Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Brian Roth]

I keep reading comments from folks that claim IC opamps sound terrible.  The Usual Suspects are TL07x and NE553x parts, apparently since they are so commonly used.

Yet, in other commentaries, others praise the sound of desks such as Neoteks, which are chock-full of TL07x chips.

I'm a bit at a loss now, since either of the two "awful" chip families measure quite well via any bench test I can concoct.  Of course, a TL07x into 600 Ohms at +20 dBu has problems due to limited output current, but that isn't the case with 553x chips.

Slew rate with either family shouldn't be an issue since they can deliver full signal-swing well into the 100's of kHz range (ignoring the current limiting problems mentioned earlier).

If these two common chip "families" are so audibly ghastly, then every common testing method we have available must be seriously flawed.

Comments?

Bri

[danlavry]

Brian Roth wrote on Tue, 10 May 2005 07:29

I keep reading comments from folks that claim IC opamps sound terrible.  The Usual Suspects are TL07x and NE553x parts, apparently since they are so commonly used. Yet, in other commentaries, others praise the sound of desks such as Neoteks, which are chock-full of TL07x chips. I'm a bit at a loss now, since either of the two "awful" chip families measure quite well via any bench test I can concoct.  Of course, a TL07x into 600 Ohms at +20 dBu has problems due to limited output current, but that isn't the case with 553x chips. Slew rate with either family shouldn't be an issue since they can deliver full signal-swing well into the 100's of kHz range (ignoring the current limiting problems mentioned earlier). If these two common chip "families" are so audibly ghastly, then every common testing method we have available must be seriously flawed. Comments? Bri

Hello Bri,

How good is an op amp? There are at least 2 ways to answer that.

The first way to look at an OP amp is subjective – how does it sound. The second way is objective – how well does it preserve the waveform (waveform in = waveform out).

Either way, when I listen and/or measure an OP amp, I make sure to keep my conclusion to a SPECIFIC design. One may have the same OP amp sound or measure OK in one circuit, and the same OP amp may be terrible in another circuit. There are many reasons for that!

Say you have an OP amp that works great in an inverting circuit configuration, but has a poor output current drive capability and a terrible power supply rejection (of noise). That OP amp will do well as an inverting stage, in an environment with a very clean supply, when the load it drives is very high resistance… But use the same part where the supply is noisy, the circuit is non inverting and the load current requirement is high, and it may disappoint you…

There are many factors that go into an OP amp and that is why there are so many OP amps devices on the market, I would not be surprised if there are 1000 devices out there. Clearly, each device offers a different set of compromises, and there is no such thing as a single device that will do it all…

Of course, there are some devices that “tend to shine” and often the price tag reflects it. There are also very many devices that are rather poor for audio, no matter what circuit one uses…

Regards
Dan Lavry
www.lavryengineering.com

[Sahib]

Hi Brian,

Yes I have come across those arguments about the op-amps but I think they are baseless. I have Walt Jung's book (1st edition)on audio opamp applications and at the time he mentions 5532-4 as an excellent sounding all round audio op-amp. This was also at a time when Neve desks were (and still are) considered as great which used 5534. So how come they are rubbish now? I think Dan has commented very well on the issue that it depends where and how one uses them. In the wrong hands I think even the best discrete module will sound bad.

Cemal

[Brian Roth]

Dan, I fully agree that there are many aspects that will impact the performance of an opamp in an audio circuit.  I clearly recall that nasty parts like 741s and 4558s were used in some gear!

I guess my main point is that *if* 5534s and TLo72s sound so ghastly, then what measurement technique can be used to come up with quantative numbers to support that?

Bri

[bobkatz]

Brian Roth wrote on Tue, 10 May 2005 23:53

Dan, I fully agree that there are many aspects that will impact the performance of an opamp in an audio circuit.  I clearly recall that nasty parts like 741s and 4558s were used in some gear! I guess my main point is that *if* 5534s and TLo72s sound so ghastly, then what measurement technique can be used to come up with quantative numbers to support that? Bri

Well, you can look at the curve of phase shift at high frequencies as you apply negative feedback versus the open loop gain. But how does that sound?  

