Dear Ecdesigns:
Thank you for taking my tip into consideration
Your explanation about "resistor's sound" and opamp's feeback is what I've comme to learn with time, but in a very clear and DIDACtic exposition.
About resistors, power supply, output section...ect, I think all DYIers have his own preferences and will experiment with many options for your brilliantly develloped DAC.
I, for example, could experiment with superregulated supplies (provided that current consumption is not too high), wich may be better than batteries (at a higher cost).
I digress...
I have a friend that sweares that Mosfet output is great, but I 've never tried it...
We are all expectant about your work...
Cheers
M
Thank you for taking my tip into consideration
Your explanation about "resistor's sound" and opamp's feeback is what I've comme to learn with time, but in a very clear and DIDACtic exposition.
About resistors, power supply, output section...ect, I think all DYIers have his own preferences and will experiment with many options for your brilliantly develloped DAC.
I, for example, could experiment with superregulated supplies (provided that current consumption is not too high), wich may be better than batteries (at a higher cost).
I digress...
I have a friend that sweares that Mosfet output is great, but I 've never tried it...
We are all expectant about your work...
Cheers
M
decoupling caps
May be one of you can throw some light on this matter. I've seen some use Black Gates as decoupling caps on TDA1541A chip. Is there an advantage using BG vs others (eg, ceramic...) in this application? BTW ecdesigns, are the decouplings caps used in your ultimate DAC ceramic? Thanks.
May be one of you can throw some light on this matter. I've seen some use Black Gates as decoupling caps on TDA1541A chip. Is there an advantage using BG vs others (eg, ceramic...) in this application? BTW ecdesigns, are the decouplings caps used in your ultimate DAC ceramic? Thanks.
Decoupling
Hi, MGH
Thanks for your reply [post#263]
> There are 3 kind of decoupling cap's in the octal D-I DAC:
1) BC components foil capacitors on the 14 "decoupling" pin's of the TDA1541A, these must have very low leakage current. If electrolytic types were used this would result in massive distortion.
2) Tantalium cap's for "low frequency" power supply decoupling, because of lower ESR and excellent stability over time. (electrolytic capacitors will loose "capacity" over time, and leaking electrolytes may destroy the circuit board and cause erratic errors. Increased temperatures speeds up this process.
3) Ceramic SMD capacitors soldered directly at the IC pin's for "high frequency" decoupling. These capacitors are 1206 X7R types.
> Black gate capacitors seem to have a different construction, they use fine grained graphite particles instead of the electrolyte paste. They have lower ESR and are non-polar like film capacitors. I read somewhere these cap's were taken out of production?
For extra noise attenuation I plan to use Murata "3 pin" filters, composed of capacitors and ferrite beads.
Mains noise is suppressed using a multiple order filter (Schaffner PCB filters)
I will post some very interesting developments about feedback optimalization in both, I/V stages and OPA627 differential amplifier soon.
Hi, MGH
Thanks for your reply [post#263]
> There are 3 kind of decoupling cap's in the octal D-I DAC:
1) BC components foil capacitors on the 14 "decoupling" pin's of the TDA1541A, these must have very low leakage current. If electrolytic types were used this would result in massive distortion.
2) Tantalium cap's for "low frequency" power supply decoupling, because of lower ESR and excellent stability over time. (electrolytic capacitors will loose "capacity" over time, and leaking electrolytes may destroy the circuit board and cause erratic errors. Increased temperatures speeds up this process.
3) Ceramic SMD capacitors soldered directly at the IC pin's for "high frequency" decoupling. These capacitors are 1206 X7R types.
> Black gate capacitors seem to have a different construction, they use fine grained graphite particles instead of the electrolyte paste. They have lower ESR and are non-polar like film capacitors. I read somewhere these cap's were taken out of production?
For extra noise attenuation I plan to use Murata "3 pin" filters, composed of capacitors and ferrite beads.
