OK. I see now. Well its a DIY'er place and simple, effective circuits for the hobbiest do very well here.
There are circuits here that are pushing things..... super low distortion at high power levels and high SR and stable into speaker loads..... and the like, as alternatives.
If we could get some of the super good IC front-ends only or a way to make a replacement OPS for them..... almost everyone would be done with this matter for low levels. Your SWOPA was an excellent discrete design but the micro-scopic sm fets used killed it for many of the DIY'er. Did for me.
The DIY'er has many choices. John's is just one.... but because his designs have garnered praise over many years, there is something else about them which isnt being discussed; A curious DIY'er is going to try his designs, if he can, and compare.
THx-RNMarsh
Last edited:
If I may, as JC has already commented on this fairly recently:
Lower noise. As John mentioned earlier the two input devices are effectively in parallel and a differential pair has each half effectively in series. If each transistor had the same e sub n, the complementary connection will have factor square root two lower noise than a single device, whereas each differential pair with, say, the same NJFET will have factor square root two higher noise than a single device. So to achieve the equivalent input noise with a differential pair you will need half the e sub n per side, which requires four devices per half. That comes to eight devices (or equivalently larger devices) than the complementary arrangement's input stage.
Typically the distortion from the diff pair arrangement can be reasonably low without global feedback or local CFP arrangements. But so can the complementary topology, although the matching of P with N is more touchy and with respect to capacitances, deteriorating at high frequencies.
Although it is not just about parts---there are still a few squeezes left in the toothpaste tube in terms of topologies, as for example the floating current source approaches discussed here and elsewhere---there are some good new discrete parts, although not developed with audio in mind, an example being the BF862, which availability is likely to remain high since AM radio is big in China. And despite the lack of cooperation from Toshiba, LIS is doing some good stuff.
Although the NEED may not be pressing, and I wouldn't try to design and pitch a truly-state-of-the-art phono preamp for incorporation in a mass-market integrated amplifier, I may still be interested in seeing what can be done. Or, in the case of I-V converters for so-called current-output DACs, there may be some enhancements when one steps outside the box of op amps used as transimpedance amplifiers. That latter chore is greatly facilitated by the degrees of freedom available from hybrid designs.
I'm not personally feedback-averse, but it can be fun to see what can be achieved if the requirements are to minimize it.
For me the most worrisome thing is the near-absence of correlation between measurements and reviewers' preferences---but we could go on for years about that.
The heart of the matter with audio "not improving" is that nearly all systems, irrespective of cost, still struggle with adequately replaying the source material - a visit to a hifi show makes this glaringly obvious; if they were halfway decent a particular recording would sound very much alike going from room to room, which of course it doesn't. Most systems are deeply flawed, and that's a great shame - the recordings have it within them to sound spectacularly good, but one would never know it from the typical standard of replay. Having on occasion hit a very nice sweet spot of system quality I'm very much aware that audio reproduction can be almost "unbelievably" good, but the thinking has to progress from the current fixations on getting one particular measurement brilliantly impressive, and the forgetting that the whole is key - one unresolved flaw brings the whole to its knees, and all the fussing about some component part being the best one can get is completely irrelevant to fixing that ..
Hi Jan, Brad,
Measurements. The trick is to interpret what you are seeing, and to understand that something you can hear very clearly might only be a small difference in your readings. I know that both of you examine the residual from a THD meter, because that is one place where differences do become more clear. We can readily see the difference between low and high order harmonics, or just non-correlated noise.
These things are all very clear to you and others who search for answers. But, we aren't very good at disseminating this information to those folks who are just doing their best to understand and be included in this audio journey. Therefore, they do believe that you can't measure audible differences and that makes them vulnerable to all the nonsensical babbling from the advert departments. It also makes our lives difficult whenever discussions take place about the art and engineering of audio design. They are just trying to be a part of something they enjoy. Thus the great arguments erupt between those two camps-that-must-not-be-named. The main reason for the split is economic, some folks directly benefit from the confusion when the market doesn't know what to believe. The mainstream audiophiles gave up and we all lost out.
