My result of opa2134 and opa627 comparison

I built an active XO for my new speakers with opa2134 and was not happy with the sound. It seemed that the dynamics and details of the music were gone with the active XO. That prompted me to investigate whether it was the design, implementation, or the opamps, or the caps that damaged the sound the most. I was determined to find out the sound of different opamps and bypass caps once more, this time, specifically, the difference between opa2134 and opa627. Although I have read various comments on this forum, getting first hand information would give me further assurance.

I previously reported that an upgrade from opa2134 to opa627 on the NAD542 gave some audible improvement but not "night and day" difference.

Often subjective listening tests really depend on many factors, including all the components in the entire audio chain, how the opamps are used, the input and output impedance, current demand, noise requirement, whether the rails are quiet enough, whether the opamps are used for the bass or the treble, etc.

This time the source was my upgraded Marantz SA11s1, a very accurate CD player.

The power amp was also very accurate with diminishing distortion figures under many test conditions. These were no caps in the signal path. There was no input cap.

I built a simple opamp unity gain buffer. Before the buffer there was only a volume control. After the buffer there was only a 200R series resistor to drive the cable.

The speaker drivers belonged to the very low distortion group. Tweeter was the top model from Dynaudio Esotar T330D. Midwoofer was the ScanSpeak Revelator 18W8531G00. The XO consisted of simple RCL for the tweeter and LCR for the midwoofer, using Solen Perfect Lay inductors and ICW SA Clarity caps. The drivers were mounted on an open baffle.

Only SACD, mostly DSD recordings, of classical music was used in the evaluation.

This was basically as good as I could have.

The power supply was of the LM317/337 type implemented strictly according to the manufacturer's datasheet. A pair of Rubycon ZL 2000uf was mounted on board with the buffer. A pair of Rubycon ZL 100uF bypass caps was mounted closer to the chip. Initially 0.01uF polystyrene bypass that was even closer to the chip.

The opa2124 was first mounted. Comparing to my 12B4 tube preamp, the 12B4 preamp sounded far more refined.

In other words, I would not be happy with the opa2134 sound at all if used as a buffer / preamp.

Since the opa627 datasheet recommends 0.1uF ceramic + 10uF Tant bypass, I added a 0.1uF polypropylene for bypass. I was still using the opa2134.

The sound was improved. It was obvious. But still, nowhere close to the 12B4.

I then replaced the opa2134 with opa627. WOW, what a difference! The music was back again.

Comparing to the 12B4, the opa627 had simular presentation. There was only a slight difference. As a matter of fact, if I did not listen carefully, I would not know which was which.

The very well executed 12B4 (choke input, shunt reg, etc) sounded slightly smoother and rounder on the top end and more "romantic" (looser) on the bottom end. The opa627 sounded slightly sharper (or harsher depending on how you look at it) on the top end but tighter on the bottom end. Overall, the 12B4 was slightly smoother but the opa627 was slightly detailed. This may well be due to the effect of the output cap (ICW SA Clarity) used in the 12B4 and absolutely no caps in the opamp buffer signal path.

I could not tell which was better. My 12B4 could still be improved slightly in which case it would be better. But then I could also try to bypass the rails for opa627 with a, say, 10uF MKP that may improve the sound?

This would be great news for those who want to build a 12B4 preamp - why be bothered! My 12B4 preamp was expensive and weights a tone. Just build the opa627 buffer! The 12B4 may still have a slight edge if the ultimate components and construction methods are used, but at the price of 20 times more costs for the improvement of less than 2% on sound?

I later added a 0.1uF MKP to bypass the two rails. i.e. a single cap connected the POS rail to the NEG rail. I have not had the time to make a detailed comparison yet but the result obviously did not improve the sound, at least not immediately audible.

So my conclusion was that on a lesser system, like the NAD542 I used to use (I still have it and still like it), there was improvement by replacing the opa2134 with opa627. But when the system is further improved and becomes more revealing, the opa627 beats the opa2134 by far. Also, I concluded that 0.1uF (and I will try larger values) MKP to bypass the rails are necessary. 0.01uF is not sufficient.

