Why do you keep them always on ? Input coupling caps are there for outputs that don't have an output cap , or have a DC voltage offset and to protect the source.
One of the reasons I want to cool the LME 49600 with a large aluminium block , is maybe to use it with little speakers . Doesn't have to be loud.
Ah , Elvee came through for me like many times before with some direct answers to my burning questions .
Well my supply lines should be very clean , they come from something you may have heard of : Elvee's DeNoiser. And my inverting inputs are fairly low impedance with NFB R's of 750 to 1,15K.
So you're saying no problem here. 👍
What do you call significant distance ? 3 mm , 1 cm ? With peak currents of maybe 100 to 150mA (into 24ohm) .
The + supply lines to both opamps must be elevated , no choice there . Hard to avoid that complication. If it was just digital logic I'd route the tracks under the feedback R's and done.
Good to know about the relay coils , no problem being so near the LME . Their supply voltage is "clean" .
So via's of 10nH for decoupling will not be a problem in this case.
I'm not sure if the T.I pdf means hollow via's , which leaves a hole you can see through and filled via's , like with a full through wire .
Is there a difference ?
Thanks Elvee .
It depends on the length for which they are parallel: if it is of the same order as the separation, it will begin to have a small effect
The difference is only for DC resistance. At HF, there is no difference between a tube and a cylinder
So I'm pretty much on the right track with the layout in post 44 .
No parallel supply tracks with sensitive signal lines and via's near decoupling caps.
If the circuit is in an aluminium casing , do shielding groundplanes even matter with only some passive cicuits nearby ?
No parallel supply tracks with sensitive signal lines and via's near decoupling caps.
If the circuit is in an aluminium casing , do shielding groundplanes even matter with only some passive cicuits nearby ?
Here's the thing: IIRC according to Scott Wurcer, designer of the AD797 opamp and former VP of engineering at ADI, no opamp has ever been designed for audio to his knowledge. Most of the latest, greatest high performance opamps used for audio today were developed for ADSL, something that for the most part didn't happen. The marketing department had to figure out how to sell the parts so they were promoted for audio use. Okay, if I recalled any of that wrong maybe someone will step in and correct me.
If no opamps were designed for audio, and if the marketing department at the IC manufacturer has a lot of say over what goes in datasheet, and maybe what not to say in there, then why should high speed buffers be any different from opamps? Bottom line in my book, IC manufacturers are trying to figure out how to market what they make. They will tell you why you should buy their products, not so much why you shouldn't.
The other thing about opamps, buffers, and discrete HPAs is something I hate to keep harping on. I know some forum members have heard it enough already. That has to do with intermodulation of audio signals and noise that I suspect can happen in high gain feedback loops. Some people have been complaining about the sound of opamps and global feedback for a long time. Mostly they have been dismissed as imagining things that aren't real by engineers who can't imagine how standard AP measurements might be able to miss anything. The theoretical arguments seem to make sense, at least they make sense until noise modulation (signal correlated noise) is considered as a possible explanation. It also may help if one can accept for the moment that some people easily notice such intermodulation effects in sound, and other people don't. I was until fairly recently someone who did not notice a difference because I never bothered to listen to a good discrete HPA and compare it directly to two different models of SOA IC-based HPAs.
Here is my impression from listening: If someone never heard a first class discrete HPA they would never have any way of knowing how bad IC designs are. What's wrong with the latter is not HD and its not baseline noise. As it happens those are the two things we spend a lot of time focused on measuring. From listening you would never know anything is wrong until an A/B comparison is done. One reason could be that someone might listen for problems like power supply hum, hissing-type noise, distortion, incorrect frequency response, etc. Those are all things you have been taught are real and you believe they are real. Many of you have been told by highly confident engineers that noise modulation in audio circuits is not real. Those guys don't actually know. They are simply overconfident they already know everything there is to know. Some of them are so confident in their belief that they have to make up conspiracy theories to explain why other people disagree. They most common theory is that people who disagree are being paid off to lie. Same type of stories people make up and then believe in the subjects that are banned from discussion in this forum. IMHO that's why the subjects are banned. People can't stop arguing and accusing each other, and then things escalate from there.