In the late 70's we had the trust of the consumer. We being the entire audio industry. We all tried to be truthful about what we were selling, and this was the peak in quality and sound quality - on average. Through the mid-late 80's some of us were lying to the public, plus business became interested in the audio business (it was too small a market before to bother with). That was the end of an industry and the start of consumers understanding they are being lied to. Today, only a few care about sound quality. That pie has been shrinking ever since.
So, want more pie? Let's try to be honest about the science behind audio products. Deliver good value and maybe even make the stuff last. Of course, we have to get the buying public over the sticker shock now. It will take a while to rebuild an industry, but it took time to tear it down to begin with.
We know that when the design starts sounding better, and the measurements are also getting better, we are on the right track. Successful design takes into account listening and measuring. I could say a lot more, but this is enough.
-Chris
Measurements. The trick is to interpret what you are seeing, and to understand that something you can hear very clearly might only be a small difference in your readings. I know that both of you examine the residual from a THD meter, because that is one place where differences do become more clear. We can readily see the difference between low and high order harmonics, or just non-correlated noise.
These things are all very clear to you and others who search for answers. But, we aren't very good at disseminating this information to those folks who are just doing their best to understand and be included in this audio journey. Therefore, they do believe that you can't measure audible differences and that makes them vulnerable to all the nonsensical babbling from the advert departments. It also makes our lives difficult whenever discussions take place about the art and engineering of audio design. They are just trying to be a part of something they enjoy. Thus the great arguments erupt between those two camps-that-must-not-be-named. The main reason for the split is economic, some folks directly benefit from the confusion when the market doesn't know what to believe. The mainstream audiophiles gave up and we all lost out.
In the late 70's we had the trust of the consumer. We being the entire audio industry. We all tried to be truthful about what we were selling, and this was the peak in quality and sound quality - on average. Through the mid-late 80's some of us were lying to the public, plus business became interested in the audio business (it was too small a market before to bother with). That was the end of an industry and the start of consumers understanding they are being lied to. Today, only a few care about sound quality. That pie has been shrinking ever since.
So, want more pie? Let's try to be honest about the science behind audio products. Deliver good value and maybe even make the stuff last. Of course, we have to get the buying public over the sticker shock now. It will take a while to rebuild an industry, but it took time to tear it down to begin with.
We know that when the design starts sounding better, and the measurements are also getting better, we are on the right track. Successful design takes into account listening and measuring. I could say a lot more, but this is enough.
-Chris
Last edited:
I see some poor analog designers getting brain fade when clocks are mentioned. Let's see how much I can get right...
Jitter.
Where does jitter come from?
A clock is an oscillator. Jitter is another way of looking at phase noise.
In the most general terms, an oscillator is an amplifier yoked to a bandpass filter. There may be other descriptions. The amplifier replenishes the energies dissipated in the tank resistance and lost to the clock sink. The amplifier is broadband and unconstrained can randomly produce output at more than one frequency. The filter constrains the output (close) to the desired frequency.
Phase noise is deviation from the centre frequency.
Because all practical filters have finite Q, the output is not always constrained to the intended frequency.
A crystal filter (resonator) is very little different from an LC filter. In fact we can get resonators made from other materials such as ceramic, and some complex structures, such as SAW filters, but essentially these circuit elements behave like LC resonators of very high Q.
So phase noise is intrinsic to this method of frequency generation, but in general, the higher the Q, the smaller the pnoise.
While it is possible to create an oscillator (relaxation?) with a square or triangle output, many oscillators have a sinewave, or near-sinewave waveform.
Most (TTL) digital circuits have a preference for a square clock with 50% duty cycle.
You can easily obtain an approximation to such by feeding a sinewave into a comparator, but a moment's thought will expose some problems. If the comparator flips exactly at zero-crossing, well and good, but if the sinewave has an added DC component it can change the duty cycle from 50%, and similarly any AM noise can push the crossover backwards and forwards in time. Jitter.
It's common for some digital chips to provide a pair of pins to drive a crystal, with a 3rd. pin supplying supplying a squared clock to the rest of the circuitry, in addition to that supplied to the device internals. Alternatively a couple of glue-logic NAND gates are commonly used.
Anyone who wants to play around with simulating clocks should try to get their hands on some of the Cadence tools. Learn how to generate that negative resistance.
Jitter.
Where does jitter come from?
A clock is an oscillator. Jitter is another way of looking at phase noise.