Mains noise is suppressed using a multiple order filter (Schaffner PCB filters)
I will post some very interesting developments about feedback optimalization in both, I/V stages and OPA627 differential amplifier soon.
Thanks ecdesign.
Black Gates are still in production. What you heard was just a rumor. Some swear by BGs for power supply application - it's suppose to be very low noise. All the professional modders seem to use BGs over other electrolytics. But they are on the expensive side. You don't think BGs have any use in your octal D-I DAC?
I look foreward to hearing about your further developments.
Black Gates are still in production. What you heard was just a rumor. Some swear by BGs for power supply application - it's suppose to be very low noise. All the professional modders seem to use BGs over other electrolytics. But they are on the expensive side. You don't think BGs have any use in your octal D-I DAC?
I look foreward to hearing about your further developments.
Differential SPDIF interface schematics
Hi, tubee,
> In post [246] you asked information about the differential SPDIF interface I am using.
Well, I attached a schematic diagram (concept) for a differential SPDIF interface. This differential interface cancels-out interference and prevents ground-loops. It increases slewrate and reduces jitter at the source.
I connected the SPDIF input to the SPDIF output of the CD player, you can connect it to the signal pin of the TOSLINK transmitter or to the coax output. The circuit runs on 5V.
The interface increases slewrate, and creates a normal (+) and a inverted (-) SPDIF output signal, by using a RS422 transceiver chip (DS8922N). But any other RS422 chip will do. If you can't find a suitable RS422 transceiver chip, you can experiment with the alternative diagram.
I used standard SVHS cable and connectors (dual screened coax cable). with chassis connectors from Farnell.
The coil was made using an insulated philips ferrite ringcore (15mm diameter blue color coding, RK series), but you can experiment with other types. I used thin insulated equipment wire with 2 different colors, but you can also use enamelled copper wire. Both primary and secondary windings consist of 15 turns each. Make sure there is sufficient gap between primary and secondary winding to minimize interference. The transformer prevents ground loops.
Impedance is around 100R, you can reduce this by using 33R for R1 and R2 and 68R for R3.
Hi, tubee,
> In post [246] you asked information about the differential SPDIF interface I am using.
Well, I attached a schematic diagram (concept) for a differential SPDIF interface. This differential interface cancels-out interference and prevents ground-loops. It increases slewrate and reduces jitter at the source.
I connected the SPDIF input to the SPDIF output of the CD player, you can connect it to the signal pin of the TOSLINK transmitter or to the coax output. The circuit runs on 5V.
The interface increases slewrate, and creates a normal (+) and a inverted (-) SPDIF output signal, by using a RS422 transceiver chip (DS8922N). But any other RS422 chip will do. If you can't find a suitable RS422 transceiver chip, you can experiment with the alternative diagram.
I used standard SVHS cable and connectors (dual screened coax cable). with chassis connectors from Farnell.
The coil was made using an insulated philips ferrite ringcore (15mm diameter blue color coding, RK series), but you can experiment with other types. I used thin insulated equipment wire with 2 different colors, but you can also use enamelled copper wire. Both primary and secondary windings consist of 15 turns each. Make sure there is sufficient gap between primary and secondary winding to minimize interference. The transformer prevents ground loops.
Impedance is around 100R, you can reduce this by using 33R for R1 and R2 and 68R for R3.
Attachments
Feedback optimalization / project progress
Hi, all
I have done some interesting experiments with feedback optimalization. I started with the I/V stage, trying to get perfect squarewave steps without "ringing" or "sagging". I used a 21 KHz test signal in order to see steps on the oscilloscope, then I zoomed in on a few steps to show the tuning effect more clearly..
Upper left oscillogram, only a 590 Ohm resistor, "feedback servo" is to fast, ringing occurs. Harmonics are added to the sound, sound get's a bit noisy and harsh.
