The first thing I did on my Marantz CD63 player when it was out of warranty was to disconnect and bypass the HDAM module which only serves to muddy the sound, and put in a better OPAs, at that time the OPA2604 brought a really nice sound improvement over the original JRC. I didn't change the voltage regulators, I didn't know what to put back then in 1997. Electrolytic capacitors were all Elna Silmic, nothing could be improved there. At the Line output, Elna Silmic coupling electrolytic capacitors (2+2pcs. back to back) were removed and high-quality PP block capacitors were placed. The difference between this slightly modified device and the factory device was heaven and earth. In the meantime, that Marantz went down in history, as a good device for its time, but far from the so-called High End sound.
As for the modern OPAs I tried, MUSES 02 performed best from the integrated ones, followed by the OPA1611/12, OPA1655/56 an some others. From the discrete ones, Burson (V6C, V6V) and Sparkos SS3601/02 are the best sonically so far.
Sparkos have discrete 3pin voltage regulators that replace the 78XX/79XX with nice improvements in sound quality, but they aren't cheap. Also on ebay there are 3pin regulators with SMD LT1963/3015 circuits which also bring improvement in sound over the 78XX/79XX and are not too expensive. I've tried it all.
The superior sound of preamplifiers and similar devices is obtained by using a Shunt voltage regulators, as the ultimate power supply solution. This of course comes at a price and requires larger transformers and more space. I had the opportunity to try the same preamplifier circuit with LM317/337 and Shunt power supply, the difference is clearly in favor of the Shunt regulators.
In building the tube preamplifier, I tried two types of HV power supplies. First diode bridge + classic CRC, then CRCRC, to get the best sound with CRC + TL783C regulator. The next HV power supply will probably have a tube rectifier and CLC filtration. Tube rectifiers have a significant effect on the sound, regardless of whether it is followed by voltage stabilization or just CRC/CLC filtering. If I hadn't heard it, I wouldn't have believed it.
It all depends on the user. A good power supply for OPAs can also be obtained using CLC or CRC filtration. The subjective character of the sound is always different, for some it means nothing or has no experience as a listener, and for some it means a lot.
Last edited:
Don't mix up the need for power of an op amp with the CD63 dac output signal needs...Here's about a headphones amplifier not image filters, pwm emi filters, jfet folded cascodes with no psrr, phono or tapehead preamps ...Is that magic regulator going to solve the hunger in Africa too? 78xx regs are specified up to 100khz, unlike lm317...Can you hear 100khz? Besides 78xx regs are not all equal.The difference between makes and manufacturers are significant.
For me, the best regulator is no regulator.
For me, the best regulator is no regulator.
Last edited:
As I said, I am passing on my many years of amateur experience from the DIY zone, related to making and listening to various gadgets. You can take my word for it, or not. Take the simplest line preamp, say with a cheap OPA1612 with voltage gain of 3x. It works linearly depending on the load on the output (20k+360pF) up to 600kHz and with a smaller capacitive load it goes solidly over 1MHz. Then add an RC filter to it at 100khz. And you will hear the difference, regardless of the fact that these things happen well above 20kHz. A voltage regulator that is declared at a limit of 100Khz will produce the same or similar effects. For some it is important, for some it is not. As the years go by, it will matter less and less to me as well. 
Last edited:
Batman is not my friend. Bats have evolved very sensitive hearing to cope with their nocturnal activity. Their hearing range varies by species; at the lowest it can be 1 kHz for some species and for other species the highest reaches up to 200 kHz. Bats that can detect 200 kHz cannot hear very well below 10 kHz. An I really love bass.
The last measurement showed that limit of my hearing is at about 14kHz.
The last measurement showed that limit of my hearing is at about 14kHz.
Here is my suggestion for a simple headphone amplifier. I didn't make it, it's still in the plan, but the concept was tested in the Line preamp with lower quiescent current of the mosfet (40mA). The advantage of this is that it sounds better than the OPA itself. In this schematic any OPA will operate in A class up to 78 mA RMS and single ended up to approximately some 2.4 VRMS at the output. There is enough current for 32ohm headphones, and enough voltage for 600ohm headphones. The supply voltage can be higher, up to the OPA limit, +-18V usually. You just have to pay attention to the heating of critical elements (mosfet and source resistors). And some circuit for DC protection of headphones should be added to the output.
Attachments
Last edited:
What's the point of running so much idle current at +-12V if you leave the 10 ohms output series resistor outside the feedback loop? You might consider such wasted power for driving 28 ohms Hifiman headphones, then the 10 ohms resistor wastes 1/3rd of the output current without any distortion mechanism applied on it as well as the headphones themselves being out of the feedback loop.It may be that such headphones are mostly resistive in nature, but that 10 ohms resistor is gonna be taking 1/3rd of the output energy for nothing.For higher inpedance and higher sensitivity headphones class A is useless already.
Like everything else, I tried it, I liked it. Most headphones are not resistive in nature, just like speakers. The resistor is there as some short-term short-circuit protection for the mosfet. They can also be kicked out. It may be necessary to add Zobel to the output, I have to try.