In the most general terms, an oscillator is an amplifier yoked to a bandpass filter. There may be other descriptions. The amplifier replenishes the energies dissipated in the tank resistance and lost to the clock sink. The amplifier is broadband and unconstrained can randomly produce output at more than one frequency. The filter constrains the output (close) to the desired frequency.
Phase noise is deviation from the centre frequency.
Because all practical filters have finite Q, the output is not always constrained to the intended frequency.
A crystal filter (resonator) is very little different from an LC filter. In fact we can get resonators made from other materials such as ceramic, and some complex structures, such as SAW filters, but essentially these circuit elements behave like LC resonators of very high Q.
So phase noise is intrinsic to this method of frequency generation, but in general, the higher the Q, the smaller the pnoise.
While it is possible to create an oscillator (relaxation?) with a square or triangle output, many oscillators have a sinewave, or near-sinewave waveform.
Most (TTL) digital circuits have a preference for a square clock with 50% duty cycle.
You can easily obtain an approximation to such by feeding a sinewave into a comparator, but a moment's thought will expose some problems. If the comparator flips exactly at zero-crossing, well and good, but if the sinewave has an added DC component it can change the duty cycle from 50%, and similarly any AM noise can push the crossover backwards and forwards in time. Jitter.
It's common for some digital chips to provide a pair of pins to drive a crystal, with a 3rd. pin supplying supplying a squared clock to the rest of the circuitry, in addition to that supplied to the device internals. Alternatively a couple of glue-logic NAND gates are commonly used.
Anyone who wants to play around with simulating clocks should try to get their hands on some of the Cadence tools. Learn how to generate that negative resistance.
Last edited:
Damn. Thought I had a problem with my display until I realized it actually was a "c/o" and not "%."
se
Regarding negative resistance: an offshore subcontractor to Harman got concerned when we had a rash of slow-starting (like 30s) crystal oscillators (due to board contamination/leakage in the vicinity of a TSSOP uC package throwing the bias of the gate oscillator off and into a region of lower gain). They read the crystal manufacturer's specification, and asked us where they could source a specific negative resistor.
It was challenging to formulate a reply.
It was challenging to formulate a reply.
FM Tuner for Jitter analysis
I did a lot on this and wrote it up here: http://www.diyaudio.com/forums/blogs/1audio/983-fm-tuner-jitter-analysis.html
It does work pretty well as a diagnostic tool but measuring the jitter/phase noise of an oscillator turns out to be pretty remote from the actual jitter induced modulation on the audio output. You need to know exactly how the DAC chip processes the data and the clocks to even guess what is important. Differential noise across the grounds could contribute enough modulation to degrade the clocks even if everything else is flawless.
Crystals are microphonic and sensitive to vibration. In some high stress precision applications the trim cap is modulated to correct for the vibration. Connecting the case in various ways will affect the vibration impacting the crystal. It may also detune the oscillator.
Look at the analog output if you need to assess jitter. Look at the clocks to see if they are the source of jitter issues. Then look at everything else.
Richard- I looked at the Yokagawa TIA's and have an HP 5370 ( https://www.febo.com/pages/hp5370b/ ) which turns out to have significantly higher resolution. In the end it did not show me much. The measured clock jitter of a stock ESI Juli@ board is around 20 pS (probably less) so I hit the wall with that method really fast. The FM tuner and various modulation analyzer methods came next with no new insights. Finally the zero beat with a reference oscillator etc. works but is not useful for in circuit measurements to see what is happening in actual operation. Measuring the output of a DAC with a PC spectrum analyzer was far more useful and a whole lot easier.
I did a lot on this and wrote it up here: http://www.diyaudio.com/forums/blogs/1audio/983-fm-tuner-jitter-analysis.html
It does work pretty well as a diagnostic tool but measuring the jitter/phase noise of an oscillator turns out to be pretty remote from the actual jitter induced modulation on the audio output. You need to know exactly how the DAC chip processes the data and the clocks to even guess what is important. Differential noise across the grounds could contribute enough modulation to degrade the clocks even if everything else is flawless.
Crystals are microphonic and sensitive to vibration. In some high stress precision applications the trim cap is modulated to correct for the vibration. Connecting the case in various ways will affect the vibration impacting the crystal. It may also detune the oscillator.
