Hi,
Okay, first, one source of jitter in "common" receiver chips is the PLL. It contains a simple VCO (Voltage Controlled Oscillator). This is one of the limiting factors.
Such oscillators have a fair amount of self-noise and hence jitter. The oscillators in the various Cirrus Logic Chips for example have around 200pS jitter as a result of the VCO used.
Now, the next point is the PLL bandwidth. Actually, the "low" 200pS jitter are only attained by making the bandwidth of the PLL very wide (~ 10KHz for CS Chips). Here the PLL actually works as feedback circuit reducing the noise of the VCO.
As a result the self noise of the PLL is low, but the suppression of jitter riding in on the SPDIF signal is very minimal or non at all.
Jitter in the SPDIF signal is a direct result of the modulation scheme, where a "1" is encoded as a double cycle of 2 X 64 X Fs while a "0" is encoded as a single cycle of 64 X Fs.
Due to the various factors in sending high frequency squarewaves the result is that any phase detector trying to lock onto this pattern (which it must to extract the bitclock) will experience signal dependent shifts. In other words, the bitpattern send modulates the recovered clock.
In addittion, any noise overlaying the signal will also cause the "trigger" points to shift, for further problems in the phase detector.
The net result is that output from a common garden Receiver usually contains much more than it's specified "self jitter" (200pS for Cirrus Logic, 50pS for Burr Brown).
Now, non of this is a problem, if we use (for arguments sake) a VCXO (Voltage Controlled Xtal Oscillator) and a PLL filter of sufficiently narrow bandwidth.
So an ideal "traditional" receiver would incorporate pin's for "pullable" crystals at 44.1 & 48KHz base sample rate, on board varicap diodes to tune these crystals and a nice low noise op-amp with suitable external PLL filter components to make a low noise, very slow PLL.
No-one ever made such a receiver. Wolfson uses advanced disgital means to in essence approximate such a system "emulated in software" but their hardware has too few samples in it's buffer to push the jitter lowpass frequency into the single Hz figure (the minimum needed IMHO) and stops killing incomming jitter at 100Hz (this is still a 100 fold improvement over Cirrus Logic though).
Finally, while a well designed digital output on a CD-Player may very well have low jitter, few examples of such exist in the commercial domain. In my own SPDIF out equipped CD-Player I do apply re-clocking and a special "high power" driver before using a simple, transformerless output, thus ensuring low jitter and a near perfectly square waveform, but most commercial designs apply much less care.
Okay, ignoring secondary PLL's (these have many implentation issues that can make the performance much worse than just the VCXO by itself), I have some data for integrated CD-Players where we do not need any PLL to recover clocks, but instead just distribute our clock around the device, so we can attain the best possible situation.
First, any logic circuits are subject to something called LIM (Logic Induced Modulation) by Meitner. What it means is that "ground bounce" and "powersupply noise" can modulate the switching point of logic devices (especially CMOS) and as a result can induce jitter.
As all commercial chips have bondwires and leadframes to the PCB, the actual silicon chip inside the case will be subject to certain amounts of both ground bounce and supply modulation that CANNOT be remedied in any way. So, this amount of "LIM" is unavoidable.
I find that using re-clocking directly before the DAC Chip from a super low jitter clock and using optimum power supply arrangements on DAC & reclocker can produce sub 50pS output Jitter, so far I have been unable to attain anything better than around 20pS myself in such systems.
This already implies individual, galvanically separated supplies for DAC, Reclocker, Clock and other Logic and quite esotheric decoupling strategies, so methodes well past anything commonly implemented by either DIY'ers or in commercial (even super high end) products.
To give you an idea how I implemented in one Products that I helped design, here a block diagram showing nthe different powersupplies.
Based on Stereophile's measurements of conmmercial DAC's and CD-players, for most the "jitter barrier" seems above 100pS, in many cases VERY MUCH above 100pS.
For what it is worth, around 100pS Jitter are good enough to not compromise CD-Audio's quality. If we double the sample rate or add one bit we need to halve the allowable jitter.
So if for example we want to do 18Bit resolution at 88.2KHz sample rate we need to keep jitter to below around 12 pS, which already seems past the means at our disposal, never mind 24 Bit/192KHz or such.
I hope this helps to clarify why the "jitter business" matter so much.
Ciao T
Okay, first, one source of jitter in "common" receiver chips is the PLL. It contains a simple VCO (Voltage Controlled Oscillator). This is one of the limiting factors.
Such oscillators have a fair amount of self-noise and hence jitter. The oscillators in the various Cirrus Logic Chips for example have around 200pS jitter as a result of the VCO used.
