Thanks! That's what i wanted to know and hoped for ....
Higher DEM frequency requirers a smaller c value, as i understood.
Thanks,
Did you try a cap pot with a cap value Under the 510 pf ? I remember advises to lock it with ears and a low frequency wave rec to be played...
Are you pleased with this value and your unpopulated 14 decoupling caps ?
Would you advise it yet ?
No I did not try with a variable capacitor, good idea. I measured the DEM frequency with a few different fixed capacitors, and the frequency is roughly 88200/C [kHz, pF].
I am pleased with it sound, but to be honest I am a bit too lazy to remove the player from the system, disassemble, modify, reassemble, install back in place, listen, compare, .... So I am also interested in other DIYers experience with omitted filter capacitors.
That would require an external oscillator? I understood that no Cap at all did not work well.
If the 88200/c formula is correct, a 470pf (and even a bit lower) value would perform better. What I understood from the original text a ratio of at least 1:4 was required for the DEM to work well.
My understanding about this thread is those DEM value are near but you need really a pot cap to lock the signal...fixed cap value is close but not enough : so the precision trim pot...to find the value (is there a concistency chip to chip with the values ?)
Philips, Marantz and most used 470 pf as close value for 44/96 and today John's last is using as near value is 100 pf.... I surmise in any way clock signal is locked with precision in both case but by luck... But never understand at the reading of the thread what was the consequence ! Jitter ? Is it really hearable if you put e.g. 470 pf and the locked real value is 463 e.g. ?
Maybe a detail.... John went very deep with test and I assume most of us could have more hearable problems with input & output stages and layout... those last hidding the hearable DEM lock !
Philips, Marantz and most used 470 pf as close value for 44/96 and today John's last is using as near value is 100 pf.... I surmise in any way clock signal is locked with precision in both case but by luck... But never understand at the reading of the thread what was the consequence ! Jitter ? Is it really hearable if you put e.g. 470 pf and the locked real value is 463 e.g. ?
Maybe a detail.... John went very deep with test and I assume most of us could have more hearable problems with input & output stages and layout... those last hidding the hearable DEM lock !
I guess you're right saying that the DEM-clock would perform best with a locked signal. Only how a pot-cap must be trimmed to obtain that "locked" condition, is not quite clear to me, since DEM-clock might not be exact under each operating condition and "lock" might be lost....
Temperature drift, air humidity ?I have a TDA1541 non A : the DEM cap is in it if my understanding is correct... I don't know the value (surmise the famous 470 pf...) and if it can sound better with the same external DEM value with a TDA1541A ?!
I will try as I have some indivudual pipe connectors for vias (said to be far better than dip16 socket for inductance problems and contacts) maid with copper & beyrelium than Peufeu fellow advised in an another thread...
That's right. In the non-A's the cap is internal, in the A's it's external.
I have a TDA1541 non A : the DEM cap is in it if my understanding is correct... I don't know the value (surmise the famous 470 pf...) and if it can sound better with the same external DEM value with a TDA1541A ?!
I don't believe the internal cap of the non-a's was published.
Using the A version at least gives the opportunity to pull the pins 16 and 17 to -15Volts with a resistor.
ECDesigns said that he uses 900KHz oscillator frequency, well, unless I am wrong and there is some kind of clock multiplier on chip and the oscilator frequency and DEM frequency are not the same...
ECDesigns said:
On post 5098: http://www.diyaudio.com/forums/digital-line-level/79452-building-ultimate-nos-dac-using-tda1541a-510.html#post3949187
Last edited:
You could look at the DEM circuit as being an automatic calibration circuit for the DAC. It consists of a cycle of 4 steps.
The DAC is fed data at a certain rate. For each of these pieces of data the DAC ought to be calibrated by the DEM. Thus the frequency of the DEM ought to be 4 times higher than the datarate. The DEM frequency is governed by that cap... and the resistors suppress the noise on the sawtooth of the oscillator.
