[jeffreyj]

Testing was conducted once again on some of the higher-performance op-amps I have laying around the shop plus a bunch of the favorites around here that Tangent loaned me. The test was eye-opening, to say the least, especially when the brutal simulated Grado's load was connected (33 ohms). Folks, it ain't pretty, is all I can say.

The test signal was a 2kHz sine wave from my B&K Precision function generator. It isn't the cleanest signal source I have, but it's definitely the most convenient. The input level was raised until the op-amps were clearly clipping, then backed off until the clipping just disappeared. The power was from an adjustable regulated supply set to 9.1V. The circuit was two inverting gain stages in series with the feedback resistors shunted by 2.2pF NPO ceramics to ensure no funny stuff ocurred. The two loads tested were 100 ohms and 33 ohms.

Without further ado, the link to the table of results follows:

* - Not a typo!

The Analog Devices offerings are starting to look a little long in the tooth. I don't know how some of these op-amps became favorites around here, but it might be time to reconsider! The AD843, in particular, turned in such a terrible performance that I'm hoping that somehow all 4 separate op-amps suffered the exact same static electricity damage or something (yeah, right...).

edit: updated test results table with a couple more op-amps; changed table format.

 

[tangent]

If low output-to-rail limitation and low supply current is a requirement, the AD843 is indeed a very poor choice. But when supply current doesn't matter and you can raise the supply voltage high enough to get the output voltage you require, the AD843 sounds wonderful and it's cheaper than the OPA627/637.

 

[jeffreyj]

Quote:

Originally posted by tangent If low output-to-rail limitation and low supply current is a requirement, the AD843 is indeed a very poor choice. But when supply current doesn't matter and you can raise the supply voltage high enough to get the output voltage you require, the AD843 sounds wonderful and it's cheaper than the OPA627/637.

You are absolutely right. I was very skeptical, but after testing the AD843 at twice the supply voltage (18.2V split into +/- 9.1V) I have to concur: the AD843 performs quite well, and distorts in a very different way from all of the other op-amps I tested, which may account for it being preferred by some folks. I was not inclined to re-test every op-amp at the higher supply voltage, but I did do three of them (the LM6172 and NE5532, additionally) and found that the AD843 did deliver the highest output into a 33R load. However, the LM6172 and NE5532 still beat it driving a 100R load (9.84, 10.16, and 10.72Vpp, respectively).

Here is a picture of the distortion the AD843 generates just prior to true clipping of the output:

 

[jeffreyj]

More op-amps will be tested soon: I just found a whole bagful of AD826AN's at a favorite nearby surplus store. Also, I bought some TL072's because I haven't personally used them in at least 10 years, but they used to be favorites for audio applications.

The AD826AN's are supposed to be cable drivers; perhaps in the same class as the LMH6655/72's from National. We'll see

 

[tangent]

That distortion characteristic you're picturing is actually quite unpleasant to listen to. If you examine the datasheet, you'll see that it informs you of this asymmetric clipping behavior in the main data table. For audio use, it's best to just take the worst number and double it; so at +/-15V, the output to rail limitation would be +/-3.5V. Down at +/-5V, the limit is apparently +/-2V, so you should be able to get 2Vrms out of the chip before it starts distorting, but I wouldn't count on even that. I never run this chip at less than 12V myself.

Interesting scope, there jeffreyj. It looks like a contemporary of the TDS320 I've borrowed from work a time or two.

 

[MegaMeee]

what about testing an njm4556 opamp? ie ra-1 opamp?

 

[ppl]

jeffreyj> Where is Pt One?

Most opamps do not like loads below 5K and alot of em will start to go into current limit at 1K @ Full output. To ask such a device to drive 100 & 33 Ohm loads is silly! this is what buffers are intended to do.

The AD-843 is a strange opamp indeed. it likes to lock up to one of the supply rails if driven into Clipping. or if powered up slowly. it;s output Voltage will swing to the rails at about 2 volts less than Most every other device. Whay one would use these might seam like a mistery However as tangent pointed out it sounds real nice and Great bass. The AD-843 sounds quite nasty at or near clipping so i avoid this. It is also a good idea to avoid clipping any Amp.

Thanks for taking the Time to do this test jefferyj the cmoy amp builders i am sure are in your debt

 

[jeffreyj]

Quote:

Originally posted by ppl jeffreyj> Where is Pt One?

In the manufacturer's datasheets!

Seriously, though, I wasn't interested in how well they performed into an easy load because, quite frankly, with 70-100dB* of negative feedback crammed down their throats I expected them to all perform pretty much the same. (* - depending on their respective Avol, of course).

Quote:

Most opamps do not like loads below 5K and alot of em will start to go into current limit at 1K @ Full output. To ask such a device to drive 100 & 33 Ohm loads is silly! this is what buffers are intended to do.

This is the conventional wisdom, and it is true 99% of the time, but I was interested in that other 1% for a commercial design I was working on that happens to have much in common with the aims of most Head-Fi'ers. Specifically, I wanted to find out which op-amp could deliver the highest volume without clipping into a "typical" headphone load while running off a single 9V battery. Knowing that, I could then make a more intelligent design choice based on sound quality, price, etc.. Given that dedicated buffer chips typically cost more than most op-amps, and that discrete buffer designs are quite cheap, but demand much more board space to implement, it behooved me to at least attempt to find an op-amp that could deliver the volume desired without needing extraneous circuitry.