Center left oscillogram, 590 Ohm resistor with 100 pF capacitor in parallell, "feedback servo is optimal, no "ringing" or "sagging", sound is neutral, without anything added or filtered out.
Bottom left oscillogram, 590 Ohm resistor with 1nF capacitor in parallell, "feedback servo is too slow, "sagging" occurs, sound looses "speed" and detail.
Photo center, the I/V / differential amplifier module (one channel). It uses a OPA627 in DIP package, and 2 X OPA627 on a self-made adapter socket (SMD OPA627's get quite hot, so I mounted a Heatsink construction, just in case...). Tantalium and 100nF X7R SMD bypass cap's are used. Each module holds one channel, so each channel has it's own power supply.
Then I tried to get the same response from the OPA627 diff amp, as with the differential tube output stage by "feedback servo" tuning. After a lot of measuring and testing, I use 4 resistors of 1.96K, and 2 capacitors of 330pF (diagram is same as schematiciv1.jpg post#50). Now both output signals (cathode of the CF) and OPA627 output looked virtually identical,
Upper right oscillogram, differential OPA627 output stage signal
Bottom right oscillogram, differential tube output stage signal
As both oscillograms already indicate, the tuned OPA627 diff amp sound now comes very close to the differential triode diff amp sound, really impressive how sound quality can improve by simply changing some resistor and capacitor values.
I started routing the analog main board. When this PCB is ready and functions properly, the "core" of the octal D-I DAC is ready.
Analog main board modules:
2 X I/V Diff-amp module
2 X Quad DAC module
1 X Timing chain module
Analog main board inputs:
+8V, -8V, +20V, -20V, GND
I2S (BCK, DATA, WS)
Analog main board outputs:
Audio L, Audio R, GND
IL+, IL-, IR+, IR-, GND (can be used to drive the differential triode stage)
The analog main board makes all connections between modules, has a BCK buffer build-in, holds an optimized DEM reclocking system (176.4 KHz) and addittional decoupling. I will post some more information about the analog mainboard later.
Hi, all
I have done some interesting experiments with feedback optimalization. I started with the I/V stage, trying to get perfect squarewave steps without "ringing" or "sagging". I used a 21 KHz test signal in order to see steps on the oscilloscope, then I zoomed in on a few steps to show the tuning effect more clearly..
Upper left oscillogram, only a 590 Ohm resistor, "feedback servo" is to fast, ringing occurs. Harmonics are added to the sound, sound get's a bit noisy and harsh.
Center left oscillogram, 590 Ohm resistor with 100 pF capacitor in parallell, "feedback servo is optimal, no "ringing" or "sagging", sound is neutral, without anything added or filtered out.
Bottom left oscillogram, 590 Ohm resistor with 1nF capacitor in parallell, "feedback servo is too slow, "sagging" occurs, sound looses "speed" and detail.
Photo center, the I/V / differential amplifier module (one channel). It uses a OPA627 in DIP package, and 2 X OPA627 on a self-made adapter socket (SMD OPA627's get quite hot, so I mounted a Heatsink construction, just in case...). Tantalium and 100nF X7R SMD bypass cap's are used. Each module holds one channel, so each channel has it's own power supply.
Then I tried to get the same response from the OPA627 diff amp, as with the differential tube output stage by "feedback servo" tuning. After a lot of measuring and testing, I use 4 resistors of 1.96K, and 2 capacitors of 330pF (diagram is same as schematiciv1.jpg post#50). Now both output signals (cathode of the CF) and OPA627 output looked virtually identical,
Upper right oscillogram, differential OPA627 output stage signal
Bottom right oscillogram, differential tube output stage signal
As both oscillograms already indicate, the tuned OPA627 diff amp sound now comes very close to the differential triode diff amp sound, really impressive how sound quality can improve by simply changing some resistor and capacitor values.
I started routing the analog main board. When this PCB is ready and functions properly, the "core" of the octal D-I DAC is ready.
