The datasheet specs and performances are measured with very basic setups, You can easily get 5 db lower noise if you have better clocks and regulation on the dac.
the data sheet just represent normal functionning with average components...
Take the mosfet for example, in proper circuits it can behave better than any tubes in current gain mode .
the data sheet just represent normal functionning with average components...
Take the mosfet for example, in proper circuits it can behave better than any tubes in current gain mode .
My last adventure...
Hi guys,
Yesterday I was bored on my job (24hr ward
) so I decided to build the simplest and cheapest DIDAC that I could and let it on my job to entertain my humble self: I used a SMPS from a printer that Kaput; I used one of DDDAC's USB receivers for which I had to solder first the PCM 2707 chip that I have previously destroyed due to a short (It is hard to solder it with a normal iron, believe me); I used RCA connectors from a broken DVD that someone let around because I ran out of stock of them; I used a 5V TeddyReg that I had lying around because I am upgrading to the new SuperteddyRegs...
All cheap as I am.
Passive I/V with common type R (50ppm) and Vref inspired on -EC-' but cheaper:
Picasa Web Albums - mauricio
It resulted very small and sounds so good that I will continue to tweak it. maybe a baloon after the SMPS and honeycomb I/V resistors.
I installed the I2S attenuators and the DJA...
The sound is more immediate than the ones with active output and the bass is less pronounced but very natural and with good bass harmonics which I discovered when I returned home and used my system:
Picasa Web Albums - mauricio
I can imagine how good does the SDplayer sound.
Cheers,
M.
Hi guys,
Yesterday I was bored on my job (24hr ward
) so I decided to build the simplest and cheapest DIDAC that I could and let it on my job to entertain my humble self: I used a SMPS from a printer that Kaput; I used one of DDDAC's USB receivers for which I had to solder first the PCM 2707 chip that I have previously destroyed due to a short (It is hard to solder it with a normal iron, believe me); I used RCA connectors from a broken DVD that someone let around because I ran out of stock of them; I used a 5V TeddyReg that I had lying around because I am upgrading to the new SuperteddyRegs... All cheap as I am.
Passive I/V with common type R (50ppm) and Vref inspired on -EC-' but cheaper:
Picasa Web Albums - mauricio
It resulted very small and sounds so good that I will continue to tweak it. maybe a baloon after the SMPS and honeycomb I/V resistors.
I installed the I2S attenuators and the DJA...
The sound is more immediate than the ones with active output and the bass is less pronounced but very natural and with good bass harmonics which I discovered when I returned home and used my system:
Picasa Web Albums - mauricio
I can imagine how good does the SDplayer sound.
Cheers,
M.
2xTDA1541A NOS Dac Distortion Problem
Hi All,
After reading this thread and some other internet documents i decided to build a 2xTDA1541 parallel DAC, similar to Lesha's DAC. The DAC consists of CS8414 , 2xTDA1541 ( with 100nF decoupling caps ) and for the output buffer there are OPA627 and AD843opamps . It is not a differential dac it's simply a parallel dac.
It tooks me 2 weeks to finish it . When it was complete i have started some listening sessions . The sound was impressive , i could compare it with some hi-end dacs such as Benchmark DAC1 and Reimyo DAP 777 , its sound was fairly good enough . Actually the main difference is that the TDA1541 have a bold sound . It has a deep and tight bass but the highs are a bit rolled off. I was thinking that i have built a good dac until i analyzed its output on the oscilloscope . I have tested the dac using some sine waves starting from 1Khz up to 15Khz . Until 8 Khz it looks fine but after that you have a distorted sine waves . Actually you can see on the scope what you listen . The sound is bold and the signal is bold too. When you magnify this bold signal you can see the distortion . I thought that two TDA1541 are not very well synchronised and there is a timing difference in between. So i have disabled one of it . Single TDA1541 is better but you can still see some distortion above 10Khz. I can't imagine what can cause this distortion . Does anybody has an idea ?
Thanks
Reha
Hi All,
After reading this thread and some other internet documents i decided to build a 2xTDA1541 parallel DAC, similar to Lesha's DAC. The DAC consists of CS8414 , 2xTDA1541 ( with 100nF decoupling caps ) and for the output buffer there are OPA627 and AD843opamps . It is not a differential dac it's simply a parallel dac.