I will do further tests on this in the coming a few days, but mainly on the different types of MKP caps.

I am very interested in comparing the opa627 to the ad825 and the new LM????

Again, this time I ordered 4 pairs of opa627 from Chris who sells in the eBay (seller: ca2832 email: Web: The much improved sound simply indicates that the parts Chris supplied were genuine BB products. His prices are very good and he delivered them very promptly. I have no connection to Chris. I am just a happy customer.

Good test and very interesting. I'm a big fan of the LME49710, as it beats the 134 and 627 by a decent margin on most of the datasheet parameters. But, and it's a significant but, the LME is also a 55MHz GBP part. I suspect the people who have been unhappy with it haven't had sufficiently good layout and bypassing. It also needs isolation resistance to drive more than 100pF. You should try it- it's only a few dollars compared to things like the 627. I used to use the 627 in a scientific application, and it's a super part. It used to be cheaper and more available. It's really suited to electrometers and such, and I'd guess the high price is due to the low input current, much testing, and higher fallout than most op-amps. It has very low overshoot and actually slews faster than the LME part. It needs guard rings on the pcb to take advantage of its low input current, otherwise it's kind of wasted. I'd be very interested in how you think they compare- in my phono preamp I liked the LME better, but I haven't tried them in my active crossover yet.
I think the AD825 has a much higher spec and is much cheaper. It compared well to the opa134. I have not compared it to opa627 but I will, because I still need quite a few more opamps and opa627 is too expensive to my pocket. I have never tried the LM4562.

Also, I will try a shunt reg in a couple of weeks on the opa627 buffer and see if it makes any difference.

By the way, what are guard rings? where can I get them?
Hi Bill,

Have you thought about biasing the OPA627 to Class A? I have been a pest about this for about 9 months around here and I have MY answer on this (It can certainly be done).

If you need, I can tell you what I have learned.

I think you might be just as surprised as when you removed the 2134 and replaced it with the 627.

Guard rings are sold in the same stores that sell buckets of prop wash ;-) No, seriously, they're just very fine traces on the pcb that fully surround the input pins. Without those traces, surface leakage on the pcb will far exceed the input current of the part. For most audio circuits they probably don't matter, but in a high impedance circuit you don't get the performance the part is capable of without them. They might also matter with active filter circuits. I believe the OPA627 data sheet shows how to do them.

Thanks for the information. I used matrix type varo board. Does it matter for audio signals up to only 20kHz?


I heard of biaing into class A but never tried it. My class B (or AB) amp sounds as fine as class A amp, I suppose, having own a Musical Fidelity A1000 pure class A 60W amp for a couple of years and built a number of my current class B amps. Given that these tiny chips are mounted on a browndog heat dissipation may be a problem with class B, let alone with class A. I guess these chips are designed to be used in class B.

I am not surprised you may achieve better sound with class A bias but unless the heat dissipation is not an issue I would hesitate to do so. I guess you will tell me that you have never had a chip overheated (IC heatsink?).

I guess the trick is to add a few resistors. I am happy to give it a try and report back here. What is your fomula?

Hi Bill,

I hope you have been using Dip Sockets. Attached are 2 files of how to take 2 JFET Transistors to make a JFET Cascode. This is the best way to bias as it is the most consistant regardless of the variation in signal passing through the op amp. It employs two 2N5484 JFETs per channel.

Notice on one of the diagrams I said pin 4 -Vs and pin 7. Pin 7 is wrong because it is the +Vs so substitute that for pin 6 in the diagram. Pin 6 is the output and that is the correct pin to connect the D from Q2 to.

Buy 10 or 12 of the JFETs. They are cheap and I know at least Fairchild makes them.

Then check for Idss. You need two categories, High Idss and Low Idss. Then find two in the High Idss category that are a match and two in the Low Idss that are another match. When I say match, I mean as close as possible. These will be the 4 that you make the two Cascodes with. The two Low Idss ones will be Q1 and the two High Idss will be Q2.

As far as heat, everything I read said it is not an issue and shouldn't even need a Heatsink but you can certainly employ one if desired.