Look at the analog output if you need to assess jitter. Look at the clocks to see if they are the source of jitter issues. Then look at everything else.
Richard- I looked at the Yokagawa TIA's and have an HP 5370 ( https://www.febo.com/pages/hp5370b/ ) which turns out to have significantly higher resolution. In the end it did not show me much. The measured clock jitter of a stock ESI Juli@ board is around 20 pS (probably less) so I hit the wall with that method really fast. The FM tuner and various modulation analyzer methods came next with no new insights. Finally the zero beat with a reference oscillator etc. works but is not useful for in circuit measurements to see what is happening in actual operation. Measuring the output of a DAC with a PC spectrum analyzer was far more useful and a whole lot easier.
Hi Demian,
I am going to do some tests pretty much as you describe. I can also bring into play several spectrum analyzers of high precision. Remember, my goal is not to improve something..... but to see what jitter differences I can see between HD download files being played and My CD player...
For example, playing files in a USB memory stick thru a Araliti PK100 to a BenchMark DAC-2 is WAY better sounding than playing a CD in a mid price CD player using its fiber optic output.
Many here have expressed -in the strongest terms - they dont think there is any difference.
THx-RNMarsh
Last edited:
I do have a practical side as well. But, I dont think of the High-End of any field that way, however. IMO, it is the same with IC opamps etc. They keep trimming the specs -- maybe well past what most need.
But, I do it because it is fun to learn about subtle things..... an intellectual curiosity. I also enjoy the sound of music a lot more when it is reproduced to a high level of perfection. But, I spend a lot of time listening to far less than the best (in-car, project room, video etc).
How could it help to have ever better test equipment? I have no idea until after the tests. But by what I have read and what people in the know say.... jitter is still a problem esp in interfacing. Besides, I am bored right now. It is an occupational hazard of being retired and having too much time on one's hands. I am waiting for the dust to settle in Nepal and then I am out of here and heading back to my condo in Bangkok and then a trip to Nepal to see what i can do there... maybe get my adopted family outta there for awhile for a much needed break. One day, I wont be able to travel due to old age wear and tear affects.... then I'll really get into this hobby....
THx-RNMarsh
Thank you Ed for replying. These are all good points.
At these low frequencies, the field inside the box is a 'near field' (up to ~1/6 wavelength), predominately magnetic.
Under these conditions, magnetic field attenuates at a rate of (1/distance)^3, the much weaker electric field attenuates at a rate (1/distance)^2.
This, in the case when there were no reflecting surfaces around the source.
At mains frequencies and first harmonics, thin steel enclosure absorption loss (due to eddy currents) is only 3-6 dBs.
The shielding –towards the outside world-of the box works mostly through reflection.
Steel enclosure reflects back almost all the magnetic field that is generated internally, so, inside this steel box, attenuation of the magnetic field with distance, change drastically from the free-field case.
Geometrical details become very important as to the strengths of direct vs reflected field.
Within a magnetic field, strong coupling occurs with low impedance circuits.
The impedance of the secondary/diodes/cap/20 Ohm resistor circuit is low, thus – seen either as a TX or a RX- it needs to form the minimum area loop possible.
(In “Noise Reduction Techniques in Electronic Systems” by Henry Ott, there is accumulated a wealth of relevant knowledge).
As for the structure of the harmonics, I thinks it is an indication that in your example, the predominant polluter is the transformer and not the diodes/cap/resistor electric circuit loop.
Test of this hypothesis: Change the transformer’s magnetic flux by changing the volts/turn of the primary through a variac-but keep the secondary voltage the same by adding or removing secondary turns-and watch the change of the spectrum.
Higher volts/turn of the primary (higher core's magnetic flux) should increase the amplitude and the number of odd harmonics of the pick-up spectrum.
For Richard
Products | AfterMaster | Audio Labs
George
Last edited:
Well said John, keep the faith going eh wot! these silly buggers keep chiming in with their physics and in most cases good advice.
So what is the best clock money can buy? Even amongst audiophiles there are disagreements...
So many, myself have stated that from experience these sort of MODs will change the sound but are probably not improving the jitter spectrum...
- Status
- Not open for further replies.