It tooks me 2 weeks to finish it . When it was complete i have started some listening sessions . The sound was impressive , i could compare it with some hi-end dacs such as Benchmark DAC1 and Reimyo DAP 777 , its sound was fairly good enough . Actually the main difference is that the TDA1541 have a bold sound . It has a deep and tight bass but the highs are a bit rolled off. I was thinking that i have built a good dac until i analyzed its output on the oscilloscope . I have tested the dac using some sine waves starting from 1Khz up to 15Khz . Until 8 Khz it looks fine but after that you have a distorted sine waves . Actually you can see on the scope what you listen . The sound is bold and the signal is bold too. When you magnify this bold signal you can see the distortion . I thought that two TDA1541 are not very well synchronised and there is a timing difference in between. So i have disabled one of it . Single TDA1541 is better but you can still see some distortion above 10Khz. I can't imagine what can cause this distortion . Does anybody has an idea ?
Thanks
Reha
15r is too low i think . Because even with 33 TDA gets hot ( not too much but quite warm lets say )
Are you sure that you are seeing distortions? If your dac is NOS, you will see stepped sine waves above 1 kHz unless you apply steep filtering above 20kHz.
regards
Martti
I am not sure whether it's distortion or signals above 20khz . The signal shape just at the output of the TDA1541 is not like a sine wave , especially after 8Khz.
There are two opamps in the I/V stage . At the first stage there is a 1nF filter cap in parallel with the opamp feedback resistor , in the second stage there is another 470pF filter cap in parallel with the feedback resistor. So the filter is not a steep filter.
Should i increase the caps values ?
There are two opamps in the I/V stage . At the first stage there is a 1nF filter cap in parallel with the opamp feedback resistor , in the second stage there is another 470pF filter cap in parallel with the feedback resistor. So the filter is not a steep filter.
Should i increase the caps values ?
Last edited:
I think i understand the situation. TDA1541 is a quite primitive dac chip when running in NOS mode . TDA1541 is just adding binary levels to get the original signal ( R-2R Dac behaviour ) . According the Shannon's law theoritecally sampling two times of the maximum frequency of the original signal is enough but this is not the case in practical situation. 44.1 Khz is enough for low frequencies but not enough for above 8 Khz . You simply get stepping square signal for high frequencies not a sine wave . This is the way what TDA1541 is doing nothing else . So I think Martti is right .
Last edited:
I would really like to know who was the first to spread that nonsense around so that others pick it up and repeat it.
1) If already the static sine wave looks bad, what would music look like ?
2) What higher V/uSec transients and harmonics than 22 kHz ?
Bernhard said:
Bernhard, why do you think that the ability of a DAC to reproduce a perfect sine wave up to 22 kHz is the most important feature of it?
For example - square wave has infinite harmonic spectra and if you try to filter it at 22 kHz you will end with rather round result. The TDA1541A DAC is specified at 8x oversampling which means that it's settling time is at most 1 / 8x44.1 = 2.8 us or you could produce with this DAC harmonics up to 176 kHz and beyound...
On the other hand despite you can't hear over 15-17kHz sinewaves you can surely feel (like nuances in music) presence or absence of frequencies even > 37 kHz (tried and confirmed). Call it "intermodulation between frequencies" if you want. It is similar with the frequencies < 25 Hz on mighty system which you can't hear but you can feel them with your body and psyche.
John tried to explain you - music and sound are not so simple as we (and the engineers) want it to be. John's development is based more on HEARING and FEELING rather than on technical asumptions rised from simplified models... And I could confirm this. I respect very much his aproach - first to do the things and after that to seek for reasons and explanations.
If you have an opportunity to listen to some true high-end systems which are able to make you goose-flesh by emotions - do it and make some experiments with them. You surely will understand what I am talking about.
Greeting to you with "Blinded by science" by Foreigner
(please, don't accept this personally...)
Last edited:
Hi John,
I replaced the LM336-2.5 Vref from my D1 trasimpedance amp for a LED based one, inspired on your own but not an exact copy because I ran out of 10mH inductors and, with my little knowledge, I only mannaged to have 3.5V by using 3 series red LEDs of which one is in series with the Vref (the others to ground). Even if the red LED is contributing with a few nV of noise, the sound became softer, cleaner, with more depth of soundstage and nice bass. Compared to passive I/V, the sound has more presence of bass midbass but is kind of Hi Fi, with the passive being more immediate and with natural harmonics; I have the sensation that passive lets listen to what is on the software better, with no artifacts.
I think if one has a bass weak system it would be preferable to go active and vice/versa.
This leds me to a question derived from my ignorance. In your Vref circuit you use the same values for L and C: wouldn't it be better to mix different values for L and C to avoid some hypothetical resonance???
Cheers,
M.
I replaced the LM336-2.5 Vref from my D1 trasimpedance amp for a LED based one, inspired on your own but not an exact copy because I ran out of 10mH inductors and, with my little knowledge, I only mannaged to have 3.5V by using 3 series red LEDs of which one is in series with the Vref (the others to ground). Even if the red LED is contributing with a few nV of noise, the sound became softer, cleaner, with more depth of soundstage and nice bass. Compared to passive I/V, the sound has more presence of bass midbass but is kind of Hi Fi, with the passive being more immediate and with natural harmonics; I have the sensation that passive lets listen to what is on the software better, with no artifacts.
I think if one has a bass weak system it would be preferable to go active and vice/versa.
This leds me to a question derived from my ignorance. In your Vref circuit you use the same values for L and C: wouldn't it be better to mix different values for L and C to avoid some hypothetical resonance???
Cheers,
M.
LC filters
Hi,
I've done it a long time ago. With the simulator and mixing some values you can optimize the FFT response trough a great bandwidth. I mean, lower the transient response of the armonics. You have to calculate or measure the current flowing to your cuircuit and then optimize the order and the values of the LC filter.
Best regards,
Hi,
I've done it a long time ago. With the simulator and mixing some values you can optimize the FFT response trough a great bandwidth. I mean, lower the transient response of the armonics. You have to calculate or measure the current flowing to your cuircuit and then optimize the order and the values of the LC filter.
Best regards,
LC filter
Wich it means that you don't have to add a lot of L-C elements to tune your 2.5 mA consumption circuit. You can start with a simple C-L-C filter, and runing some stepped simulations changing the value parameter of the L element basically. If you change the C element capacity, you will have more or less ripple atenuation, but the most important is the L value. You can run from 1 to 10mH and you'll see how much it changes, you'll be surprised. Then you have to look both transient response and bandwidth one too.
Regards,
Wich it means that you don't have to add a lot of L-C elements to tune your 2.5 mA consumption circuit. You can start with a simple C-L-C filter, and runing some stepped simulations changing the value parameter of the L element basically. If you change the C element capacity, you will have more or less ripple atenuation, but the most important is the L value. You can run from 1 to 10mH and you'll see how much it changes, you'll be surprised. Then you have to look both transient response and bandwidth one too.
Regards,
Because otherwise it will sound dispersed.
I have the choice between unfiltered , partially filtered, or brickwall filter, and from listening I clearly chose the brickwall.
Perhaps people have different hearing but the problem might be that they do not have the choice.
The fact is that there is no information beyound 20 k contained on a CD, so nothing to reproduce either. No more than a 20 k sine or the first harmonic of a 10 k sine or the 20th harmonic of a 1k square.
So if there is nothing beyond 20 k, you can cut it all and it is beneficial to correctly reconstruct the signal that is on the CD.
Again, there are no such frequencies on the CD originally.
By not filtering, you create every kind of aliasing and intermodulation garbage that was not there.
The price for getting some signal > 20k is paied by what is < 20k.
If you are worried about what happens > 20k, every power amplifier will add some harmonics, so > 20k will not be dead even with a brickwall filter.
Really ?
Okay, look at John's speakers. The HF response of such dome tweeters is very limited, so the filtering is done partially by the tweeters. What was it all for ?
Plus an unfiltered nonos DAC will sound different on other speakers that have a wider HF bandwith.
Do you like metal dome tweeters?
In Munich we have High End fair every year.