All voltages are good after start (8.06V / 3.3V, 2.4 at pin 20...), but when I play music, the left channel becomes crazy : first time the 3.3V drops, next start it's the 8V dropping...
I guess there's something wrong appearing when the chip is on duty, like a short or something, putting the Bellesons on their knees, thus the voltage drops.
Several working regs swapped and it's the same.
Anybody has an other idea, or you also think the chip is dead ?
I guess there's something wrong appearing when the chip is on duty, like a short or something, putting the Bellesons on their knees, thus the voltage drops.
Several working regs swapped and it's the same.
Anybody has an other idea, or you also think the chip is dead ?
LM334 Tempco CCS results
I have about 150 hours of play time on LM334 Tempco CCS boards.
Bass and midrange are more prominent but musical detail and treble is compressed and smeared in an unnatural sounding way. Not musical at all. I prefer the 6K resistor over this.
Unless someone tells me that they have great sounding results with the LM334 on pin 20 I'm going to pull the plug on this CCS device and try the LT3092 next.
I have about 150 hours of play time on LM334 Tempco CCS boards.
Bass and midrange are more prominent but musical detail and treble is compressed and smeared in an unnatural sounding way. Not musical at all. I prefer the 6K resistor over this.
Unless someone tells me that they have great sounding results with the LM334 on pin 20 I'm going to pull the plug on this CCS device and try the LT3092 next.
Bad news carlsor, it was too easy to be true ?
Could you remind me the best sounding system you mount on pin20 ?
Could you remind me the best sounding system you mount on pin20 ?
I found the LM334 tempco circuit to have a stable current within +/-1% over the 25-40C range. Too bad it is unlistenable to me.
My current favorite CCS is the single 2SK170 JFET. The cascoded 2SK170 JFET also sounds very clean, dynamic, and detailed but is a bit more analytical sounding. To me the single 2SK170 contributes more musicality, realism, and emotion to the DAC than the cascoded version. This difference is small and others may prefer the cascoded JFET sound if their system is already warm or bright sounding. The downside of any JFET CCS is that it changes current with temperature.
My current favorite CCS is the single 2SK170 JFET. The cascoded 2SK170 JFET also sounds very clean, dynamic, and detailed but is a bit more analytical sounding. To me the single 2SK170 contributes more musicality, realism, and emotion to the DAC than the cascoded version. This difference is small and others may prefer the cascoded JFET sound if their system is already warm or bright sounding. The downside of any JFET CCS is that it changes current with temperature.
I've had the lm334 with zero tempco in for a couple of weeks, but I've also recently changed amps and a few other bits. I was thinking yesterday that I was missing some of the magic sparkle and holographic imaging at the top end and maybe the output valves in my amps were old and tired.
Your post got me thinking, so I swapped back to the 954-E301 diodes (300uA rated, but actually giving 360uA) and that lovely, gentle detail is back.
dwjames,
The LM334 sucks the details out of music and blurs the treble. I switched back to the single JFET CCS and everything is wonderful again. Internally the 954-E301 diode is a JFET with the G-S connected so it should sound similar to the 2SK170.
"lovely, gentle detail" is a phrase that comes to mind when the last change in my stereo system has fully burned in. Have you switched the buffer in and out of circuit to hear the difference it makes since removing the LM334?
The LM334 sucks the details out of music and blurs the treble. I switched back to the single JFET CCS and everything is wonderful again. Internally the 954-E301 diode is a JFET with the G-S connected so it should sound similar to the 2SK170.
"lovely, gentle detail" is a phrase that comes to mind when the last change in my stereo system has fully burned in. Have you switched the buffer in and out of circuit to hear the difference it makes since removing the LM334?
dwjames,
How many of those did you have to test to get one that works? They are not cheap but then not audiophile expensive either.
Can't beat that for simplicity!
How many of those did you have to test to get one that works? They are not cheap but then not audiophile expensive either.
Can't beat that for simplicity!
Did you find one you could use?
How many did you buy?
AND most importantly, did you like what you heard?
THANKS!
Yea the diodes sounded good but the values are so inconsistent I think bulk buying to match them up could be unfruitful
It affects the bias voltage and vcom voltage so matching settings on the dac chip will be impossible
I have the 3 leg lm334 no temp coefficient
I didn't notice a huge drop in sq But my testing is not like with like values so it's inconclusive
I just bought 2 x 300uA and 100uA so I could use them parallel and get 400uA together. This was before I realised how far out they could be!
So my 2 x e-301 measured 365uA and 369uA when tested with a 3.3v power supply. The 2 x e-101 are 105uA and 155uA.
So I don't know whether it's luck that my 2 x e-301 are around the same and fairly close to 400uA, but for now I'm just using those and I'm happy with the sound.
Last night my friend came round and I got to have my first proper comparative listen with the buffer stage now it's had over 300 hours of burn in. I've got switches set into the board so I can instantly swap between the standard dddac outputs into the cinemag transformers and the dddac outputs into jfet buffer stage, then into the cinemag transformers.
I'm still trying to digest the results properly, but in essence; with the buffer stage in the mix, the sound is fuller, richer, chunkier, more luscious and with more authority than without. It's definitely very musical and enjoyable, but it depends on the music and the recording as to how this translates and I guess it comes down to a matter of personal taste. Most recordings really benefit from this treatment, but some I think I might prefer in their natural state. I'm not sure, I need to live with it some more. I need to hear more to get my head clearly around the sound and how it fits with my system and my musical tastes.
I'd be keen at this stage to see what others reckon, either if people wanted to hook up for a listening session one mid-week evening (near Hereford) or if that's not practical, I'd be happy if someone else with a single deck dddac and cinemags wanted to borrow the buffer and power supply for their own explorations.