Quote:
Originally Posted by ECDesigns
I use a DEM frequency of around 900 KHz now (2 * 6K8 plus a silvered mica SMD 100pF timing cap) and 14 * 330nF 1206 size SMD film caps for filtering.
Unqte
Based on that and some experience of jameshillj /thanks/ I have at the moment 12 caps 1 uf , two big film/foil 4 uf at the MSB, four 6K8 resistors - I have two chips piggybacked- and two timing caps 120 pf. Sounds good for my ears - I have no scope and do not need one.
Cheers,
Originally Posted by ECDesigns
I use a DEM frequency of around 900 KHz now (2 * 6K8 plus a silvered mica SMD 100pF timing cap) and 14 * 330nF 1206 size SMD film caps for filtering.
Unqte
Based on that and some experience of jameshillj /thanks/ I have at the moment 12 caps 1 uf , two big film/foil 4 uf at the MSB, four 6K8 resistors - I have two chips piggybacked- and two timing caps 120 pf. Sounds good for my ears - I have no scope and do not need one.
Cheers,
Thank you too for the answer. I have a Tube-I-Zator SRPP output stage with current injection that is working very well, and I will make something like yours with D3a to compare with. I'm also planning a Kondo inspired stage with 6072s. We will see how wins the shoot out.
Vidar
Hi John,
About the Rohpoint resistor you advise for the I/V conversion : do you try also the Vishay naked VAR also and how do you benchmark it with each other for the SQ result ? Any experience please about that ?
A Mills could be a good enough swap or too noisy in relation to the SOTA Rohpoint ?
It's about sourcing, the Rohpoint is expensive to source when bought alone !
About the Rohpoint resistor you advise for the I/V conversion : do you try also the Vishay naked VAR also and how do you benchmark it with each other for the SQ result ? Any experience please about that ?
A Mills could be a good enough swap or too noisy in relation to the SOTA Rohpoint ?
It's about sourcing, the Rohpoint is expensive to source when bought alone !
I'm not John but this info might be useful to you ...
It's a bit cheaper to get a couple of single resistors from RS Components or Farnells - the individual resistors aren't too expensive from the Rhopoints company but there's a covering/service charge that makes it a crazy price for just a few resistors from them - it used to be about a 20GBP fee/charge and the resistors were about 4 GBP ea, not sure these days.
With the Neohm and Rhopoint resistors from both Farnell's and RS Components, it's definitely worth an enquiry to them about their final price as some of these components have been on their shelves for years and they're quite happy to be rid of them for a reduced price, and their pricing structure is all over the place - it's much the same with a lot of their slower more expensive items, just like a lot of retail/wholesale supply companies
TC2575 (naked Vishays) are a lot 'brighter' in comparison to say, the Rhopoints, and might suit your system if you're using 'softer' speakers/amp/etc - I found them to be a bit annoying as the extra detail was quite fatiguing over longer periods (couple of hours) but just my experience here - other people like them just fine.
It's pretty easy to just try out some of the less expensive resistors and see if that suits your requirements before spending 'big'.
It's a bit cheaper to get a couple of single resistors from RS Components or Farnells - the individual resistors aren't too expensive from the Rhopoints company but there's a covering/service charge that makes it a crazy price for just a few resistors from them - it used to be about a 20GBP fee/charge and the resistors were about 4 GBP ea, not sure these days.
With the Neohm and Rhopoint resistors from both Farnell's and RS Components, it's definitely worth an enquiry to them about their final price as some of these components have been on their shelves for years and they're quite happy to be rid of them for a reduced price, and their pricing structure is all over the place - it's much the same with a lot of their slower more expensive items, just like a lot of retail/wholesale supply companies
TC2575 (naked Vishays) are a lot 'brighter' in comparison to say, the Rhopoints, and might suit your system if you're using 'softer' speakers/amp/etc - I found them to be a bit annoying as the extra detail was quite fatiguing over longer periods (couple of hours) but just my experience here - other people like them just fine.
It's pretty easy to just try out some of the less expensive resistors and see if that suits your requirements before spending 'big'.
