Hi John,
I've been looking for a good analogue scope for a long time. I'll keep looking.
Yes i made that mistake of not playing music when i took the photos last night, so I've made some more today.
In both situations a 1nf cap was used across DEM pins, but the frequency and the voltage amplitude increased with the resistors in place.
I used jumper pins so i could easily change from one situation to the other.
I also made video recordings of both situations, and you can tell quite easily which of the two has the most jitter by the frequency counter range. Ill post the links.
With 6k8s - http://goo.gl/YRgsjd
Without 6k8s - http://goo.gl/7iVW3x
I've been looking for a good analogue scope for a long time. I'll keep looking.
Yes i made that mistake of not playing music when i took the photos last night, so I've made some more today.
In both situations a 1nf cap was used across DEM pins, but the frequency and the voltage amplitude increased with the resistors in place.
I used jumper pins so i could easily change from one situation to the other.
I also made video recordings of both situations, and you can tell quite easily which of the two has the most jitter by the frequency counter range. Ill post the links.
With 6k8s - http://goo.gl/YRgsjd
Without 6k8s - http://goo.gl/7iVW3x
Attachments
Hi John,
Any thoughts as to whether construction type of the resistors used could make a meaningful difference?.
Shane
Hi Ryan,
The videos tell quite a story compared with the screen shots. Thanks for this.
Shane
Hi Ceglar,
I suggest to use small SMD MELF or thin film resistors, 1206 or 805 size and use shortest possible wiring in order to minimise EMI pick-up.
I now use a 100pF SMD mica cap from Cornell Dubilier
This capacitor should be placed as close to pin 16 & 17 as possible.
I attached a picture of the SD2-player prototype.
You can see the two MELF resistors and the mica cap close to pins 16 & 17.
I suggest to use small SMD MELF or thin film resistors, 1206 or 805 size and use shortest possible wiring in order to minimise EMI pick-up.
I now use a 100pF SMD mica cap from Cornell Dubilier
This capacitor should be placed as close to pin 16 & 17 as possible.
I attached a picture of the SD2-player prototype.
You can see the two MELF resistors and the mica cap close to pins 16 & 17.
Attachments
Thanks, John.
Better question might have been: any potential issues using carbon film resistors with <3mm lead length, reason for is that in situe there are some new (small but present) anomalies, apparent with (HQ) headphones above and beyond what was before, but still with some annoyance as compared without (6k8 ea)
Sincerely,
Shane
Better question might have been: any potential issues using carbon film resistors with <3mm lead length, reason for is that in situe there are some new (small but present) anomalies, apparent with (HQ) headphones above and beyond what was before, but still with some annoyance as compared without (6k8 ea)
Sincerely,
Shane
Not a problem, I really enjoyed playing around with it all yesterday.
Yes, they do indeed show more of what's happening. Its would be nice to have the equipment and time to measure exactly by how much it has improved... Ill keep dreaming and just buy a new scope...
Thanks for sharing this simple and effective modification john.
Hi John,
when looking at the SD2 picture it seems that many things have changed completely:
- no tuning fork with ECL divider, all clock signals generated by the DSP?
- Rhopoint resistors only for IU conversion
- 3.5mm Jack Stereo sockets
- no trimmer for REG-GND adjustment: does your output stage with reactor circuit include an DC offset
Hi JOSI1,
There have been a lot of modifications that were not covered on this thread.
Many things changed because of the reactor circuit, it enabled me to pick up TDA1541A design again. This circuit is used in the UD1, UD2, and SD2-player. It has effect on jitter, output spectrum, feedback loop stability in connected circuits and equipment, and impact of power supply noise spectrum.
- The masterclock is now based on a single crystal that is used for both, fixing the clock frequency and as crystal bandpass filter. It produces a sine wave output signal with high spectral purity. The output signal is taken from a capacitive divider / impedance matching circuit. The advantage is that it is no longer necessary to fine tune the oscillator, match multiple crystals and adjust the crystal bandpass filter.