Now, the next point is the PLL bandwidth. Actually, the "low" 200pS jitter are only attained by making the bandwidth of the PLL very wide (~ 10KHz for CS Chips). Here the PLL actually works as feedback circuit reducing the noise of the VCO.
As a result the self noise of the PLL is low, but the suppression of jitter riding in on the SPDIF signal is very minimal or non at all.
Jitter in the SPDIF signal is a direct result of the modulation scheme, where a "1" is encoded as a double cycle of 2 X 64 X Fs while a "0" is encoded as a single cycle of 64 X Fs.
Due to the various factors in sending high frequency squarewaves the result is that any phase detector trying to lock onto this pattern (which it must to extract the bitclock) will experience signal dependent shifts. In other words, the bitpattern send modulates the recovered clock.
In addittion, any noise overlaying the signal will also cause the "trigger" points to shift, for further problems in the phase detector.
The net result is that output from a common garden Receiver usually contains much more than it's specified "self jitter" (200pS for Cirrus Logic, 50pS for Burr Brown).
Now, non of this is a problem, if we use (for arguments sake) a VCXO (Voltage Controlled Xtal Oscillator) and a PLL filter of sufficiently narrow bandwidth.
So an ideal "traditional" receiver would incorporate pin's for "pullable" crystals at 44.1 & 48KHz base sample rate, on board varicap diodes to tune these crystals and a nice low noise op-amp with suitable external PLL filter components to make a low noise, very slow PLL.
No-one ever made such a receiver. Wolfson uses advanced disgital means to in essence approximate such a system "emulated in software" but their hardware has too few samples in it's buffer to push the jitter lowpass frequency into the single Hz figure (the minimum needed IMHO) and stops killing incomming jitter at 100Hz (this is still a 100 fold improvement over Cirrus Logic though).
Finally, while a well designed digital output on a CD-Player may very well have low jitter, few examples of such exist in the commercial domain. In my own SPDIF out equipped CD-Player I do apply re-clocking and a special "high power" driver before using a simple, transformerless output, thus ensuring low jitter and a near perfectly square waveform, but most commercial designs apply much less care.
Okay, ignoring secondary PLL's (these have many implentation issues that can make the performance much worse than just the VCXO by itself), I have some data for integrated CD-Players where we do not need any PLL to recover clocks, but instead just distribute our clock around the device, so we can attain the best possible situation.
First, any logic circuits are subject to something called LIM (Logic Induced Modulation) by Meitner. What it means is that "ground bounce" and "powersupply noise" can modulate the switching point of logic devices (especially CMOS) and as a result can induce jitter.
As all commercial chips have bondwires and leadframes to the PCB, the actual silicon chip inside the case will be subject to certain amounts of both ground bounce and supply modulation that CANNOT be remedied in any way. So, this amount of "LIM" is unavoidable.
I find that using re-clocking directly before the DAC Chip from a super low jitter clock and using optimum power supply arrangements on DAC & reclocker can produce sub 50pS output Jitter, so far I have been unable to attain anything better than around 20pS myself in such systems.
This already implies individual, galvanically separated supplies for DAC, Reclocker, Clock and other Logic and quite esotheric decoupling strategies, so methodes well past anything commonly implemented by either DIY'ers or in commercial (even super high end) products.
To give you an idea how I implemented in one Products that I helped design, here a block diagram showing nthe different powersupplies.
An externally hosted image should be here but it was not working when we last tested it.
Based on Stereophile's measurements of conmmercial DAC's and CD-players, for most the "jitter barrier" seems above 100pS, in many cases VERY MUCH above 100pS.
For what it is worth, around 100pS Jitter are good enough to not compromise CD-Audio's quality. If we double the sample rate or add one bit we need to halve the allowable jitter.
So if for example we want to do 18Bit resolution at 88.2KHz sample rate we need to keep jitter to below around 12 pS, which already seems past the means at our disposal, never mind 24 Bit/192KHz or such.
I hope this helps to clarify why the "jitter business" matter so much.
Ciao T
ThorstenL,
thanks! that is a great explanation and is very helpful. I work with RF electronics at my current job; so I have done a bit of work on low jitter clock sources, but these are running at 2.4GHz. My experience with audio clocking is non-existent, so this is a great learning experience.
thanks! that is a great explanation and is very helpful. I work with RF electronics at my current job; so I have done a bit of work on low jitter clock sources, but these are running at 2.4GHz. My experience with audio clocking is non-existent, so this is a great learning experience.
