Hi,
Bernhard. I have lived nearly halve my life in England. English and German differ in the way they are used. German is a very precise language where you say precisely what you want to say.
English is not like that (note that I mean ENGLISH, not american, which is a whole other language that at best is tenously connected to English and follows many different rules in use [like at least every fifth word must be an expletive]).
You need to to speak, listen between the lines and words, where is the emphasis placed? That is why on-line many use extra emphasis in English to replace these non-linguistic cues.
For example the title "Building the ultimate NOS DAC using TDA1541A" can have the following Emphasis options:
Building the ultimate NOS DAC using TDA1541A
Building the ultimate NOS DAC using TDA1541A
Building the ultimate NOS DAC using TDA1541A
Building the ultimate NOS DAC using TDA1541A
You need to ask John which was his...
More fun, with certain soft subjects the emphasis changes over time.
I have said elsewhere that this makes English the perfect language for a German to express ones frustration over the lack of precision in the language.
Ciao T
most of us are not native english speakers, but I think that "creating the best DAC that can be made with the TDA1541 chip"
would be "Building the ultimate TDA1541A DAC"
Bernhard. I have lived nearly halve my life in England. English and German differ in the way they are used. German is a very precise language where you say precisely what you want to say.
English is not like that (note that I mean ENGLISH, not american, which is a whole other language that at best is tenously connected to English and follows many different rules in use [like at least every fifth word must be an expletive]).
You need to to speak, listen between the lines and words, where is the emphasis placed? That is why on-line many use extra emphasis in English to replace these non-linguistic cues.
For example the title "Building the ultimate NOS DAC using TDA1541A" can have the following Emphasis options:
Building the ultimate NOS DAC using TDA1541A
Building the ultimate NOS DAC using TDA1541A
Building the ultimate NOS DAC using TDA1541A
Building the ultimate NOS DAC using TDA1541A
You need to ask John which was his...
More fun, with certain soft subjects the emphasis changes over time.
I have said elsewhere that this makes English the perfect language for a German to express ones frustration over the lack of precision in the language.
Ciao T
This is what I expected.
Aerial thermal noise masks 1 bit / 10 °C
-->
Limitations of the 1541A:
MSB adjust not possible
Passive I/V not possible
Selection difficult, 2 channels in one package
No ECL compatible input (not confirmed)
I totally agree with this interpretation. Title tells us that the goal is "ultimate NOS DAC" and the secondary caveat is "using TDA1541".
You didn't reply my question, which was: "Which is the best NOS DAC to your view?"
Hi,
Or necessary.
Of course it's possible.
We select them all the time for level, it is easy. Channel match is excellent.
Not sure what that is supposed to mean. All Chip DAC's I know are not ECL compatible (it is after all negative voltage logic), but instead are TTL/CMOS compatible.
Ciao T
Or necessary.
Of course it's possible.
We select them all the time for level, it is easy. Channel match is excellent.
Not sure what that is supposed to mean. All Chip DAC's I know are not ECL compatible (it is after all negative voltage logic), but instead are TTL/CMOS compatible.
Ciao T
It is necessary if you want the best possible low level performance.
I meant no post amplification, no stepup transformers, no "tricks".
I meant selection for low level distortion.
No problem ;-)
Hi Bernhard,
Volume control attenuates the 2V signal to say 100mV. This signal is then amplified say 50 times by the power amp. We then have 5V rms, this would produce approx. 3 watts in 8 Ohm.
With 2V full scale, LSB corresponds to 2 / 2^16 = 30.5uV, this signal is clearly measurable as you already demonstrated. System noise floor could be something like 4uV.
After attenuation to 100mV (volume control) LSB represents 0.1 / 2^16 = 1.5uV, this is already below system noise floor.
When you play music recorded with a transfer level of say -70dB (Sheffield Lab test CD "My Disc track #40) it is likely you will hear just ambient noise when listening few meters away from the speaker using average volume setting. If I am correct the attenuated signal has dynamic range of approx. (16 * 6) -70 = 26dB. This corresponds to approx. 26 / 6 = 4.3 bits resolution. These signals will sound distorted when "blown up" (volume on maximum, listening inches away from the speaker), with or without DLE of 0.5 LSB or lower.
The point I am trying to make is why go through all the trouble attempting to reduce DLE to zero if the result is likely to be inaudible under normal listening conditions? The "blown up" signal distorts anyway because it's only 4.3 bits.
Yes, because the 6 MSBs are automatically adjusted by the DEM (Dynamic Element Matching) circuit. The 10 LSBs are fixed. With the PCM56 all bits are fixed except MSB that can be trimmed.
It -is- possible, but placing the passive I/V resistor between output and GND introduces amplitude limitations because of the output compliance that every current output DAC has. If a bias current is present (2mA with the TDA1541A) this will cause an additional DC offset voltage across the passive I/V resistor, further limiting allowable ac signal amplitude at the DAC chip output.