- Using a separate divider for BCK in combination with a microcontroller with built in frequency multiplier causes problems with phase locking between BCK and WS / DATA. This phase locking issue has to do with the ECL clock divider for BCK and the clock divider that is part of the frequency multiplier inside the microcontroller. The random phase locks are a result of pulse skipping (interference) and random counter start positions. These issues cannot be fixed and phase changes occur during every cold or hot reset and as a result of external interference sources (sparking mains switch for example).
With the reduced jitter sensitivity provided by the reactor circuit we were finally able to generate all I2S signals by the controller while maintaining sound quality. The phase lock issues are now eliminated as all 3 I2S signals now “follow” the phase shifts introduced by the counter for the clock multiplier.
I/V conversion requires resistors that offer lowest possible noise and a low inductance. I tested many I/V resistors, carbon, metal film, bulk metal foil, wire wound and so on. I also tested high wattage multiple paralleled resistors. The Honeycomb Mobius and Rhopoint econister were the preferred I/V resistors after a very long testing period.
Resistors in the stepped shunt volume control only need to offer lowest possible noise, so inductive wire wound resistors can be used here. The resistors used in the SD2-player volume control provide very similar sound quality as the Rhopoint econister. This is not the case with all wire wound resistors and has to do with properties of the resistive element, welding and lead wire material.
The UD1, UD2, and SD2-players were miniaturized as this offers a number of advantages. So smaller audio sockets were required, even this turned out to offer advantages.
I use sockets that have a silver plated copper alloy spring contact on GND, L, and R. This forms a point contact (Think of Eichmann RCA plugs) and this has advantages.
here a link to a 6moons review of RCA interlinks:
6moons audio reviews: 3 RCA*Connectors
The copper alloy springs are soldered directly to the copper PCB traces, this results in fewer thermo couplings compared to a wired RCA socket. The socket is mounted flat on the PCB and the PCB has a solid ground plane with no embedded pads nor traces.
This forms excellent screening and very low interference levels. Finally all RF interference signals will be coupled directly into the ground plane. Listening tests showed that this results in clearly improved sound quality over high quality RCA sockets that do not have point contacts.
Both volume sections , the one in the SD1-player and in the SD2-player are identical, same relays, same resistors and so on. Yet the sound quality of the SD2 volume unit alone (using AUX input) is clearly better than that of the SD1-player. The only differences are the 3.5mm jack socket and the solid ground plane.
Interlink tweaking is no problem as there are also audiophile 3.5mm jack to RCA interlinks like these for example:
Hi-Fi Cables + Interconnect — TurntableLab.com
The SD2-player contains a new I/V & output stage. TDA1541A output is connected to a 1K bias resistor that connects to +5V. It sources 5mA bias current and ensures that all TDA1541A bit currents flow back into +5V. The 1K resistor can be used as constant current source as both 5V supply voltage and the voltage on the TDA1541A outputs remain virtually constant. With a constant 5V across the 1K bias resistor this will result in a constant 5mA passive bias current source. Since the load impedance on the TDA1541A is extremely low, the 1K impedance has neglible effect on output signal amplitude. The advantage of a 1K resistor over a CCS is that it forms a linear impedance, offers low noise, and low stray capacitance.
The current difference (5mA bias current minus DAC current) is fed into a PNP grounded base buffer. This buffer is controlled by a servo circuit that together with the PNP transistor emulates a very low load impedance. It also maintans 0V DC on the TDA1541A output. Measured ac voltage at the TDA1541A output is below 50uVpp. This translates to a load impedance of approx. 12.5 milli Ohms (0.012 Ohms).
The very low ac on the TDA1541A output ensures low distortion and very good bit switching characteristics as the TDA1541A output is basically short circuited to GND.
The current buffer drives the I/V resistor that connects to a negative voltage. We now have almost 15V headroom (TDA1541A -15V supply). The I/V resistor connects to a second servo circuit that automatically matches the DC voltage on the I/V resistor (audio output) with the -5V supply. The -5V supply is now used as ground reference.