Analog main board modules:
2 X I/V Diff-amp module
2 X Quad DAC module
1 X Timing chain module
Analog main board inputs:
+8V, -8V, +20V, -20V, GND
I2S (BCK, DATA, WS)
Analog main board outputs:
Audio L, Audio R, GND
IL+, IL-, IR+, IR-, GND (can be used to drive the differential triode stage)
The analog main board makes all connections between modules, has a BCK buffer build-in, holds an optimized DEM reclocking system (176.4 KHz) and addittional decoupling. I will post some more information about the analog mainboard later.
Attachments
Mixed mode "magic" / analog mainboard
Hi, all
Yesterday, I redesigned the DEM clock synchronizing circuit [post#90]. This circuit makes sure that all D/A conversions start simultaneously and are synchronized with BCK. The aim of DEM clock synchronizing is to make sure all 8 TDA1541A's put their sample on their output simultaneously. When using 8 free running DEM clocks that are also not in sync with BCK, there will be varying delays. Again these delays are rather small, but in High-end equipment they can make quite a difference.
Basically I reduced the amplitude of the 176.4 KHz clock (BCK/16), and filtered it using 2 Murata filters. The TDA1541A datasheet indicates DEM (Dynamic Element Matching) clock of minimal 150 KHz, typical 200 KHz and maximal 250 KHz. So 176.4 KHz is within range. The lowered amplitude was then routed to a tiny ringcore transformer (similar as the one in post #268]. This transformer is used to isolate the digital GND noise from analog GND. The secondary winding is loaded with a 100 Ohm resistor and connected to analog ground. Then all 8 DAC pin's 16 were connected to the transformers output using 470pF capacitors.
After all modifications made to the prototype it was time for some more listening sessions on the main set (in the workshop I have smaller sonic resonators and a different control amplifier)
Modifications are: I/V optimizing, OPA627 diff amp optimizing and adding the DEM clock synchronizing circuit.
After switching-on, the sound of the OP-amp stage seemed a lot clearer, more open and more natural. Then I switched to the tube output stage, expecting not much difference, wrong! the tube stage still had it's characteristic sound, but was also significantly improved. The DEM clock synchronizing and tuning sure made a difference. Then I switched to mixed mode because I was curious about the sound, since the OP-amp stage was tuned to closely match the tube output stage waveform. This was something else, almost magic, it seemed both OP-amp stage and tube output stage completely "merged" producing a sound quality I never heared before during testing. The sound was crystal clear / pure, and increadibly "harmonic", instruments and voices sounded ever so realistic.
I also added the module lay-out of the analog mainboard. The analog mainboard is routed and I can start optimizing the tracks and adding ground planes. I will post some more information about the analog mainboard soon.
Hi, all
Yesterday, I redesigned the DEM clock synchronizing circuit [post#90]. This circuit makes sure that all D/A conversions start simultaneously and are synchronized with BCK. The aim of DEM clock synchronizing is to make sure all 8 TDA1541A's put their sample on their output simultaneously. When using 8 free running DEM clocks that are also not in sync with BCK, there will be varying delays. Again these delays are rather small, but in High-end equipment they can make quite a difference.
Basically I reduced the amplitude of the 176.4 KHz clock (BCK/16), and filtered it using 2 Murata filters. The TDA1541A datasheet indicates DEM (Dynamic Element Matching) clock of minimal 150 KHz, typical 200 KHz and maximal 250 KHz. So 176.4 KHz is within range. The lowered amplitude was then routed to a tiny ringcore transformer (similar as the one in post #268]. This transformer is used to isolate the digital GND noise from analog GND. The secondary winding is loaded with a 100 Ohm resistor and connected to analog ground. Then all 8 DAC pin's 16 were connected to the transformers output using 470pF capacitors.
After all modifications made to the prototype it was time for some more listening sessions on the main set (in the workshop I have smaller sonic resonators and a different control amplifier)
Modifications are: I/V optimizing, OPA627 diff amp optimizing and adding the DEM clock synchronizing circuit.