Build them and do one channel first to see if you like what you hear in that channel! If so, make it permanent in both channels!



  • class a bias jfet cascode pcb for op amps.pdf
    57.7 KB · Views: 496


2007-04-27 12:54 am
It has been my experience that despite what the manufacturers
may say about ceramics and tantalums used for bypassing,and this applies to the AD8066 as well, that they usually make the treble sound metallic, and give a hard edge in general. I find it far better SQ wise,to wherever possible , use voltage regulators ,e.g. 78L12 and 79L12, closer to the audio I.C. and use 100nF poly. bypass capacitors at the I.C. The typical
47-100uF at the output of the close by VR is normally fine. I guess that the manufacturers specify those particular components to guarantee stability?
The OPA2134 is not a bad sounding I.C. , it is the way it is implemented. The LM4562 can also sound excessively detailed, and fatigueing in some arrangements.


Flame suit is on !

Thanks for sharing the information. I will keep what you posted and do some experiment when time allows in the future. The only difficulty I can see is where to buy the matched parts of the JFET.


I also thought of using descrete circuits but don't like the complexity. While one can achieve something much better in one way he can also lose in another way. For example, what about noise rejection?


Where did you get the LM4562 from? Given that you are in Sydney, have you got one on the browndog for me to loan? I need a dual one for the existing test circuit. I have just read the datasheet and could not find any notes on the sizes of the LM4562? single? dual? If you do have one spare, I am happy to buy or loan it from you and pick it up from your place, or you could come over to my place and have a listen to your LM4562 in comparison to the opa627 in my set up.

Hi Bill,

Maybe try here or any other distributor in Australia.

They are under $0.14 US here at and no minimum charge so buy 25 for $2.88 (you get a break at 25) and check the Idss with a DMM (milliVolt meter), 9-volt battery and a 1K resistor yourself to match pairs. So you will have a few left over. After hearing the impact, you might continue doing more and some recommending too!

If you need, I can explain how to set up the test! (beginner stuff, that's why I know now). LOL. Actually, I have to thank AndrewT for the tutorial!

Cap Test results

Last night, I used the opa627 buffer and tested a number of caps. The caps were connected as output caps in the buffer. The power amp had an input impedance of 10k. All caps had a value of 0.1uF. That means the rolloff was very early but it was of no concern because the purpose was to test how much the caps were damaging the sound in the midrange and high frequencies. Low frequencies were of no concern.

The value of 0.1uF was chosen because it is the most commonly used value. I mainly use 0.1uF in my filters.

1. Vishay MKP1837, polypropylene - highly recommended as bypass caps in passive XO by Tony Gee, Humble Homemade HiFi.
2. BC Component MKP, polypropylene - have not taken note of the series number. Smaller voltage. Fairly squared.
3. BC Component MKP, polypropylene - have not taken note of the series number. Slightly larger voltage. Flater appearance.
4. Wima MKS4 - Polyester

I was planning to also test the following for a comparison:

5. ICW SA Clarity MKP - I used them a lot. I have only 0.22uF only, the lowest value I could get.
6. Mylar - the cheap green cap.
7. Falco - a big round thing I got from an amplifier built by somebody.
8. Solen MKP

However, I got distracted by something else so did not have the time to do No 5-8. I will possibly do them tonight.

The result:

Overall they are very close to each other. They are much better than I expected. I remember a couple of years ago I did compare MKPs of the same value on a passive XO and found huge difference between them. The Jaycar MKP / polypropylene sounded absolutely awful. The Solen MKP, in comparison, was better. But when ICW SA Clarity caps were compared, the Solen became harsh. So since then I used only SA caps in passive XO. I have not tried anything else.

In an active circuit, the situation could be quite different. I remember I once compared the SA caps with the ordinary Mylar caps in a basic tube preamp with the caps being used as output caps, and I could not hear any difference.

This time, I found that those caps veiled the sound slightly, but were certainly not detrimental to the sound. My conclusion was that, lesser opamps, even like the good opa2134, when comparing to better opamps, like opa627, make a large difference. Good capacitors used in an active circuit, like those ones tested, make far less difference comparing to opamps.