ThortsenL,
I have been investigating the Pass D1 I/V stage a bit more. Reading through Owen's thread, it occured to me that the power on transient of the circuit causes the DAC output pin to be pulled negative for some time until the gate of the MOSFET is pulled up to it's steady state resting point.
I assume you are familiar with the circuit. I attached Owen's schematic.
Since the gate is driven by a resistor divider (with a large filter cap) it takes a long time for the voltage to come up. However, the resistor to the negative supply comes on immediately with the supply.
I am worried if I use this with the PCM1794 chip the outputs will be damaged during power on. The datasheet says the maximum current for any pin is 10mA, but that is all. No other information given. Since the part is powered from only +5V and ground, I can see a large current could be "pulled" from the chip to the negative supply through the MOSFET source resistor.
Do you have any experience with this?
I have been investigating the Pass D1 I/V stage a bit more. Reading through Owen's thread, it occured to me that the power on transient of the circuit causes the DAC output pin to be pulled negative for some time until the gate of the MOSFET is pulled up to it's steady state resting point.
I assume you are familiar with the circuit. I attached Owen's schematic.
Since the gate is driven by a resistor divider (with a large filter cap) it takes a long time for the voltage to come up. However, the resistor to the negative supply comes on immediately with the supply.
I am worried if I use this with the PCM1794 chip the outputs will be damaged during power on. The datasheet says the maximum current for any pin is 10mA, but that is all. No other information given. Since the part is powered from only +5V and ground, I can see a large current could be "pulled" from the chip to the negative supply through the MOSFET source resistor.
Do you have any experience with this?
BTW, I checked out AMR website. Very nice products! I wish I could afford them.
Can you give me any hints on the tube section? I am interested in trying a tube output stage on my DAC instead of the ALeph P design, but I am not sure if there is a particular topology that would be the best.
Can you give me any hints on the tube section? I am interested in trying a tube output stage on my DAC instead of the ALeph P design, but I am not sure if there is a particular topology that would be the best.
Hi,
Sorry, not really. If I am really worried about stuff like that I add protection diodes, at least until I am ready to blow up the Chip without them by trying without them.
Ciao T
Sorry, not really. If I am really worried about stuff like that I add protection diodes, at least until I am ready to blow up the Chip without them by trying without them.
Ciao T
Hi,
Thank you, we try.
Well, I designed one that is commercial but not super expensive:
Universal Stereo Tube Output Stage | Diy HiFi Supply
Have a look. And there is always my old article, until the updated version is published:
http://www.fortunecity.com/rivendell/xentar/1179/theory/vasfda/vasfda.html
Ciao T
Thank you, we try.
Well, I designed one that is commercial but not super expensive:
Universal Stereo Tube Output Stage | Diy HiFi Supply
Have a look. And there is always my old article, until the updated version is published:
http://www.fortunecity.com/rivendell/xentar/1179/theory/vasfda/vasfda.html
Ciao T
Hi sir
And what is the "Right" kind of I/V opamps for the TDA1541A ?
Can you list them starting from the best one ?
Btw, wen we modifie a cd player with a better and faster I/V opamps, cd players use a resistor and a cap in the nfb of the I/V opamps (like in my image), the cap are quite high value for a slow I/V opamps, do we need to reduce the cap value with a faster I/V opamps ?
Thanx
Paul
Attachments
Last edited:
Hi,
There is now such a profusion. Much depends on the context.
My personal recommendation are any current feedback Op-Amp'swith high input stage quiescent current. These generally retain their low input impedance even when the feedback loop opens up due to transients and at the same time their spped is very high. Depending on the precise arrangements such an op-amp needs the lowpass capacitor in the feedback loop removed or a suitable resistor (to assure stability) in series with the inverting input.
One also needs to watch the power supplies, many modern CFB Op-Amp's only run +/-5V.
From the older generations afavourites are AD811 and LM6181 (or the LM6182 Dual), I am uncertain if either/any remain in production.
An alternative are the Burr Brown OPA860/OPA861 "diamond transistors", these allow I'V without feedback, Pedja "Audioal" Rogic documented their use with the TDA1541.
Ciao T
There is now such a profusion. Much depends on the context.
My personal recommendation are any current feedback Op-Amp'swith high input stage quiescent current. These generally retain their low input impedance even when the feedback loop opens up due to transients and at the same time their spped is very high. Depending on the precise arrangements such an op-amp needs the lowpass capacitor in the feedback loop removed or a suitable resistor (to assure stability) in series with the inverting input.
One also needs to watch the power supplies, many modern CFB Op-Amp's only run +/-5V.
From the older generations afavourites are AD811 and LM6181 (or the LM6182 Dual), I am uncertain if either/any remain in production.