Hi Eldam,
DACs use / generate large bandwidth signals. Since the I/V resistor is tied directly to the TDA154x output, it is hit by large bandwidth RF signals. When the same resistor was only exposed to audio signals it would have neglible impact on perceived sound, only the resistor thermal noise could possibly cause some problems.
Similar, any active, non-linear component such as transistor, OP-amp, FET, tube connected directly to the output of a multibit DAC means big trouble in the form of inter modulation distortion. This is why I recently ditched all current conveyor circuits in my DAC designs and replaced these with a new I/V & output stage.
The inter action of this RF signal and resistor properties like stray inductance, stray capacitance, thermal noise, current noise (carbon resistors) can lead to a modified RF spectrum. This filtered RF spectrum can translate to changes in the audio spectrum and related sound quality.
Strictly speaking we need a good RF resistor for passive I/V conversion.
Wire wound resistors will always have some stray inductance (Ayrton Perry or bifilar winding) and are therefore less suitable.
Metal film resistors have helical grooves that cause stray inductance.
Viahay (naked) Bulk metal foil resistors should be ideal as RF resistor but I tested these often and was never satisfied with the "sound signature".
For 16 bits and 50R (TDA1541A, 4mA full scale current), LSB corresponds to 3uV. So it would be a good idea to choose a resistor that produces low enough thermal noise to resolve the LSB.
I had good results using multiple paralleled metal film resistors. For 50R passive I/V resistor one could parallel 2 * 100R, 3 * 150R or 10 * 500R for example.
Paralleling resistors reduces overall stray inductance, increases wattage (lower noise) and reduces Eddy current losses as one basically creates a “litz wire” resistor.
One could also try resistors that were specifically designed for RF applications:
http://uk.farnell.com/jsp/search/productdetail.jsp?SKU=1435937&MER=bn-me-pd-r3-cust-all-2
DACs use / generate large bandwidth signals. Since the I/V resistor is tied directly to the TDA154x output, it is hit by large bandwidth RF signals. When the same resistor was only exposed to audio signals it would have neglible impact on perceived sound, only the resistor thermal noise could possibly cause some problems.
Similar, any active, non-linear component such as transistor, OP-amp, FET, tube connected directly to the output of a multibit DAC means big trouble in the form of inter modulation distortion. This is why I recently ditched all current conveyor circuits in my DAC designs and replaced these with a new I/V & output stage.
The inter action of this RF signal and resistor properties like stray inductance, stray capacitance, thermal noise, current noise (carbon resistors) can lead to a modified RF spectrum. This filtered RF spectrum can translate to changes in the audio spectrum and related sound quality.
Strictly speaking we need a good RF resistor for passive I/V conversion.
Wire wound resistors will always have some stray inductance (Ayrton Perry or bifilar winding) and are therefore less suitable.
Metal film resistors have helical grooves that cause stray inductance.
Viahay (naked) Bulk metal foil resistors should be ideal as RF resistor but I tested these often and was never satisfied with the "sound signature".
For 16 bits and 50R (TDA1541A, 4mA full scale current), LSB corresponds to 3uV. So it would be a good idea to choose a resistor that produces low enough thermal noise to resolve the LSB.
I had good results using multiple paralleled metal film resistors. For 50R passive I/V resistor one could parallel 2 * 100R, 3 * 150R or 10 * 500R for example.
Paralleling resistors reduces overall stray inductance, increases wattage (lower noise) and reduces Eddy current losses as one basically creates a “litz wire” resistor.
One could also try resistors that were specifically designed for RF applications:
http://uk.farnell.com/jsp/search/productdetail.jsp?SKU=1435937&MER=bn-me-pd-r3-cust-all-2
Last edited:
Did I understand what you wrote (^) correctly when I conclude that feeding the output of the TDA directly into the inverting input of the opamp, is not the way to go? (Philips does in their players)
If I did not come to the correct conclusion, can you please point out where I did go wrong?