We have DC-coupling and DC-offset is adjusted automatically. Because of the large correction range, this DC servo also acts as DC protector.
This I/V & output stage runs on TDA1541A +5V, -5V, and -15V supplies, so no additional supply voltages are needed. The TDA1541A supply voltages are derived from 3 cascaded quatrode shunt regulators that offer extremely low output impedance. The precision voltage references are powered by the very clean output voltage of the quatrode shunt regulator itself. This results in excellent ripple and noise rejection.
These shunt regulators are powered by an external 24V DC power supply. Sound quality does not change with external power supply quality and even modern (high frequency) universal 24V switch-mode adapters can be used without problems. I tested 24V lithium ion battery supply (two 12V battery packs connected in series), linear supply and 24V SMPS with no audible change in sound quality.
There have been a lot of modifications that were not covered on this thread.
Many things changed because of the reactor circuit, it enabled me to pick up TDA1541A design again. This circuit is used in the UD1, UD2, and SD2-player. It has effect on jitter, output spectrum, feedback loop stability in connected circuits and equipment, and impact of power supply noise spectrum.
- The masterclock is now based on a single crystal that is used for both, fixing the clock frequency and as crystal bandpass filter. It produces a sine wave output signal with high spectral purity. The output signal is taken from a capacitive divider / impedance matching circuit. The advantage is that it is no longer necessary to fine tune the oscillator, match multiple crystals and adjust the crystal bandpass filter.
- Using a separate divider for BCK in combination with a microcontroller with built in frequency multiplier causes problems with phase locking between BCK and WS / DATA. This phase locking issue has to do with the ECL clock divider for BCK and the clock divider that is part of the frequency multiplier inside the microcontroller. The random phase locks are a result of pulse skipping (interference) and random counter start positions. These issues cannot be fixed and phase changes occur during every cold or hot reset and as a result of external interference sources (sparking mains switch for example).
With the reduced jitter sensitivity provided by the reactor circuit we were finally able to generate all I2S signals by the controller while maintaining sound quality. The phase lock issues are now eliminated as all 3 I2S signals now “follow” the phase shifts introduced by the counter for the clock multiplier.
I/V conversion requires resistors that offer lowest possible noise and a low inductance. I tested many I/V resistors, carbon, metal film, bulk metal foil, wire wound and so on. I also tested high wattage multiple paralleled resistors. The Honeycomb Mobius and Rhopoint econister were the preferred I/V resistors after a very long testing period.
Resistors in the stepped shunt volume control only need to offer lowest possible noise, so inductive wire wound resistors can be used here. The resistors used in the SD2-player volume control provide very similar sound quality as the Rhopoint econister. This is not the case with all wire wound resistors and has to do with properties of the resistive element, welding and lead wire material.
The UD1, UD2, and SD2-players were miniaturized as this offers a number of advantages. So smaller audio sockets were required, even this turned out to offer advantages.
I use sockets that have a silver plated copper alloy spring contact on GND, L, and R. This forms a point contact (Think of Eichmann RCA plugs) and this has advantages.
here a link to a 6moons review of RCA interlinks:
6moons audio reviews: 3 RCA*Connectors
The copper alloy springs are soldered directly to the copper PCB traces, this results in fewer thermo couplings compared to a wired RCA socket. The socket is mounted flat on the PCB and the PCB has a solid ground plane with no embedded pads nor traces.
This forms excellent screening and very low interference levels. Finally all RF interference signals will be coupled directly into the ground plane. Listening tests showed that this results in clearly improved sound quality over high quality RCA sockets that do not have point contacts.
Both volume sections , the one in the SD1-player and in the SD2-player are identical, same relays, same resistors and so on. Yet the sound quality of the SD2 volume unit alone (using AUX input) is clearly better than that of the SD1-player. The only differences are the 3.5mm jack socket and the solid ground plane.