An alternative are the Burr Brown OPA860/OPA861 "diamond transistors", these allow I'V without feedback, Pedja "Audioal" Rogic documented their use with the TDA1541.
Ciao T
Hi Sir
I have some voltage feedback opamp like the OPA627, LM4562 and THS4031. But I don't have any current feedback Op-Amp.
But, looking at the ecdesign thread, I seen a mosfet I/V amp, I include the circuit image, how about this one ?
Thanx
Paul
Attachments
Hi,
I have used the OPA627/637 in my re-build of the Philips LHH-1000 DAC (basically a CD/DA-12 by another name). I used OPA637 for the I/V with a silver/mica cap from inverting input to ground for both stability and to slow down any "spikes". I used OPA627 for the lowpass filter and BUF634 for the output driver and Elna Silmic/Elna Cerafine/Sanyo Os-Con's all over in the various positions where they make sense. As my work was more a "restauration" than a modification I left the original circuit design, so no non-os or other such stuff.
This "refreshed" LHH-1000 was a very good DAC, outperforming a lot of stuff it came up against.
It is not REALLY an Amp. It is a current conveyor, similar to the circuits using Diamond Transistors.
The issues here are that the I/V resistor is not referenced to ground but to the supply rail, which means there is zero noise immunity to power supply noise and the output carries substantial DC levels neccesitating a suitable coupling capacitor (I do not feel the quality of the one ecdesigns promotes suffices BTW). So the apparent simplicity in the I/V section is paid for with increased complexity elsewhere.
Remember TANSTAAFL.
So to make this circuit work really well you need a very esotheric powersupply with very low noise and an essentially linear impedance that should be in the audio range preferably 100000 times lower than the I/V resistor and a high quality coupling cap.
This gets very complex and expensive quickly, but the results of such simple circuitry (and the complex external circuitry) when it is done well can be quite outstanding.
Ciao T
I have used the OPA627/637 in my re-build of the Philips LHH-1000 DAC (basically a CD/DA-12 by another name). I used OPA637 for the I/V with a silver/mica cap from inverting input to ground for both stability and to slow down any "spikes". I used OPA627 for the lowpass filter and BUF634 for the output driver and Elna Silmic/Elna Cerafine/Sanyo Os-Con's all over in the various positions where they make sense. As my work was more a "restauration" than a modification I left the original circuit design, so no non-os or other such stuff.
This "refreshed" LHH-1000 was a very good DAC, outperforming a lot of stuff it came up against.
It is not REALLY an Amp. It is a current conveyor, similar to the circuits using Diamond Transistors.
The issues here are that the I/V resistor is not referenced to ground but to the supply rail, which means there is zero noise immunity to power supply noise and the output carries substantial DC levels neccesitating a suitable coupling capacitor (I do not feel the quality of the one ecdesigns promotes suffices BTW). So the apparent simplicity in the I/V section is paid for with increased complexity elsewhere.
Remember TANSTAAFL.
So to make this circuit work really well you need a very esotheric powersupply with very low noise and an essentially linear impedance that should be in the audio range preferably 100000 times lower than the I/V resistor and a high quality coupling cap.
This gets very complex and expensive quickly, but the results of such simple circuitry (and the complex external circuitry) when it is done well can be quite outstanding.
Ciao T
Hello sir
The Adcom cd player that I have do use, for the I/V, an opamp with only the number 6AA mark on them, so I don't know wich brand and number are this 6AA opamp.
One thing I don't like with opamp are that they need lot of negative feedback to have low thd and the input transistors on the die are so close to the output transistors that they are prone to oscillate.
I've look again in the very long ecdesign thread and there was another one with bjt transistor, here it is.
How about this one ?
Thanx
Paul
Attachments
Last edited:
Hi,
You may wish to get a service manual for your player.
This is debatable. Feedback is just another tool.
Ideally the open loop bandwidth of a circuit before applying feedback is at least twice that of the (fully bandwidth limited using a function parallel to that of the Amplifier open loop gain rolloff) signal.
Using very fast (CFB) structures can fulfil this requirement easily, even for the fast changing output current of a DAC. If using rather "slow" amplifiers within feedback loops it may be necessary to to slow down the signals to something the closed loop system can handle, with a cap on the inverting input/output of DAC.
I don't think you have this right. Op-Amp's, discrete, monolithic etc. oscillate due to the feedback loop and the fact that feedback signal eventually rotates phase such that the negative feedback becomes positive feedback.
I feel you just want someone telling you "build this one", please do not expect such answers from me. Analyse the schematic you have shown and consider the issues and problems with it. This will then allow you to make correct choices.