BK

[danlavry]

Brian Roth wrote on Wed, 11 May 2005 04:53

Dan, I fully agree that there are many aspects that will impact the performance of an opamp in an audio circuit.  I clearly recall that nasty parts like 741s and 4558s were used in some gear! I guess my main point is that *if* 5534s and TLo72s sound so ghastly, then what measurement technique can be used to come up with quantative numbers to support that? Bri

I don't use the 5534 and I don't use the TL07X. It was long ago since I tested them at various configurations.

The 5534 is very inexpensive and can be made to yield good results in some circuits. The TL07X seems to yield pretty bad distortions in many circuit configurations, and I do not see it as an audio device.

What tests? One can write a book about testing analog, but when a device yields terrible distortions with a basic circuit, and basic sine wave testing, it is enough reason not to use it, in my book.

Regards
Dan Lavry

[Brian Roth]

Dan, that is perhaps my point, I just wasn't clear perhaps.  I've measured tons of gear that uses opamps, and even the pedestrian stuff commonly has excellent square wave response (bandlimiting being the obvious "flaws" to various degrees) as well as THD levels well into the noise floor over the audio band at "nominal operating level", and typically 0.01% just below clipping over the audio band.  The latter test does shake out the weaker designs at 10 or 20 kHz.

I need to breadboard a 553x and run some tests again.  It's been too long since I did that, but my recollection was that THD was extrememly low at any frequency and level (up to near clipping).

Perhaps 2-tone IMD tests with HF tones will reveal something different...haven't tried that.  Also, topology (inverting vs non inverting, output load-both R and C, closed loop gain, etc) could  result in variable test bench results.

In any case, if bench tests show vanishingly low THD and excellent square wave response, what additional tests can be run to show why chips like 553x's sound so (allegedly) awful?  Or, for that matter, even the higher priced Burr Brown, Analog Devices, or Linear Technology chips are allegedly ghastly as well.

Bri

[Brian Roth]

Bob, at low to moderate closed loop gain, any semi decently designed opamp circuit has extremely low phase shift at the extreme ends of the audio band.  Or did I miss your point?  <g>

Bri

[bobkatz]

Brian Roth wrote on Wed, 11 May 2005 23:01

Bob, at low to moderate closed loop gain, any semi decently designed opamp circuit has extremely low phase shift at the extreme ends of the audio band.  Or did I miss your point?  <g> Bri

I was actually referring to gain-bandwidth product. My only point was that apparently this is an important spec for an opamp. I wish I knew more about how to correlate the measurements with the audio as I've heard some "pencil-sounding" opamps, to describe their particular coloration, and inevitably seem to correlate with slew rate and drive capability even if bridge-loaded.

As to whether it has any effect on the real sound, I dunno, but I tend to use opamps with a high GBP and run the neg feedback conservatively for those reasons. But measurements? I never learned how to correlate anything I knew how to measure with the sound except maybe to steer away from opamps with a low GBP.

[danlavry]

bobkatz wrote on Thu, 12 May 2005 08:26

Brian Roth wrote on Wed, 11 May 2005 23:01 Bob, at low to moderate closed loop gain, any semi decently designed opamp circuit has extremely low phase shift at the extreme ends of the audio band.  Or did I miss your point?  <g> Bri I was actually referring to gain-bandwidth product. My only point was that apparently this is an important spec for an opamp. I wish I knew more about how to correlate the measurements with the audio as I've heard some "pencil-sounding" opamps, to describe their particular coloration, and inevitably seem to correlate with slew rate and drive capability even if bridge-loaded. As to whether it has any effect on the real sound, I dunno, but I tend to use opamps with a high GBP and run the neg feedback conservatively for those reasons. But measurements? I never learned how to correlate anything I knew how to measure with the sound except maybe to steer away from opamps with a low GBP.

Hi Bob,

You said:
"but I tend to use opamps with a high GBP and run the neg feedback conservatively"

That statement by itself can be a contradiction. Most Opamps with high GBP used in audio have a lot of open loop gain at audio frequency which means a lot of negative feedback. Most often, you choose a high GBP amp and you automatically end up with a lot of negative feedback.
(I am not talking for or against feedback, which is another issue).