Interlink tweaking is no problem as there are also audiophile 3.5mm jack to RCA interlinks like these for example:
Hi-Fi Cables + Interconnect — TurntableLab.com
The SD2-player contains a new I/V & output stage. TDA1541A output is connected to a 1K bias resistor that connects to +5V. It sources 5mA bias current and ensures that all TDA1541A bit currents flow back into +5V. The 1K resistor can be used as constant current source as both 5V supply voltage and the voltage on the TDA1541A outputs remain virtually constant. With a constant 5V across the 1K bias resistor this will result in a constant 5mA passive bias current source. Since the load impedance on the TDA1541A is extremely low, the 1K impedance has neglible effect on output signal amplitude. The advantage of a 1K resistor over a CCS is that it forms a linear impedance, offers low noise, and low stray capacitance.
The current difference (5mA bias current minus DAC current) is fed into a PNP grounded base buffer. This buffer is controlled by a servo circuit that together with the PNP transistor emulates a very low load impedance. It also maintans 0V DC on the TDA1541A output. Measured ac voltage at the TDA1541A output is below 50uVpp. This translates to a load impedance of approx. 12.5 milli Ohms (0.012 Ohms).
The very low ac on the TDA1541A output ensures low distortion and very good bit switching characteristics as the TDA1541A output is basically short circuited to GND.
The current buffer drives the I/V resistor that connects to a negative voltage. We now have almost 15V headroom (TDA1541A -15V supply). The I/V resistor connects to a second servo circuit that automatically matches the DC voltage on the I/V resistor (audio output) with the -5V supply. The -5V supply is now used as ground reference.
We have DC-coupling and DC-offset is adjusted automatically. Because of the large correction range, this DC servo also acts as DC protector.
This I/V & output stage runs on TDA1541A +5V, -5V, and -15V supplies, so no additional supply voltages are needed. The TDA1541A supply voltages are derived from 3 cascaded quatrode shunt regulators that offer extremely low output impedance. The precision voltage references are powered by the very clean output voltage of the quatrode shunt regulator itself. This results in excellent ripple and noise rejection.
These shunt regulators are powered by an external 24V DC power supply. Sound quality does not change with external power supply quality and even modern (high frequency) universal 24V switch-mode adapters can be used without problems. I tested 24V lithium ion battery supply (two 12V battery packs connected in series), linear supply and 24V SMPS with no audible change in sound quality.
Hi John,
Yes, Congratulations on your SD-TDA1541A Masterpiece...
Must be a great feeling after developing it for about 8 years...
It's also nice to that you have a higher output signal now (1,4V RMS vs 0,8V RMS) with your new I/V & output stage...
Have you also implemented the circuit that you mention in post #5042 ??
Quote:
{It was not easy but I managed to convert I2S 64 bits / frame (44.1/16) two’s complement into offset binary, 16 bit data bursts for TDA1541A in simultaneous mode.
The whole point of this exercise is to obtain longest possibe period of silence (absence of I2S interference in the TDA1541A and attached circuits) during each sample period.
I2S stream now looks as follows, LE pulse is generated, lasting one BCK clock cycle. 16 bits of data are clocked in simultaneously. Then the I2S stream is shut down completely (WS, DATA and BCK shut down for 75% of the sample period).
The created silence (absence of switching noise) has positive impact on sound quality.}
Thanks,
Peter
Yes, Congratulations on your SD-TDA1541A Masterpiece...
Must be a great feeling after developing it for about 8 years...
It's also nice to that you have a higher output signal now (1,4V RMS vs 0,8V RMS) with your new I/V & output stage...
Have you also implemented the circuit that you mention in post #5042 ??
Quote:
{It was not easy but I managed to convert I2S 64 bits / frame (44.1/16) two’s complement into offset binary, 16 bit data bursts for TDA1541A in simultaneous mode.
The whole point of this exercise is to obtain longest possibe period of silence (absence of I2S interference in the TDA1541A and attached circuits) during each sample period.