After switching-on, the sound of the OP-amp stage seemed a lot clearer, more open and more natural. Then I switched to the tube output stage, expecting not much difference, wrong! the tube stage still had it's characteristic sound, but was also significantly improved. The DEM clock synchronizing and tuning sure made a difference. Then I switched to mixed mode because I was curious about the sound, since the OP-amp stage was tuned to closely match the tube output stage waveform. This was something else, almost magic, it seemed both OP-amp stage and tube output stage completely "merged" producing a sound quality I never heared before during testing. The sound was crystal clear / pure, and increadibly "harmonic", instruments and voices sounded ever so realistic.
I also added the module lay-out of the analog mainboard. The analog mainboard is routed and I can start optimizing the tracks and adding ground planes. I will post some more information about the analog mainboard soon.
Attachments
Thanks for the update. Your descriptions of the sound are making us drool.
I am curious. I know you designed the DAC module to accept the TDA1541A directly into the PCB. But I would like to know if using a socket device for the TDA1541A would be OK - I may want to upgrade my TDA1541A to single crown version in the future. This way I can switch different versions of the chip on the fly without having to desolder the chips. Would a socket degrade the signal?
I am curious. I know you designed the DAC module to accept the TDA1541A directly into the PCB. But I would like to know if using a socket device for the TDA1541A would be OK - I may want to upgrade my TDA1541A to single crown version in the future. This way I can switch different versions of the chip on the fly without having to desolder the chips. Would a socket degrade the signal?
Analog mainboard schematics
Hi, MGH
Thanks for your reply [post#274]
I use high-quality turned pin IC sockets (Harwin, Augat or Aries). The low cost IC sockets with wipe contacts often make bad connection and can degrade the signal. I primarily use IC sockets to simplify updates, repairs and test the assembled board without the IC's (voltages) as each board has it's own power supply regulators, so a fault can be detected before destroying a IC. The module sockets are similar high-quality turned pin versions, power supply connections and output signals use 2 pin's in parallell, just in case...
I have added the schematic diagram of the analog mainboard. I use Murata LCL filters on every power supply input of every module (28 X). The diagram indicates 100pF for these, but I am planning tu use much higher values for the power supply lines. Digital and analog ground noise is separated by using the 2 ferrite beads of one of these filters (N29). The BCK signal is buffered using a 74HCT08, since it gave the cleanest signal, 74HCT244 buffer caused to much interference and "ringing" on BCK. The DEM clock circuitry uses a synchronous devider 74HCT161 and a small transformer to isolate the DEM clock signal. Filters N32, N33 and C10 filter the DEM clock signal. R2 and R3 attenuate the signal. The analog mainboard has it's own digital power supply for the BCK buffer and the DEM clock circuit. I used labels to indicate the interconnections (labels with the same name are interconnected).
J25 is the power supply connection, J26 the I2S input and J27 is the analog output (audio + I/V outputs)
Routing was quite a challenge, since I had to use manual routing. Today I plan to optimize the tracks and add ground planes. Can't wait to assemble
Hi, MGH
Thanks for your reply [post#274]
I use high-quality turned pin IC sockets (Harwin, Augat or Aries). The low cost IC sockets with wipe contacts often make bad connection and can degrade the signal. I primarily use IC sockets to simplify updates, repairs and test the assembled board without the IC's (voltages) as each board has it's own power supply regulators, so a fault can be detected before destroying a IC. The module sockets are similar high-quality turned pin versions, power supply connections and output signals use 2 pin's in parallell, just in case...
