain

the paper link moved, today I found it at:

http://citeseer.ist.psu.edu/cache/p...zcas91.pdf/a-general-relationship-between.pdf

google or citeseer should be able to find it given the reference info:

E. Sackinger, J. Groette, and W. Guggenbuhl, “A General Relationship Between Amplifier Parameters, And Its Application to PSRR Improvement,” IEEE Trans. Circuits And Systems, vol. 38, pp 1171-1181, October 1983

 

[lumanauw]

PSRR

Thanks JCX for the link

I read your paper, also read Doug Self and Bruno's post about PSRR
http://www.diyaudio.com/forums/showthread.php?postid=458167#post458167

What is the mechanisme that the VAS has Zero PSRR at open loop?

The paper and Bruno seems to say that PSRR loss in VAS stage is due to miller cap (one end connecting to supply, the other end connecting to VAS output)

But DougSelf said that Miller cap itself is not to blame for bad PSRR in VAS stage. He experimented with buffer (so signal can only leaving, not able to enter), and PSRR is the same.

 

[foo]

I was so curious to see that someone said he saw the OLG plot of LM4562.
I thought the datasheet was renewed and I downloaded it again.
Sadly, there's nothing new.
Of course, no OLG plot as well.
That plot is what I wanna see first aquired the datasheet and what missing in the datasheet.

I would be glad to see the OLG plot of LM4562 or LME49710.

 

[foo]

I found it in an old version of datasheet.
It surely perform not that good for the linearty of maximum OLG.
But just like other good audio OPAMPs, due to it's very high initial OLG, it still has about 75db OLG at 20khz, which is a very good result among existing audio OPAMPs.

BTW, for OLG within audio frequency, OPA637 LT1115 THS4031 all perform well.