After mining through some of National's offerings, I became intrigued by the capabilities of their op-amps designed to drive cable loads; specifically, the '6655 and '6672. Frankly, their performance here says it all: the '72, in particular, seemed virtually unfazed by the 33R load, delivering 5.2Vpp without clipping from a measly 9V supply. Granted, a full-on shakedown with the same loading but RMAA 5.1 calling the shots would give a more complete objective measure of its performance, but this simple load-testing at least gives a simple "go/no-go"


tangent: the AD843 may have it's merits elsewhere, but not w/r/t driving heavy loads! The hump on the positive excursion seems to be 3rd harmonic, while the sharpening of the negative excursion looks for all the world like the beginnings of slew-limiting?! Not likely, I know, but that's what it looks like. My eyeball Fourier analysis is not accurate enough to even guess what exactly is going on below the zero crossing

The scope was the single most agonizing purchase I made, but has proven it's worth since then. There are two things I love most about it: it is truly portable and it is fully isolated from the AC line even when powered by it. This makes working on switchmode power supplies, AC motor drives, etc., much more practical; nay, possible, really, as I have never really been inclined to lug around isolation transformers.

MegaMeee: The next time I order parts from Mouser I'll pick up a couple of the '4556s as I am also curious about their "reknowned" load driving capabilities.

 

[tangent]

Quote:

the sharpening of the negative excursion looks for all the world like the beginnings of slew-limiting?!

No, it's that B+K 4011 you're using: it uses triangle wave reshaping to obtain its "sine wave" output, so it goes to peaks like you're seeing. Distortion is nearly 2% in the best case with my 4017A. Bleh.

 

[jeffreyj]

Quote:

Originally posted by tangent No, it's that B+K 4011 you're using: it uses triangle wave reshaping to obtain its "sine wave" output, so it goes to peaks like you're seeing. Distortion is nearly 2% in the best case with my 4017A. Bleh.

I'm not so sure this is the case here, but it is certainly a possible culprit. I say "not so sure" because no other op-amp sharpened the negative excursion prior to clipping like the AD843 did. Also, the sine wave looked pretty good on the scope on it's own, which is another reason I am skeptical the 4011's tracking filter is the culprit.

Granted, though, that this is was all done using my "eyeballs" to perform the FFT, so YMMV!

 

[jamont]

Quote:

Originally posted by jeffreyj Granted, though, that this is was all done using my "eyeballs" to perform the FFT, so YMMV!

You mean you don't have a spectrun analyzer?

Seriously, thanks for making these measurements!

 

[ppl]

if pure Power is needed in an opamp of reasonable Perfiormance is required check out the LT-1210 @ 1.2.Amps of output current. This is a CFB type of Op amp and somthing like a Jfet folower hanging on it's input would allow this chip to cope with a volume Pot without inducing too much DC offset.

 

[jeffreyj]

Quote:

Originally posted by jamont You mean you don't have a spectrum analyzer? Seriously, thanks for making these measurements!

I did own one several years ago but sold it because it was so expensive yet rarely used. It came in handy for doing RF stuff, but the audio frequency range definitely seemed to be an afterthought! A software-based FFT program I have, WinAudioMLS is far superior at the audio stuff as long as a good soundcard is used (I have a M-Audio Audiophile USB connected to my laptop to make a semi-portable spectrum analyzer).

Oh, as a more general announcement, check the results table again as I added a couple more op-amps and the current prices/suppliers of each. That information really puts things into perspective!


ppl: LT definitelys makes some great products and I'll be looking at a couple of their video-optimized op-amps in the near future. When a circuit needs output currents approaching 1A, though, I completely abandon the thought of monolithic buffer ICs and go discrete because there just isn't any comparison from a performance per dollar perspective (if not a performace per area of pc board perspective!).

 

[ppl]

Quote:

ppl: LT definitelys makes some great products and I'll be looking at a couple of their video-optimized op-amps in the near future. When a circuit needs output currents approaching 1A, though, I completely abandon the thought of monolithic buffer ICs and go discrete because there just isn't any comparison from a performance per dollar perspective (if not a performace per area of pc board perspective!). [/B]

I agree with this as the last real good ! amp Buffer the EL-2008/2009 are discontinued and yes these sounded as good as a well designed Discreet component Buffer and alot better than a porly designed one. I al so do discreet design in fact i like this method because i can have complete control over each cell of the Circuit and this makes it aalot easyer to impliment my Design Ideas and most importently my Perfered compensation Methods. By controlling the Real Open loop gain i can optomize the circuit so as to avoid Frequency compensation. I understand and can respect your reasons for using Compensation and alt of times light compensation is required but as a rule i tend to avoid it.
some interesting links
http://peufeu.free.fr/audio
The Lavaradim Circuit
http://www.linear-tech.com/pub/docum...pp&document=50
AN-47 High Speed Amp technics from Linear tec.
This covers almost all aspects of High speed design and shows examples of Discreet component buffers and the Advantages of both IC and Discreet design. (Pg 46) Starting on Pg 84 is a good discussion on Frrequency compensation. jeffreyj should like this note as most of the Circuits like most all other App notes advcate compensation capacitors. however if the Complete note is read word for word every page certen things come to light such as Pg 88 Fig. C3 " When the op- Amp is much slower than the Booster stage compensation is free" on Pg 96 is a low capacitence High Z scope probe made from an EL2005 FET input buffer. I use this same thing as a probe on sensitive circuits. Pg 128 showes the layout i use around op amps in my own Board layouts. This is a good read

over at http://diyaudio.com/forums/showthrea...ler+compensate is a thred showing mine as well as others view on compensation. you can see an equal amount support both mine as well as jefferyj view. on this matter. No I am not opening up pandoras box just mearly pointing out data that hopefully some will find of interest and dod not concern with if it supports my view or not.