Regards
Dan Lavry

[bobkatz]

danlavry wrote on Thu, 12 May 2005 17:12

You said: "but I tend to use opamps with a high GBP and run the neg feedback conservatively" That statement by itself can be a contradiction. Most Opamps with high GBP used in audio have a lot of open loop gain at audio frequency which means a lot of negative feedback. Most often, you choose a high GBP amp and you automatically end up with a lot of negative feedback. (I am not talking for or against feedback, which is another issue). Regards Dan Lavry

What I meant by the word "conservatively" was the choice of HF rolloff of the negative feedback in relation to the phase shift at audio frequencies. As a rule of thumb I've been going for an F sub C of about 300 kHz or perhaps higher.

[dcollins]

bobkatz wrote on Thu, 12 May 2005 14:57

What I meant by the word "conservatively" was the choice of HF rolloff of the negative feedback in relation to the phase shift at audio frequencies.

A lot of classic gear was struggling to be flat at 20kHz!  They didn't have enough gain before feedback to make it.  And the transformers phase-shift was such that they couldn't include them in the NFB loop!

How did they ever make records like that?

-3 at 300k?  More like -10 at 30k.

DC

[bobkatz]

dcollins wrote on Thu, 12 May 2005 22:54

bobkatz wrote on Thu, 12 May 2005 14:57 What I meant by the word "conservatively" was the choice of HF rolloff of the negative feedback in relation to the phase shift at audio frequencies. A lot of classic gear was struggling to be flat at 20kHz!  They didn't have enough gain before feedback to make it.  And the transformers phase-shift was such that they couldn't include them in the NFB loop! How did they ever make records like that? -3 at 300k?  More like -10 at 30k. DC

Absolutely. It was all part of the Neve "sound", along with transformers, aluminum electrolytic couplic capacitors and class-A discrete transistor modules.... But one of the reasons I think that in my circles we liked to bypass part of the old Neves to get a more "transparent-sounding" signal path was that SOME OF cumulative HF rolloff (or transformer coloration) became too much. If I want today to emulate the "old sound" I prefer to insert a piece of colored gear in the middle of as neutral a chain as possible, and I prefer to construct the rest of that chain with a 300 kHz bandwidth or so. By the time it goes through a few more parts of that chain it tends to be -3 at 100K or "worse". Wonder why that happens .

BK

[Zilla]

Brian Roth wrote on Tue, 10 May 2005 20:53

... *if* 5534s and TLo72s sound so ghastly, then what measurement technique can be used ... to support that? ...  every common testing method we have available must be seriously flawed. Comments?

Common testing methods are not flawed.  They are able to detect common physical electronic parameters for observation.  This informs engineers how to correct/improve circuit performance.  

The opinion as to whether an ic sounds ghastly or not is the sole domain of human culture and perception.  No scientific test is going to interpret how artful or musical an ic performs.  Our own ears are the quickest, most accurate test rig we could have for making this evaluation, and at no cost.  I do not think a "ghastlometer" will ever be invented.  But if it is, I will not trust a machine to tell me what musical expression ought to be.

[danlavry]

bobkatz wrote on Thu, 12 May 2005 22:57

danlavry wrote on Thu, 12 May 2005 17:12 You said: "but I tend to use opamps with a high GBP and run the neg feedback conservatively" That statement by itself can be a contradiction. Most Opamps with high GBP used in audio have a lot of open loop gain at audio frequency which means a lot of negative feedback. Most often, you choose a high GBP amp and you automatically end up with a lot of negative feedback. (I am not talking for or against feedback, which is another issue). Regards Dan Lavry What I meant by the word "conservatively" was the choice of HF rolloff of the negative feedback in relation to the phase shift at audio frequencies. As a rule of thumb I've been going for an F sub C of about 300 kHz or perhaps higher.

I really do not understand what you are saying. Fc of 300Khz? What kind of a spec is it for opamps?

Regards
Dan Lavry
www.lavryengineering.com