Source resistors can be replaced with CCS, but this increases the parts count and adds two heatsinks. And it changes the sound of course.
Source resistors can be replaced with CCS, but this increases the parts count and adds two heatsinks. And it changes the sound of course.
Nice and simple, but not high fidelity with a long tail of higher-order harmonics at the output and asymmetric clipping. The reason is the same as in Elliott's: the output source follower is not a particularly linear circuit, esp. at large signal swings, and the opamp doesn't gave enough gain to correct it, esp. at the high end.
If your heatsink is large enough, you can jack up the quiescent current of the MOSFET to, say 500mA (its source resistor will be hot, too). This will clear up things a bit. A CCS (another N-ch MOSFET + small NPN) will help, too.
Thanks Nixie - it looks like a well-thought-out design, and I have a couple of my own thoughts:
If I did include those source follower FETs or similar, in my design, I assume that will reduce the upper voltage the circuit can produce. probably not an issue for me, since most IEMs are high sensitivity, and low impedance, maybe even less than 32ohms. So I may not actually need much voltage swing out of this thing. I am also curious how close to the bottom rail those source followers can get. I suppose I could just simulate it. 🙃
Some background on my project: This will be for driving IEMs in a musical performance situation, not driving high-end reference headphones for hearing every nuance in Dark Side of the Moon, haha. Although I do want as great a sound as I can get in a simple-ish design.
I'm currently trying to decide whether I want to bother with dual supplies at all, or use a single supply and AC coupling capacitors. Space, and especially height, will be at a premium for me, so no big poly cap components allowed, they will likely all be ceramics. I'm going to be designing a PCB for this, and there's about 1/4" of height available, so everything will be SMT and short. The board will be approximately 2" x 4", all components on the topside only as the backside will be close to the bottom of the case.
If I do go with a dual supply, I will have to build a negative DCDC converter (and probably and LDO too?) The only supply available to me is a 12VDC, from which everything else must flow.
I'd build a composite amp based on either the LME49600 and LM4562 or the LME49600 and the OPA1612. The latter is a bit more difficult to get stable as it loves to oscillate in the 100-150 MHz range, but I was able to get it working well in the Neurochrome HP-1.
The "typical application schematic" in the LME49600 data sheet works well.
The OPA1622 is another nice li'l headphone driver. It's hard to solder by hand, but there's an evaluation board available for around $30 that solves that problem nicely. Add an opamp input buffer and you're set. That's what I do in the HP-22.
Tom
The "typical application schematic" in the LME49600 data sheet works well.
The OPA1622 is another nice li'l headphone driver. It's hard to solder by hand, but there's an evaluation board available for around $30 that solves that problem nicely. Add an opamp input buffer and you're set. That's what I do in the HP-22.
Tom
I've read again your post after I discovered your Omicron design here on diyaudio which has basically the same output buffer as ESP... and I have smth to ask you:
1. why did you choose D.Self's measurements done on the class B version as opposed to the class AB of ESP version , then recommended your class A amp again? This looks like a " magician" job to me. The ESP design actually follows the best class AB bias prescription of D.Self..you know ...the ideal one!
Even Elliot said that his design's distortions were below his measuring equipment at the time.Not sure what that measuring equipment was at the time , but I know exactly what measuring equipment SSL and Soundcraft used and it was back then and still is the best on the market...you'll see what I mean down this page!
2. what class are most op amps you recommended for your Omicron drive which basically used the same output buffer as ESP and Studer and what is essentially their load when driving the output transistors?
I found your comment unfair because I have built similar amps to ESP'S with lf356, lf351, ne5532, lm4562, opa1612, opa604 opa2134, 32,opa2227, 2228 and they don't have anything wrong attached to the sound.Imagine that SSL and SOUNDCRAFT used similar amps in their 500k$ recording consoles that mixed any music you can think of in the past 40...50 years and no sound mixing engineer on Abbey Road protested. RME is using just modern paralleled op amps in their last equipment and they are no 1 on the small bussiness sound recording market for about 15 years.
I have nothing against using your Omicron design or TPA6120 or whatever technology you like, but you better prove there's sibillants and glassy mids produced by every commercial headphones amp based on similar designs to ESP !
Beware though cause they were produced in the numbers of tens of millions outhere!
Check out page 98 of SSL4000 service manual and you'll see it's even worse than the ESP version.Never heard of Peter Gabriel turning around because of its headphones monitor output.They used bfr39 which is worse than bd139 in terms of dissipated power and has no cap bypass over the bias diodes...
If I am to compare D.Self's Soundcraft K1 monitor output using paralleled 2sc2240 to your more elaborate design I'd choose the K1 anytime with maybe a better op amp than TL072 simply because it does all the essential work it's needed to do.I would have gone 2n3904/06 instead and go class B with 1n6263 schotky diode bias, but that's just me... Besides it has essentially identical bias to ESP version while being part of a professional recording console. I tried class B bias with slow 20Mhz, 400pF cob 100watts transistors and I couldn't hear a damn difference from a pure 50mA class A drive although I could measure crossover distortions starting with 4khz when driving 4 ohms loads, yet almost no crossover distortion at 20khz on 64 ohms load. You can drive headphones with anything and with no Zobel network...they have practically zero back EMF . Put in the most agressive zobel and no zobel with class A on 32 ohms headphones...no difference that you can hear or measure with sine signals.Yes you can stretch out to megaherts domain or step signal response if you want, but that will not change a thing with real program music which is just sine with most material program found below 12khz...Read Bob Katz on that if you can't believe me!