They are all quite acceptable. It is highly likely that if in a blind test I would not be able to tell the differences. They are more common than different. If I have to exaggerate their differences based on my highly unreliable impressions, I would say the MKP1837, the most expensive one from Farnell, sounds possibly the smoothest but a bit dull. I guess it does not do very well comparing to others in high frequencies. It lacks of the "sparkles" others have. No 2 is possibly what I would choose in my active circuit, and is an overall balanced performer. No 3 and No 4 sound the same, which is to my surprise. No 4 is polyester only, not polypropylene. Both are very smooth without any nastiness, brightness or hashness in the sound. They are perhaps lack of the finest details or resolution that No 2 would give. But No 2 does not appear to be the smoothest sounding cap. Basically, I don't trust my hearing that much so who knows on different machines using different music what the results would be?

The good news for me is that I should not be worried about caps that much in the active circuit.

I think the next stage for me is to compare the opa627 to LM4562 and AD825. I will do that once I receive them, possibly in a weeks time.

I would like to ask a question and get a reply from you experts.

In the OPA2134 datasheet:

For lowest distortion with a source or feedback network which has an impedance greater than 2kOhm, the impedance seen by the positive and negative inputs in noninverting applications should be matched. The p-channel JFETs in the FET input stage exhibit a varying input capacitance with applied common-mode input voltage, ..., and the gate-to-source voltage is not constant. The effect is increased distortion due to the varying capacitance...

Does it apply to using OPA2134 as an unity gain buffer? Does it apply to other JFET opamps, such as opa627, ad825?

For some filter networks, such as dipole EQ / shelving low pass, if lower than 2k impedance is chosen the capacitor in the feedback network needs to be very large that is not practical. Also the impedance vary with frequencies.
Re: Cap Test results

HiFiNutNut said:
I would say the MKP1837, the most expensive one from Farnell, sounds possibly the smoothest but a bit dull.

I think the next stage for me is to compare the opa627 to LM4562 and AD825. I will do that once I receive them, possibly in a weeks time.


Go AD825 go. ;)

I never had much luck with the MKP1837 as a by-pass in speakers and seemed to take some life from the music.
If you're on copper square Vector board, there's significant capacitance between squares. I've certainly used a ton of it, but it's not the stuff use high bandwidth parts on, then make comparisons of sound quality that apply to other situations. IMO, the OPA627 comes out on top of most other parts not only because it's inherently very good, but because it occupies a spot near the limit of casual construction techniques. Linear Technology's application note on High Speed Amplifier Techniques (AN-47 ?) would be an excellent read for many here. Note the breadboard techniques used. This stuff isn't just theory- it has a huge effect on overshoot and settling.

Thanks for the info. I will find that note and have a read.

Yes I am using copper square Vector boards. I can understand the issue of capacitance but thought within audio bandwidth it does not really matter as it may affect only very high frequencies that are above at least 15kHz. But given that opamps have Mohm of input impedance I can be wrong. I don't have any equipment to measure it.

I guess I will first read that article.

Last night I added two new opamps to my comparison tests:


I ran the tests in this order:


First, thanks to SandyK who kindly offered me a dual LM4562. I asked for paying for it but SandyK insisted giving it to me free so that I can publish the result here for other members' benefits, and thanks to Chris, who supplied 4 pairs of opa627 to me and sent me 4 evaluation samples of AD825 from Tasmania.

I had been running the OPA627 for days. That was my reference opamp upon which others were judged.

I first played an hour music with the OPA627. I mainly used two pieces of music that I was overly familiar with: "Beethoven: Complete Sonatas & Variations - Wispelwey/Lazic" in SACD from Channel Classics. This is the most gorgeous cello and piano SACD I have found, even better than many from Sony, Decca, DG and Phillips SACDs. It has first class performance and is a first class recording (slightly close-mike), and the "Grieg: Choral Music - Grex Vocalis" in SACD from 2L. I believe that the sound quality of this 2L recording matches the recordings from isoMike, but with better singing / performance in a more natural acoustic environment in a church instead of in a recording studio.