Anyway, I only suggest to build some HPAs as diy projects. Sure, build an opamp and buffer HPA. Build a Pass HPA-1 too, and then compare. Why HPA-1? Because it is probably the best one with a schematic now in the public domain, not because someone couldn't design better. That's why I also suggest experimenting to see if anyone can figure out how to improve it. If someone can and they want to share it, they the diy audio community could benefit from the know how. Personally speaking, I can only give general hints. That's a limitation I accepted and will abide by.
Last edited:
^ Well , if we're going there ...
I am skeptical about things like what T.I calls : SoundPlus TM opamps and what I
think is binning , like CPU manufacturers do . Same opamp but slighly different
properties like OPA2156 / OPA1656 . At least they stopped with A and B versions,
like OPA627A & B. The OPA828 has only 1 version .
Since ADSL is something from the past , the newest opamps are not build for that , but any precision opamp
with good enough numbers ( noise , speed, THD,...) can be used as an HD audio opamp. 5 V 24 bit = 0,3uV steps !
Not all can be so precise.
There is a lot of talk here about which opamp or amp sounds best or better than the other.
Here is my (amateur) opinion about that :
Many opamps have nearly all around the same very low THD and noise.
AFAIK you sweep a sine wave on the input and compare that with the output over
the audio frequency range. Some give a 3 tone comparison and
intermodulation distortion .
A good baseline comparison , but too artificial.
They should measure distortion with complex signals , like music . I think opamps
differ in how they handle such signals composed of a very high number of frequencies , not a clinical sine wave.
But it is ofcourse not practical . What would be a reference music piece to compare all with ? A piece of Bach/Mozart ?
Some opamps will distort/change the multitude of frequencies in music more than
others and that is what we can hear as in this one sounds beter than that one.
Like I said before , choosing an opamp is a crapshoot . For me choosing a
reasonably priced precision opamp like OPA828 , gives me a better chance of
having a good sounding HPamp , than someone on a forum saying this or that
discrete amp sounds really good.
Building with discretes , means more care in
choosing and selecting because of large differences/tolerances between the same transistors .
Again this is my amateur opinion why some opamp / amps can sound better to someones ears than another.
I am skeptical about things like what T.I calls : SoundPlus TM opamps and what I
think is binning , like CPU manufacturers do . Same opamp but slighly different
properties like OPA2156 / OPA1656 . At least they stopped with A and B versions,
like OPA627A & B. The OPA828 has only 1 version .
Since ADSL is something from the past , the newest opamps are not build for that , but any precision opamp
with good enough numbers ( noise , speed, THD,...) can be used as an HD audio opamp. 5 V 24 bit = 0,3uV steps !
Not all can be so precise.
There is a lot of talk here about which opamp or amp sounds best or better than the other.
Here is my (amateur) opinion about that :
Many opamps have nearly all around the same very low THD and noise.
AFAIK you sweep a sine wave on the input and compare that with the output over
the audio frequency range. Some give a 3 tone comparison and
intermodulation distortion .
A good baseline comparison , but too artificial.
They should measure distortion with complex signals , like music . I think opamps
differ in how they handle such signals composed of a very high number of frequencies , not a clinical sine wave.
But it is ofcourse not practical . What would be a reference music piece to compare all with ? A piece of Bach/Mozart ?
Some opamps will distort/change the multitude of frequencies in music more than
others and that is what we can hear as in this one sounds beter than that one.
Like I said before , choosing an opamp is a crapshoot . For me choosing a
reasonably priced precision opamp like OPA828 , gives me a better chance of
having a good sounding HPamp , than someone on a forum saying this or that
discrete amp sounds really good.
Building with discretes , means more care in
choosing and selecting because of large differences/tolerances between the same transistors .
Again this is my amateur opinion why some opamp / amps can sound better to someones ears than another.
First off, I highly doubt most OpAmps used in audio were made primarily for ADSL applications. Many of the high-speed Line Driver ICs used to drive headphones these days were made for ADSL or powerline communication however. That is for example the omnipresent TPA6120/THS6012.
Secondly, OPA2156/1656 is not just "binned" like CPUs are and if you read the datasheets carefully there are obvious differences. OPA2156 is laser trimmed for lower input offset voltages. It is focussed towards instrumentation, so DC precision matters more than it does for audio applications. It also has a rail-to-rail input stage that utilises slew-boosting. OPA1656 doesn't have the rail-to-rail IPS with slew-boosting but has an additional pole-zero pair for higher loop gain in the audio band.