I have added the schematic diagram of the analog mainboard. I use Murata LCL filters on every power supply input of every module (28 X). The diagram indicates 100pF for these, but I am planning tu use much higher values for the power supply lines. Digital and analog ground noise is separated by using the 2 ferrite beads of one of these filters (N29). The BCK signal is buffered using a 74HCT08, since it gave the cleanest signal, 74HCT244 buffer caused to much interference and "ringing" on BCK. The DEM clock circuitry uses a synchronous devider 74HCT161 and a small transformer to isolate the DEM clock signal. Filters N32, N33 and C10 filter the DEM clock signal. R2 and R3 attenuate the signal. The analog mainboard has it's own digital power supply for the BCK buffer and the DEM clock circuit. I used labels to indicate the interconnections (labels with the same name are interconnected).
J25 is the power supply connection, J26 the I2S input and J27 is the analog output (audio + I/V outputs)
Routing was quite a challenge, since I had to use manual routing. Today I plan to optimize the tracks and add ground planes. Can't wait to assemble
Attachments
SVHS cable
Hi, real
Thanks for your reply,
> Yes, of course you're right. A shielded twisted pair (networking) cable would be better, especially when using long cables. In fact, the red "SVHS" cable on the prototype pictures is self made and uses high quality "van damme" balanced studio cable and Professional gold plated Nakamichi SVHS connectors (order code 804-1326) from Farnell.
However, tests with standard SVHS cables showed these work fine as well, provided high quality cable is used. The cable length of approx. 2m is also less critical.
Hi, real
Thanks for your reply,
> Yes, of course you're right. A shielded twisted pair (networking) cable would be better, especially when using long cables. In fact, the red "SVHS" cable on the prototype pictures is self made and uses high quality "van damme" balanced studio cable and Professional gold plated Nakamichi SVHS connectors (order code 804-1326) from Farnell.
However, tests with standard SVHS cables showed these work fine as well, provided high quality cable is used. The cable length of approx. 2m is also less critical.
dem reclock in CD304, D-I pcb
Hi Ecdesigns
*Will you pcb-design be as shown? Then its easy to pick the wanted sub-pcb's, someone can cut his sub pcb's off, or it could kept as one big pcb for who wants to build the D-I dac complete.
*What input receiver CS8412, or is CS8414 possible too?
Some off topic here but:
I want to take as much as possible the kwak's 11.2896Mhz as reference in my 304 CDP which i am modifiing now. To get dem signal i have to /4 clock, and then /16 with HCT161, then i have 176.4 kHz for dem reclock to pin 16 available.
Btw, i reclocked /2 to 1541 dac (HCT163) allready on kwak 7 board, and i am thinking to distribute clock 11mhz into Xin of SAA7210 too, besides the SAA7220. This could enhance sound again, because a cleaner clock signal is injected in 7210 then, normal the clock is redistibuted by the 7220 to 7210, a polluted signal i suppose. Planning to build 2 shunt ps's for 7210/20, but first things first.
So no Non-os for the 304mk2, want to keep dig-out (and don't want to spend it a CS8402 either)
Hi Ecdesigns
*Will you pcb-design be as shown? Then its easy to pick the wanted sub-pcb's, someone can cut his sub pcb's off, or it could kept as one big pcb for who wants to build the D-I dac complete.
*What input receiver CS8412, or is CS8414 possible too?
Some off topic here but:
I want to take as much as possible the kwak's 11.2896Mhz as reference in my 304 CDP which i am modifiing now. To get dem signal i have to /4 clock, and then /16 with HCT161, then i have 176.4 kHz for dem reclock to pin 16 available.
Btw, i reclocked /2 to 1541 dac (HCT163) allready on kwak 7 board, and i am thinking to distribute clock 11mhz into Xin of SAA7210 too, besides the SAA7220. This could enhance sound again, because a cleaner clock signal is injected in 7210 then, normal the clock is redistibuted by the 7220 to 7210, a polluted signal i suppose. Planning to build 2 shunt ps's for 7210/20, but first things first.
So no Non-os for the 304mk2, want to keep dig-out (and don't want to spend it a CS8402 either)