I2S stream now looks as follows, LE pulse is generated, lasting one BCK clock cycle. 16 bits of data are clocked in simultaneously. Then the I2S stream is shut down completely (WS, DATA and BCK shut down for 75% of the sample period).
The created silence (absence of switching noise) has positive impact on sound quality.}
Thanks,
Peter
I'm curious to how these "LIC" Lithium Ion Capacitors might sound in audio applications ??
They are low voltage so you would have to wire them in series...
Anybody have any experience with "LICs" ??
TAIYO YUDEN CO., LTD.
They are low voltage so you would have to wire them in series...
Anybody have any experience with "LICs" ??
TAIYO YUDEN CO., LTD.
Nice to see the price on high-capacity SD cards has come down...
Like the PNY 256GB for $120
PNY Technologies 256GB Elite Performance SDXC P-SDX256U1H-GE B&H
And Newegg has it for $106 - $96
Newegg.com - 256gb sdxc
Like the PNY 256GB for $120
PNY Technologies 256GB Elite Performance SDXC P-SDX256U1H-GE B&H
And Newegg has it for $106 - $96
Newegg.com - 256gb sdxc
Impressive work, as always.
I will have to write a letter
Max, open a Curento restaurent in Europe ! It will have a lot of sucess and please sonorise it with SD player with your first gains... We will see you there !
Like your country Max, beautifull landscapes!
Thank you, dear Eldam.
I guess you meant "curanto" perhaps?
http://www.espejodeluna.cl/blog/wp-content/uploads/2012/04/curanto-serve.jpg
Unfortunately, I am allergic to sea-food
You Nederlands are all welcome to visit me (not all at once) and I hope I will be also welcome when I have the time and resources to visit our digital guru there, a trip that I am postponing for too long now...
Best wishes to you,
M.
Last edited:
Yes "curanto".... Curanto, curento : same same but different : I eated it and heard the name without readed it ! (and remember a funny song)
Not dutch myself but french : same, same but different : our flag is the same but colors are horizontal... we never suceed to let it float, so we put its horizontal... you know they are specialists of floating poldairs !
Remenber now with your sea food alergie than we had this conversation already before (do you remember : i eated a curanto on Chiloe!). Never found again this South Chilie speciality elswhere ! I liked it... haha I'm more chilian than you, i'm a mapuch who like the Curanto !
With the years memorie gone... I remember begann this thread when I was a child, now I'm 45 and realize if I keeped 100 euros/year, I will have a SD1 ... But nowadays, players haven't not rca plugs anymore ! Strange are the things !
off topic, sorry... adds moment about Chili Tours...
Not dutch myself but french : same, same but different : our flag is the same but colors are horizontal... we never suceed to let it float, so we put its horizontal... you know they are specialists of floating poldairs !
Remenber now with your sea food alergie than we had this conversation already before (do you remember : i eated a curanto on Chiloe!). Never found again this South Chilie speciality elswhere ! I liked it... haha I'm more chilian than you, i'm a mapuch who like the Curanto !
With the years memorie gone... I remember begann this thread when I was a child, now I'm 45 and realize if I keeped 100 euros/year, I will have a SD1 ... But nowadays, players haven't not rca plugs anymore ! Strange are the things !
off topic, sorry... adds moment about Chili Tours...
Last edited:
Ooops!
Ou plûtot, m***! Veuillez excuser ma maladresse...je mourrais de sommeil.
Vous autres Francais êtes bien sûr mes invités si vous me faites l' honneur de visiter mon pays et je serais très fier d'être vôtre très humble serviteur
Sorry for obliged OT.
M.
I do not have full knowledge of proposed reactor circuit. But from what i could understand does it mean that John is coming back to a full circle? Is the reactor circuit doing the linear interpolation like DI-DAC (in analog domain instead of digital domain)?
I am quoting post # 61 as reference.
I am quoting post # 61 as reference.