I will let someone else continue this list if they feel like it. I have no interest to engage in the usual debates about feedback or discrete vs IC-based design approaches.
Secondly, OPA2156/1656 is not just "binned" like CPUs are and if you read the datasheets carefully there are obvious differences. OPA2156 is laser trimmed for lower input offset voltages. It is focussed towards instrumentation, so DC precision matters more than it does for audio applications. It also has a rail-to-rail input stage that utilises slew-boosting. OPA1656 doesn't have the rail-to-rail IPS with slew-boosting but has an additional pole-zero pair for higher loop gain in the audio band.
I will let someone else continue this list if they feel like it. I have no interest to engage in the usual debates about feedback or discrete vs IC-based design approaches.
The diyAudio Store sells the WHAMMY HPA which is a composite amplifier, consisting of an opamp plus a discrete, high bias, class-A buffer stage, all within a single global NFB loop. People love its sonics.
The Store also sells the ACP+ HPA which is an all-discrete circuit containing 3 JFETs, 2 N-MOSFETs, and a 4N35 optoisolator for bias stabilization. It has two amplifying stages (output stage is high bias Class A), and a single global NFB loop. People love its sonics.
The Store also sells the Noir HPA which is an all discrete circuit containing 1 BJT and 1 N-MOSFET. These are arranged as two cascaded amplifying stages (both are high bias & Class A) within a single global NFB loop. People love its sonics.
And of course there are zillions more HPAs available from other places besides the Store. Lots of them are owned by very very happy audiophiles who love the sonics.
There seem to be several quite different approaches to sonic bliss. Which is good news for builders, experimenters, and enthusiasts.
The Store also sells the ACP+ HPA which is an all-discrete circuit containing 3 JFETs, 2 N-MOSFETs, and a 4N35 optoisolator for bias stabilization. It has two amplifying stages (output stage is high bias Class A), and a single global NFB loop. People love its sonics.
The Store also sells the Noir HPA which is an all discrete circuit containing 1 BJT and 1 N-MOSFET. These are arranged as two cascaded amplifying stages (both are high bias & Class A) within a single global NFB loop. People love its sonics.
And of course there are zillions more HPAs available from other places besides the Store. Lots of them are owned by very very happy audiophiles who love the sonics.
There seem to be several quite different approaches to sonic bliss. Which is good news for builders, experimenters, and enthusiasts.
I congratulate you to good choices: OPA828 + LME49600.
I like this myself. This is why I have made some try to make such circuit.
I think you should go for one OPA828 and LME49600.
^ Like I said , it is a gamble choosing and opamp , but if you have tried this combination , for real not SPICE , share your design and experience .
Yes I get some negative comments on using 2 opamps , that's ok . If it had no gain or lower , sure . And yes an opamps doesn't really see how much it is amplifying , whether that be 2 x or 10 , it is just the same at the inv & non-inv input . Still I do it in 2 steps and sticking by it . Besides I'm too far in making the PCB to change it , even if I wanted , which I don't.
Yes I get some negative comments on using 2 opamps , that's ok . If it had no gain or lower , sure . And yes an opamps doesn't really see how much it is amplifying , whether that be 2 x or 10 , it is just the same at the inv & non-inv input . Still I do it in 2 steps and sticking by it . Besides I'm too far in making the PCB to change it , even if I wanted , which I don't.
Semiconductor manufacturers are quick to point out the virtues of composite amplifiers. There is a natural division of labor: the output stage (current booster LME49600 / BUF634A / LH0033 / LT1010 / discrete_buffer) squirts giant amounts of current into the load, while the opamp provides low offset voltage, high input impedance, giant amounts of negative feedback which reduces distortion, and easy configuration of overall gain. Since the opamp isn't required to drive the load, its output current is tiny, thus it doesn't heat itself up and ruin the near-perfect balance of the input pair. Read Robert Widlar's application note about the LT1010 (document AN16) -- it waxes eloquently about the virtues of separating the opamp from the load-driver.
squirts giant amounts of current 🙂
Yes . An interesting read , not easy for an amateur . AN16
https://www.analog.com/media/en/technical-documentation/application-notes/an16f.pdfEven quality opamps struggle a bit with overshoot or ringing as non-inverting buffers with low gain.