I have to say that I have a bizarre feeling when I talk about comparing opamps. All of these opamps have a distortion figure that is so low that is supposed to be well below human perception. I don't think I have a pair of ears much different from most of my own species. So comparing opamps judging by ears sounds very silly to me, but, a big BUT, I found opamps sound so different in a high resolution system, and replacing an opamp on my buffer preamp has the same sound effect as changing a CD player, preamp or power amp. Although the result is highly subjective, I found I have had very consistent results. So here it goes. Bearing in mind that I will have exaggerated their differences than stated their similarities.


When I replaced the OPA627 with the OPA2134 I found the music to be a bit uncomfortable to listen to. High resolution was lost. Details were lost. The sound came out "thick" and I could feel the artificially enlarged sound stage. It was less "real", less "live" and less "musical" to my ears. Don't get me wrong. The OPA2134 is not harsh. Comparing to the opa627 it is muddy but not harsh. It is actually the warmest opamp among this group. It is the thing that is closest to the "tube" sound. I feel the OPA2134 has the highest 2nd order harmonic distortions that I can not live with. Price wise, the opa627 is much more expensive than the opa2134.


On first listen, just like what I experienced when I replaced the opa134 with the ad825 on a DAC more than a year ago, I had smile on my face. The high level of distortion of the OPA2134 was gone. I found the sound to be so much cleaner, clearer and faster than the OPA2134, a similar feeling I got from the OPA627. But soon I realised that the cello and the piano sound, comparing to the OPA627, is thinner and sharper. When changed to the Choral music the devil in the AD825 was revealed - the AD825 sounded "hard" on voices in comparison to opa627. I did not think I could live with the AD825. My guess is that the AD825 has a substantially lower harmonic distortions (HD) than the OPA2134, but the HD in the OPA2134 is mainly even order, and in the AD825, odd order. The price of a AD825 is less than half of that of the OPA627.

LM4562 and OPA627:

Wow! The LM4562 sounded nice! Comparing to the OPA2134 and AD825, the LM4562 is a lot closer to the OPA627. But it was different in some ways I could not immediately describe. So I played the LM4562 for a lot longer trying many other CDs I was familiar with. Even after one hour playing, I still could not conclude if the LM4562 was better than the OPA627 or not, though the difference was obvious.

Note that I had lived with the opa627 for some days so I would unconsciously use the opa627 as the reference.

The LM4562 provides more details than opa627. It has higher resolution. The sound is richer. It is fast in speed and very dynamic. So overall, the LM4562 gives a more "live" feeling to the music, comparing to the opa627. My love of the LM4562 was growing when listening to this chip. I could discover some new stuff in the music with new sensations that I had not previously obtained from the opa627.

But the opa627 does not seem to do anything wrong except that it does not have the details and resolution as the LM4562, though it still sounds very detailed with high resolution. The opa627 does not appear to have its own sound, though it might have deleted some sound, unlike the OPA2134 and AD825 that may add harmonics to the sound.

Comparing the two, to the slightest extend, the opa627 sounds smoother, softer, lighter, calmer. The LM4562 sounds more detailed, live, colourful, and eventful. On musical instruments and even voices, the LM4562 sounds, to the slightest extend, more real and closer to live. The opa627 sounds sweeter, delightful on human voices.

I could not draw a conclusion which one I preferred. I will need to do more tests and play the LM4562 and OPA627 for a few more days before I can possibly come to a conclusion. Or perhaps which one is better is subject to personal preference.

I suspect that the LM4562 may be better in many ways and I may end up with using LM4562 rather than the opa627. But there might be a chance that on long listening the LM4562 might be tiring, as I know very well that I can listen to the opa627 for hours and days without any fatigue whatsoever.

I don't think I will publish the end result because it can be meaningless. If I could not tell which one I liked better in a two hour contest, they must be at pretty much the same level of performance. I will prefer to leave it to you to conclude. This kinds of tests are highly subjective, and the results depend on too many other factors to be relieable for someone else's use. I am posting my result for your reference only, not as a guideline.