That's why I inserted a current buffer between DAC chip output and passive I/V resistor that simply passes the DAC current to this passive I/V resistor. This way large signal swings can be obtained across the passive I/V resistor while minimizing ac / dc at the DAC output. Grounded base / gate / grid circuits offer large bandwidth (up to transient frequency), low input impedance and high output impedance.
I agree, there can be small tolerances between L and R channels, but tolerances between L and R speakers can be significantly higher (also affected by speaker placement and room acoustics).
It is possible to design a I2S splitter circuit that outputs DATA L/L for DAC chip #1 and R/R for DAC chip #2. This in combination with chip matching (matching chips with similar differences between L and R channels) would fix the "problem".
It has TTL compatible inputs that also seem to work with CME (Current Mode Logic), major problem with low level drive signals say 400mV is DC-offset in combination with thermal drift (diodes in the TDA1541A I2S input circuit).
Volume control attenuates the 2V signal to say 100mV. This signal is then amplified say 50 times by the power amp. We then have 5V rms, this would produce approx. 3 watts in 8 Ohm.
With 2V full scale, LSB corresponds to 2 / 2^16 = 30.5uV, this signal is clearly measurable as you already demonstrated. System noise floor could be something like 4uV.
After attenuation to 100mV (volume control) LSB represents 0.1 / 2^16 = 1.5uV, this is already below system noise floor.
When you play music recorded with a transfer level of say -70dB (Sheffield Lab test CD "My Disc track #40) it is likely you will hear just ambient noise when listening few meters away from the speaker using average volume setting. If I am correct the attenuated signal has dynamic range of approx. (16 * 6) -70 = 26dB. This corresponds to approx. 26 / 6 = 4.3 bits resolution. These signals will sound distorted when "blown up" (volume on maximum, listening inches away from the speaker), with or without DLE of 0.5 LSB or lower.
The point I am trying to make is why go through all the trouble attempting to reduce DLE to zero if the result is likely to be inaudible under normal listening conditions? The "blown up" signal distorts anyway because it's only 4.3 bits.
Yes, because the 6 MSBs are automatically adjusted by the DEM (Dynamic Element Matching) circuit. The 10 LSBs are fixed. With the PCM56 all bits are fixed except MSB that can be trimmed.
It -is- possible, but placing the passive I/V resistor between output and GND introduces amplitude limitations because of the output compliance that every current output DAC has. If a bias current is present (2mA with the TDA1541A) this will cause an additional DC offset voltage across the passive I/V resistor, further limiting allowable ac signal amplitude at the DAC chip output.
That's why I inserted a current buffer between DAC chip output and passive I/V resistor that simply passes the DAC current to this passive I/V resistor. This way large signal swings can be obtained across the passive I/V resistor while minimizing ac / dc at the DAC output. Grounded base / gate / grid circuits offer large bandwidth (up to transient frequency), low input impedance and high output impedance.
I agree, there can be small tolerances between L and R channels, but tolerances between L and R speakers can be significantly higher (also affected by speaker placement and room acoustics).
It is possible to design a I2S splitter circuit that outputs DATA L/L for DAC chip #1 and R/R for DAC chip #2. This in combination with chip matching (matching chips with similar differences between L and R channels) would fix the "problem".
It has TTL compatible inputs that also seem to work with CME (Current Mode Logic), major problem with low level drive signals say 400mV is DC-offset in combination with thermal drift (diodes in the TDA1541A I2S input circuit).
So, you have no ground to attack the thread's header as you interpret it.
So, it's clear that you are on personal crusade on NOS DACs in general and on TDA1541A in particular.
Also it is quite clear that you judge electronics implementations by their technical data, not by the way they sound.
Hi Joshua G, I agree with you. Maybe technical data are not "supreme", but I've been listening my Philips CD player, with 5 independent supplies (my last design) for TDA, reclock, and for my last clock (-184dbc/hz@0.1ppm), and all the servo supplies in other chassi, the HV and filaments in another chassi, without the SAA7220, with 1uf decoupling capacitors from Cornell Dubilier (the same SMD ones that John uses but with more capacity), with ECD reclock scheme, Gomes output, I/V resistor, good quality components, etc...., and the detail level, the silence level, the naturality, the open scene, bass extension, sweet voices, real crystal clear highs, the instruments location, etc, etc, are all really astounding, far way better than modern DAC's like Esoteric, Wadia, Electrocompaniet, Krell, Audioresearch, etc. And I can hear all this brands daily, and I can compare between them. So, dear Bernhard, something is wrong with the way you think about the TDA1541A......
Kind regards,
Kind regards,
It's very simple: Absence of oversampling is enough to prefer a nos tda over modern DACs. 1541A nos becomes a honey-like taste.
Before you ask..
I neither make business with my DAC nor do I publish schematics.
Reasons are that it is expensive to build as it now uses 32 DAC chips ( could be upgraded to 64 chips in future ) and many other parts, it also is physically very large and needs measuring equipment to make custom parts and do adjustments. It is not possible to reduce the number of chips or change other details without redesigning the whole thing. Also it requires a CD transport that has a word clock input.
Maybe one day when I have a final version...
... it is NOS.
Last edited: