Building the ultimate NOS DAC using TDA1541A - Page 141 - diyAudio
Go Back   Home > Forums > Source & Line > Digital Line Level

Digital Line Level DACs, Digital Crossovers, Equalizers, etc.

Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Reply
 
Thread Tools Search this Thread
Old 9th May 2007, 08:07 AM   #1401
diyAudio Member
 
-ecdesigns-'s Avatar
 
Join Date: May 2006
Hi tubee,

Thanks for your reply [post #1394]

Resistor ladder DAC chips like the PCM56 use a voltage reference source that feeds a resistor ladder network. The output impedance of the ladder network is not constant, and varies with the bit switch settings. When a passive I/V resistor is connected to it, this will lead to audible bit errors. The PCM56 can only function correctly if it feeds into virtual ground (OP-amp). With TDA1541A / TDA1543 it's similar, violating the maximum allowable output compliance of 25mV produces higher bit errors than specified in the datasheet.

So I personally don't use passive I/V converters because I try to get lowest possible bit errors.

I use highly optimized the OP-amp I/V stages now, this includes external biasing (cascode). External biasing is not the same as increasing the (OP-amp) internal bias current. For I/V converters it's important to use fast OP-amps, operate them in the output voltage range where they produce lowest distortion, and keeping capacitive loads as low as possible.

If you accept bit errors by using passive I/V, and also add a tube output stage, some new problems occur:

- Tubes produce relatively high noise levels,
- The anode / filament supply can easily introduce hum
- Tubes have relatively high output impedance
- A coupling capacitor is often needed, causing sound coloration

The "sound" of a DAC chip relies heavily on the circuit it's used in. In order to achieve optimal results, extensive tuning is necessary.

DAC chips like the TDA1541A were probably never used optimally, because I didn't see a correct working DEM clock circuit yet. Using the TDA1541A internal DEM clock causes huge jitter, this can be easily measured (PIN 16 & 17). Jitter on the DEM clock causes incorrect active divider operation (divided currents are no longer accurately averaged / matched). Without a perfect working DEM clock, the TDA1541A produces fluctuating bit errors (bit errors vary with the DEM clock duty-cycle variations. The effect on sound quality is similar to jitter on the timing signal (BCK).

The TDA1541A-S2 is selected for lowest bit errors, probably the standard DEM clock was used during this selection process (so DEM clock wasn't optimal). A possible way to still achieve low bit errors with a bad DEM clock is that the active dividers already have very high initial accuracy.

Another problem is that DEM clock interference finds it's way to the TDA1541A output (crosstalk), when using a free running DEM clock, this will introduce intermodulation distortion. Therefore it's important to synchronize the DEM clock with BCK. In order to complete a full DEM cycle (DEM switch matrix has 4 switch settings) within one sample, optimal DEM clock (with a NOS DAC) equals 44100 * 4 = 176.4 KHz (DEM clock frequency is specified between approx. 150...250 KHz).

With oversampling (4X) this means that there isn't enough time to complete a full DEM cycle before the next sample has to be output. So bit errors / fluctuations might be slightly higher now.

I have tried to find the reason for the natural sound quality of NOS DACs. By experimenting with band-limiting (brickwall filtering) signals from tape decks and record players, it seems like the band-limiting itself already produces a synthetic, unnatural sound. Further experiments show that a NOS DAC (because it's not band-limited like a OS DAC) is capable of producing faster signal transients (similar like with the tape deck and record player that are also not band-limited to 20 KHz). This might well be one of the reasons why NOS DACs produce a sound that appears to be more natural.

Next "problem" with digital filters is that they replace the original samples with calculated ones. Each sample leaving a digital interpolation filter is a result of a averaging / calculation process using 100+ samples. So the outcome of this calculation process is entirely responsible for the sound quality. The slightest trunctuations can completely change the sound spectrum. Basically, when listening to a OS DAC, one is listening to the results of calculations, not the original samples.

An example, suppose we average between 100 samples and 5 of them have a slightly different value (few bit difference), it's very likely that these relatively small fluctuations are completely "filtered" out by averaging / trunctuation process. So basically, micro-information present in the original samples could be simply removed this way.

A NOS DAC will re-produce this micro-information, but it's likely to disappear below the noise level caused by quantization noise (16 bit 44.1 KHz), as the (time) resolution is rather low.

The DI DACs change this, by significantly increasing both time, and analog resolution, quantisizing error amplitude will drop significantly. Now all micro-information is lifted above the noise floor and becomes clearly audible. Because the DI DACs only interpolate between few samples, and amplitude "calculations" are performed by a analog circuit (multiple DAC chips), no trunctuation errors occur. In fact this system even increases bit resolution by providing additional analog values.

The DI DAC analog output signal now has both, Higher amplitude resolution (19 or 20 bits) and higher time resolution (8 or 16 times higher). So a DI 8(M) ends up with approx. 64x resolution enhancement, the DI 16 with 256x resolution enhancement.

By comparison the 4x oversampled signal from a SAA7220 / TDA1541A combination only provides 4x resolution enhancement!

Another problem with OS DACs is that they shift the mirror images up the frequency spectrum (with a NOS DAC they are located relatively close to the audio range). By doing so, the HF signal / HF energy has to pass both, the analog part of the D/A converter AND the I/V stage / analog output stage, before it enters the analog corrective filter. Any non-linearity will now introduce errors. This might also be the reason for the OS DAC high jitter sensitivity (frequency modulated HF images).

The DI DACs interpolate as well, they also produce higher virtual sample rates, but each DAC chip still runs in NOS mode, the mirror images at higher frequencies have less power now (each image has less power than the fundamental image). The jitter sensitivity of a DI 8M is much lower than a OS DAC with 8x oversampling, this seems to indicate that the jitter "problem" is not merely caused by slight sample timing errors, but rather by analog circuit disruption, caused by the frequency modulated HF images. The FM HF images introduce problems with non-linear components / circuits.

OS DACs have pre-ringing, this almost never occurs with real-life sounds, so we might perceive it as unnatural sounding. It could also limit dynamics.

DI DACs have zero pre- and post ringing. The impulse response is totally flat. Dynamics are optimal.
  Reply With Quote
Old 9th May 2007, 08:46 AM   #1402
oshifis is offline oshifis  Hungary
diyAudio Member
 
Join Date: Mar 2004
Location: Budapest, Hungary
Quote:
Originally posted by -ecdesigns-
DAC chips like the TDA1541A were probably never used optimally, because I didn't see a correct working DEM clock circuit yet. Using the TDA1541A internal DEM clock causes huge jitter, this can be easily measured (PIN 16 & 17). Jitter on the DEM clock causes incorrect active divider operation (divided currents are no longer accurately averaged / matched). Without a perfect working DEM clock, the TDA1541A produces fluctuating bit errors (bit errors vary with the DEM clock duty-cycle variations. The effect on sound quality is similar to jitter on the timing signal (BCK).
Hi -ecdesigns-

I intend to use an external 176.4 kHz clock here, divided from the low-jitter master clock. Which pin should I use for the input, pin 16 or pin 17, and do I need the other pin left open? Can I use TTL level signal?

BTW, if you spent so much effort in the digital part, you should try my transformer/tube I/V circuit. It sounds much better than the OPA627 that I used before.

Laszlo's Valve Output Stage with Lundahl transformer

Regards,

Laszlo
  Reply With Quote
Old 9th May 2007, 10:39 AM   #1403
tubee is offline tubee  Netherlands
diyAudio Member
 
tubee's Avatar
 
Join Date: Jul 2004
Location: Netherlands
Hi John

Thanks for explaining of your TDA choice and tips for PCM experiments. Will read it over thourough.
My first comment now is: I first did dem in the simple way: divided with 74/163 to 176Khz, no extra components, only a coupling cap to pin 16. Now i removed Dem reclock, replaced a 470pF Philips KP cap between 16-17, because i sometimes heard artefacts in treble. I should use your transformer coupled dem schematic with bead filtering.

The sound without dem divider: a little warmer (more jitter?) and maybe a slightly more diffuse stereo image, that's all. But the artefacts are gone.

"
- Tubes produce relatively high noise levels,
- The anode / filament supply can easily introduce hum
- Tubes have relatively high output impedance
- A coupling capacitor is often needed, causing sound coloration "

I made a regulated PS for f-f
Have some Ropel 3.3 uF MKP caps.
Output impedance is not important to me, the stage has to drive a 50cm cable to tubed pre-amp.
Only the noise can be some concearn.
__________________
Daisy Bell
  Reply With Quote
Old 9th May 2007, 11:07 AM   #1404
adhoc is offline adhoc  Australia
diyAudio Member
 
adhoc's Avatar
 
Join Date: Jul 2005
Location: Australia
Hello EC,

A bit of a late reply, but I have generated some PDFs off you kit configurator - what does "PAL" stand for?

Thanks!
__________________
jitter critter.
  Reply With Quote
Old 9th May 2007, 12:04 PM   #1405
diyAudio Member
 
Bernhard's Avatar
 
Join Date: Apr 2002
Location: Munich
Quote:
Originally posted by -ecdesigns-


Resistor ladder DAC chips like the PCM56 use a voltage reference source that feeds a resistor ladder network. The output impedance of the ladder network is not constant, and varies with the bit switch settings. When a passive I/V resistor is connected to it, this will lead to audible bit errors. The PCM56 can only function correctly if it feeds into virtual ground (OP-amp).
I can't agree with that.

With the PCM56, the passive I/V works very good up to > 500 ohms.
  Reply With Quote
Old 9th May 2007, 01:24 PM   #1406
tubee is offline tubee  Netherlands
diyAudio Member
 
tubee's Avatar
 
Join Date: Jul 2004
Location: Netherlands
Quote:
Originally posted by Bernhard


I can't agree with that.

With the PCM56, the passive I/V works very good up to > 500 ohms.
I guess the lower the I/V resistor, the lower the distorsion. but lower output voltages has to be amplified more, it's a matter of the right choices. Triodedick advises 100 ohm for I/V res.

(btw i don't know where the zobel is for)
Attached Images
File Type: gif dac1.gif (8.1 KB, 1149 views)
__________________
Daisy Bell
  Reply With Quote
Old 9th May 2007, 02:32 PM   #1407
diyAudio Member
 
Bernhard's Avatar
 
Join Date: Apr 2002
Location: Munich
Zobel ?

I don't know if the analog filter in front of the I/V resistor is a good idea.
  Reply With Quote
Old 9th May 2007, 02:53 PM   #1408
diyAudio Member
 
Bernhard's Avatar
 
Join Date: Apr 2002
Location: Munich
Default Re: Re: Re: Re: DAC chips

Quote:
Originally posted by tubee


Don't know what selection criteria you handled at that time Bernard. I purchased them together with a CD304, remember? (june 2005??)

Almost two years, at that time I selected chips, so should be good ones.
Did they have any stickers/labels ?
  Reply With Quote
Old 9th May 2007, 03:35 PM   #1409
djn111 is offline djn111  Netherlands
diyAudio Member
 
Join Date: Sep 2006
Default tda541a where to buy

Hello,

For those looking for a tda1541a source check the following site:

http://www.diyparadiso.com/

got the sales or "uitverkoop" part, tda1541a 10 euro each.

Keep on building
  Reply With Quote
Old 9th May 2007, 06:35 PM   #1410
diyAudio Member
 
-ecdesigns-'s Avatar
 
Join Date: May 2006
Hi oshifis,

Thanks for your reply [post 1402]


You need to use a synchronous divider like the 74HCT161, if you need a higer division factor, multiple 74HCT161 chips can be cascaded (examples can be found in the datasheet).

Now make a voltage divider by connecting 2 x 100 Ohm resistor in series between 176.4 KHz output and AGND, creating a 2.5Vpp signal. Now connect a 100pF capacitor between the 2.5Vpp tap and pin 16 of the TDA1541A. Use a high quality polypropylene (WIMA) or polystyrene capacitor (LCR components).

Connect a 100nF film bypass capacitor (WIMA or BC components) between pin 17 and AGND.

The upper oscillogram shows the typical "datasheet" DEM clock, needless to say it contains jitter.

The lower oscillogram shows the synchronized low jitter DEM clock. This is how the output signal on pin 16 should look like (you may need to experiment with both signal amplitude (2.5V) and capacitor value (100pF).


No offense but I already have a hybrid output stage (tube OP-amp plus semiconductor Op-amp in parallel). I am convinced that it's able to outperform both a semiconductor-only or tube-only output stage. It has new sonic properties, unlike typical tube or semiconductor circuits.

I feed both tube Op-amp and semiconductor OP-amp with a OP-amp I/V converter, so I fully meet TDA1541A output compliance, achieving very low non-linear distortion.

I avoid using transformers and capacitors whenever possible, they are non-linear devices that can easily colour / smear the sound. I use DC coupled circuits whenever possible.

I no longer use the OPA627, I can imagine that a good tube output stage could outperform it. The OPA627 sound darks / laid-back, and veiled. It can't achieve the LM4562 transparency, even when forced into class A.

I now use the LM4562 "sweet spot" for lowest THD, and added a external tuning circuit that significantly reduces crossover distortion, and stabilizes the negative feedback loop. I also kept capacitive load very low (settling time). Used this way, the LM4562 is hard to beat.

Don't forget, my aim is not to achieve a nice sounding DAC.

My aim is to achieve very realistic "live" performance from my audio set. I use sonic resonators (point source) to achieve this. These have a special mechanism to re-construct a 3D sound image, using only 2 channels.

This requires staggering performance from both DAC and power amplifier, and makes DAC design much more difficult. Playing both high-quality live recordings and Chesky audio recordings, it seems I almost reached that aim.

It's not possible to obtain this effect with (bad) studio recordings, it only works with good live recordings that are not tampered with too much.

This more or less proves that the system works.
Attached Images
File Type: jpg demclock1.jpg (47.7 KB, 1011 views)
  Reply With Quote

Reply


Hide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off



New To Site? Need Help?

All times are GMT. The time now is 10:58 AM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio

Content Relevant URLs by vBSEO 3.3.2