If this is too much for you, just fit OPA627 without any further modification. This is safe, will sound good and does not require any special skills or knowledge.
Ciao T
You may wish to get a service manual for your player.
This is debatable. Feedback is just another tool.
Ideally the open loop bandwidth of a circuit before applying feedback is at least twice that of the (fully bandwidth limited using a function parallel to that of the Amplifier open loop gain rolloff) signal.
Using very fast (CFB) structures can fulfil this requirement easily, even for the fast changing output current of a DAC. If using rather "slow" amplifiers within feedback loops it may be necessary to to slow down the signals to something the closed loop system can handle, with a cap on the inverting input/output of DAC.
I don't think you have this right. Op-Amp's, discrete, monolithic etc. oscillate due to the feedback loop and the fact that feedback signal eventually rotates phase such that the negative feedback becomes positive feedback.
I feel you just want someone telling you "build this one", please do not expect such answers from me. Analyse the schematic you have shown and consider the issues and problems with it. This will then allow you to make correct choices.
If this is too much for you, just fit OPA627 without any further modification. This is safe, will sound good and does not require any special skills or knowledge.
Ciao T
Hello Thorsten
This weekend there was the Montreal audio show, and Avatar Acoustic was there and they did some public listening sessions of the AMR CD-777, it was very nice sounding at the Direct Master II NOS mode , I was very please by the sound. Very good work indeed.
At the Direct Master II mode, how much db are the analog filter of the AMR CD-777 ?
Bye
Gaetan
This weekend there was the Montreal audio show, and Avatar Acoustic was there and they did some public listening sessions of the AMR CD-777, it was very nice sounding at the Direct Master II NOS mode , I was very please by the sound. Very good work indeed.
At the Direct Master II mode, how much db are the analog filter of the AMR CD-777 ?
Bye
Gaetan
Last edited:
So...not being sure where to enter this...I'm looking to actually build my own dac - and by build, I mean...a kit that gives me a PCB (and a project box wouldn't hurt) and then gives me a component list that I need and lets me tinker some.
Basically, I was hoping to do coax/optical/USB all in one, with a selector for which one I want (and if need be, coax can be dropped as it's the least importent to me.)
What kits might be good? Where should I be looking? It seems simple to find USB dacs and all...
And check my logic too: I want a optical DAC because I have a Creative X-Fi in my PC; the XFi lets me EQ stuff which, when I have my speakers on at night is great - kill the bass and I'm a good citizen in my apartment building...I'd prefer not to lose that. By having optical off the front panel or coax off the back panel, I can get full EAX sound with games, and keep the EQ options. Right? Is there anything wrong with my reasoning here?
Thanks in advance!
Basically, I was hoping to do coax/optical/USB all in one, with a selector for which one I want (and if need be, coax can be dropped as it's the least importent to me.)
What kits might be good? Where should I be looking? It seems simple to find USB dacs and all...
And check my logic too: I want a optical DAC because I have a Creative X-Fi in my PC; the XFi lets me EQ stuff which, when I have my speakers on at night is great - kill the bass and I'm a good citizen in my apartment building...I'd prefer not to lose that. By having optical off the front panel or coax off the back panel, I can get full EAX sound with games, and keep the EQ options. Right? Is there anything wrong with my reasoning here?
Thanks in advance!
Hi,
Sorry, but AMR would not like to see such information made oublic. The filter in the CD-777 and in the CD-77.1 of current vintage do not work with LCR resonance circuits, this much I can say, so they differ much from the original designs of mine and the ones Pedja Rogic made.
However Pedja's design is pretty good, I use it myself in one of my experimental DAC's at home.
Ciao T
Sorry, but AMR would not like to see such information made oublic. The filter in the CD-777 and in the CD-77.1 of current vintage do not work with LCR resonance circuits, this much I can say, so they differ much from the original designs of mine and the ones Pedja Rogic made.
However Pedja's design is pretty good, I use it myself in one of my experimental DAC's at home.
Ciao T
Hello Thorsten
Pedja, in his dac schematic, used an asynchrone reclock (50 or 60 mhz) and for the IV amp he used an AD844, did you use those same reclock and IV amp ?
I'm not a big fan of asynchrone reclock.
Thank
Bye
Gaetan
Last edited:
Hi,
Not at all, I referred to his "pre commercial" phase, I think time-machined/mirrored at decibeldungeon... His analog filter design.
Ciao T
Not at all, I referred to his "pre commercial" phase, I think time-machined/mirrored at decibeldungeon... His analog filter design.
Ciao T
- Status
- Not open for further replies.