You can still go RME ADI-2 beast mode with 3 paralleled opa1688 per channel and no output transistors, but will you really hear a differrence?
Well here's a 20 years old " no bs phylosophy " you think you can undermine with bullet proof space alien technology applied to normal people ears...and you wouldn't be the first competent engineer( I don't doubt you are a competent engineer!) to exagerate people's needs in order to prove the necessity of a solution to a problem that doesn't exist, but human needs are way more modest and double blind tests are simply horrific scenarios to most high end companies.I think there still is a 50k$ prize available for anyone able to proove in a blind test that 24 bits sound better than 16 bits...
https://sound-au.com/philosophy.htm
Last edited:
Damn I have 4 lme49600 buffers for "ages" around me and never felt the impulse to build anything with them...maybe because it's too simple ...It's how human brain sometimes works...we always feel that more apparently complicated visual designs are better or worthy of our attention , yet lme49600 isn't really a simple buffer on the inside.You know what the ordinary diy man thinks...what would I then do if I see one IC can do all the job? We really fear that there's no place for diy in our conscience so we ignore it and build our astronomically complex, weird and useless designs that makes us feel creative.We are for maybe 2 decades now in a dire place where we can't really improve on major trends found in mass production so we like to fool ourselves with our little inventions and stole the 60's japanese spirit to prove our world is still in 1950's and no nuke ever destroyed our small Hiroshima...
We live in a different century now. and technology moved on. Forty years ago, DeLorean was a futuristic looking car, and Studer A810 was a state-of-the-art reel-to-reel audio recording machine with 63dB SNR and 0.5% distortion:
You can still enjoy both DeLorean and Studer, but state of the art is very different these days. Audio gear with 120dB SNR and less than 0.0005% distortion is inexpensive and easily available - that's 60dB, or one thousand times, better than Studer's. And yes, it does sound better (although not necessarily a thousand times better).
W.r.t. Omicron, you completely miss the point. It's not about Class A or B or AB. The Omicron thread shows measurements where the output stage leaves class A, and it is still 10,000 times more linear than the Studer. Omicron's ultra-high performance is due not to its 50mA of quiescent current, but to its sophisticated high-order contol loop and ultra-high loop gain, equivalent to a 700MHz Gain-Bandwidth Product. BTW, that control loop is one example of new technology developed since the times of DeLorean.
You can still enjoy both DeLorean and Studer, but state of the art is very different these days. Audio gear with 120dB SNR and less than 0.0005% distortion is inexpensive and easily available - that's 60dB, or one thousand times, better than Studer's. And yes, it does sound better (although not necessarily a thousand times better).
W.r.t. Omicron, you completely miss the point. It's not about Class A or B or AB. The Omicron thread shows measurements where the output stage leaves class A, and it is still 10,000 times more linear than the Studer. Omicron's ultra-high performance is due not to its 50mA of quiescent current, but to its sophisticated high-order contol loop and ultra-high loop gain, equivalent to a 700MHz Gain-Bandwidth Product. BTW, that control loop is one example of new technology developed since the times of DeLorean.
Studer was a discussion found two pages back.My last discussion was about SSL, Soundcraft( D.SELF design) and RME monitor output...you know...the monitor outputs used by the very engineers who made most of the music you're gonna use to test your equipment and draw conclusions on high endish audio theology...I'd give any audio equipment I ever touched for a full functioning A810...and you have no ideea what that means!I can source or replicate any digital technology available, but I can't build any A810...
It would be nice if someone could do a simulation of that circuit. I did not try that circuit for headphones, but I tried an identical circuit with a lower current through a Mosfet (line preamp). I have a modest Hantek oscilloscope with some FFT of limited capabilities. When measuring a sine signal up to 3V RMS no harmonics are visible, but I think that oscilloscope can only display THD down to -60dB (0.1%).
There is no problem with voltage swing, raise the power supply to +-15V and increase the coolers. The output of the OPA without the signal is approximately at +3.4V for IRF510, so you have at least 10V peak before clipping the positive period. This may only be relevant for high impedance headphones. As for the single power supply, I'm against it. Also the coupling capacitors must be electrolytic for 32ohm, I'm against that too. DC coupling in turn requires a DC protection circuit for the headphones. Maybe you should add a Zobel say 100ohm+10nF in parallel to the output.
I said I tried the Burson V6 classic single OPA for direct 32ohm headphone drive. It plays very nice, but it works in AB class in that case. Connection was full DC. That would be the simplest possible headphone amplifier for you. The +-12VDC power supply is quite enough so that they don't heat up too much.