When putting a buffer like the LME behind it , the opamps only sees the input and the stray cap in the PCB , but the AN16 suggests that it still needs compensation with high cap loads , like the long wire to the HP.
Then again , isn't overshoot and ringing the worst with square waves , which are a lot less steep in a audio signal with a bandwidth of less than 100k for 24bit HD tracks ?
Yes . An interesting read , not easy for an amateur . AN16
https://www.analog.com/media/en/technical-documentation/application-notes/an16f.pdfEven quality opamps struggle a bit with overshoot or ringing as non-inverting buffers with low gain.
When putting a buffer like the LME behind it , the opamps only sees the input and the stray cap in the PCB , but the AN16 suggests that it still needs compensation with high cap loads , like the long wire to the HP.
Then again , isn't overshoot and ringing the worst with square waves , which are a lot less steep in a audio signal with a bandwidth of less than 100k for 24bit HD tracks ?
Overshoot and ringing are defined for square waves since it takes a fast risetime signal to produce such artifacts. What they mean in the frequency domain is that there is gain peaking at high frequencies, often well above the audio band. While at first that my not seem of any significance, given that we now live in a world full of household RFI/EMI, it means that antennas such as long speaker leads can bring RF incursion into an amplifier. Once that happens, and if the amplifier is as good or better at amplifying those frequencies than it is at amplifying audio frequencies...Well, maybe opening the door more widely than necessary to chances of audio becoming intermodulated with RF noise. Good luck seeing if any other people agree on that since almost no one seems to bother to measure for it 🙂
Last edited:
I’m on the same course, designing circuit with OPA828 and LME49600. It will be input buffer for amplifiers from Pass Labs forum section, that employ transformer as VAS (i.e. LuDEF & SissySIT), and which will replace original buffer that drives transformer.
Input opamp choice was not hard as I had OPA828 in component drawers and some experience with them. Bought some OPA1641 as well. Between BUF643A and LME49600, the latter seemed as better suited for intended purpose and easier to solder.
Now, I’m trying to determine whether there would be any benefit from using super regulators as supply for this buffer. OPA 828 has high PSRR. LME49600 not so great, but as it will be inside NFB loop of OPA828 its PSRR should be improved.
As is, I’m leaning on conclusion that super regulators would bring small or no benefit at all. Any experience or opinions are welcome.
Input opamp choice was not hard as I had OPA828 in component drawers and some experience with them. Bought some OPA1641 as well. Between BUF643A and LME49600, the latter seemed as better suited for intended purpose and easier to solder.
Now, I’m trying to determine whether there would be any benefit from using super regulators as supply for this buffer. OPA 828 has high PSRR. LME49600 not so great, but as it will be inside NFB loop of OPA828 its PSRR should be improved.
As is, I’m leaning on conclusion that super regulators would bring small or no benefit at all. Any experience or opinions are welcome.
Lay out your PCB with super regulators AND jumpers or switches which allow you to remove/bypass the super regulators completely. Then have an audiophile buddy or spouse secretly set them one way (you are out of the room) and seal the chassis. Listen to the gear in this configuration for two or three weeks. Then have the same person set the jumpers/switches the other way and seal the chassis. Listen to that configuration for two or three weeks. Now you know which one you like most ... AND you also know whether it makes a big difference sonically, or a small difference, or no difference at all.
Some good points here , and something I haven't thought of. Where I live the nightwatch uses these walkie-talkies and an EMS service 200 meter further has a big antenna to communicate with their drivers . Maybe I will notice this when they use it , because of the fast opamp-buffer.
I considered the OPA1641 , a very affordable opamp with excellent numbers and has duals and quads , but with a typical 1 mV offset x 5 gain = 5mV on the output > no thanks ...or a servo which I want to avoid. Also T.I 's datasheet starts with : Superior Sound Quality ... Soundplus ... I start to roll my eyes .
Not just is the LME easier to solder , it is easier and better to cool.
I use Elvee's DeNoiser , not really a super regulator , but if you look at "the Wire" https://www.diyaudio.com/community/...-performance-headphone-amplifier-pcbs.179298/
OPC/Owen shows some pretty impressive graphs of noise/ distortion with only the regular LM3X7 regulators .
Like Mark says , is it going to make a dfference you can hear ?