One thing's for sure though, at this stage in the game, and with my new Sun Audio amps singing away, it's all sounding pretty damn amazing and I'm getting carried away and lost in a world of amazing music more often than not these days and for that I'm thankful and happy 🙂
J
Nigel,
Forgive me if you have made this plain and I missed it but are you using your brilliant computer power supply scheme with the DDDAC?
Seems to me it would be just as worthwhile here.
Just ordered some of those little regulators from Hong Kong. Seem to be better than decent and at a price one cannot do it for themselves.
The wait is about a month so I will have to be patient.
Nigel's scheme, for those who have not read it, is to use the A123 batteries as a giant LOW ESR capacitor trickle charged by a decent regulated power supply. He does not see (hear?) a need for anything extravagant. Set the "charger" for the voltage needed and let it go! No danger of over-charging and since the charging is constant no worries about battery discharge.
It is the definition of an elegant concept. And he is much too modest about it.
Forgive me if you have made this plain and I missed it but are you using your brilliant computer power supply scheme with the DDDAC?
Seems to me it would be just as worthwhile here.
Just ordered some of those little regulators from Hong Kong. Seem to be better than decent and at a price one cannot do it for themselves.
The wait is about a month so I will have to be patient.
Nigel's scheme, for those who have not read it, is to use the A123 batteries as a giant LOW ESR capacitor trickle charged by a decent regulated power supply. He does not see (hear?) a need for anything extravagant. Set the "charger" for the voltage needed and let it go! No danger of over-charging and since the charging is constant no worries about battery discharge.
It is the definition of an elegant concept. And he is much too modest about it.
Too Many Upgrades?
dwjames - we have both gone down the road of fully tweaking a single board DDDAC. Unlike you, I did the buffer first instead of last which made a huge improvement in detail and dynamics at the time. I suspect that the Shunt PS on the Analog section would have provided most of this improvement anyway. The only remaining upgrade for me is the 2-chip 1/2 clock circuit for which I have ordered all the parts.
It seems that every upgrade provided more detail and dynamics - regardless of the order of improvements. At some point there could be a point of too much detail and dynamics at the cost of musicality and naturalness depending on the rest of the system. We mustn't forget that this is an unfiltered DAC which must have the higher frequency digital artifacts smoothed out somewhere along the line. More detail also demands more of the source material and rest of the equipment.
It makes sense to remove added components like the buffer if the system sounds too extreme. All the other improvements are a direct replacement of a necessary lower performance component. The type of CCS circuit used or a buffer may be a matter of personal taste and the SQ of rest of the system.
dwjames - we have both gone down the road of fully tweaking a single board DDDAC. Unlike you, I did the buffer first instead of last which made a huge improvement in detail and dynamics at the time. I suspect that the Shunt PS on the Analog section would have provided most of this improvement anyway. The only remaining upgrade for me is the 2-chip 1/2 clock circuit for which I have ordered all the parts.
It seems that every upgrade provided more detail and dynamics - regardless of the order of improvements. At some point there could be a point of too much detail and dynamics at the cost of musicality and naturalness depending on the rest of the system. We mustn't forget that this is an unfiltered DAC which must have the higher frequency digital artifacts smoothed out somewhere along the line. More detail also demands more of the source material and rest of the equipment.
It makes sense to remove added components like the buffer if the system sounds too extreme. All the other improvements are a direct replacement of a necessary lower performance component. The type of CCS circuit used or a buffer may be a matter of personal taste and the SQ of rest of the system.
I'm looking forward to hearing more opinions and listening tests, and welcome the contributions this far and also their diversity - not only that my cup of tea may not be your cup of tea, but we all know that sometimes a very revealing segment added (or substituted in)to the audio chain (a positive thing) may be perceived as bad because it suddenly shows a weakness not previously apparent... Not saying this applies to any opinion this far, but I mean to invite other contributions and flavors of opinion.
By and large, it seems that the single JFET seems to gain some ground. I confess to being a little underwhelmed by this as we've started this endeavor trying to best the K&K Audio's design and a lowly 2SK170 JFET CCS is far from a challenge... But heck, truly, whichever circuit best serves the application.
By and large, it seems that the single JFET seems to gain some ground. I confess to being a little underwhelmed by this as we've started this endeavor trying to best the K&K Audio's design and a lowly 2SK170 JFET CCS is far from a challenge... But heck, truly, whichever circuit best serves the application.
Hi Palmito,
Do you know if this can act similar to Botic Cap? I eager giving this a go with BBB post Raspberry Pi sq appreciation. In anticipation of Rune 0.3-beta for BBB which consists of Miero Botic driver, I wonder this combo can be an ideal CA player for all 44.1 to 192k?
Hi James,
If I'm not mistaken, you have just taken the Crown of Stefan in DDDAC Mod! 😛
While I am no doubt very satisfied with my current DDDAC, reading your excellent feedback have intriguing me to do more...! 😀 Pls excuse my laziness here, has this 'buffer stage' discussed in schematic detail? I will give this ago but pin 20 seems too challenging for me both time and technical competency. 😉
Best,
Chanh
Don't give up yet on CCS
I just ordered some 2SK246 JFETs which, according to Erno Borbely, have a 0 tempco point right around 0.4ma. He contrasts this with the 2SK170 which has a 0 tempco point around 10ma and a huge current variation with temperature changes at 0.4ma. The 2SK246 datasheet for linearity at a low Vds looks excellent too. Stay tuned.
I expect to receive some LT3092 current regulating chips on Monday. I will also test the BF862 because I already have them on hand for buffers. Also, a fellow diyAudio-er and I designed a CCS miniboard which can accommodate a variety of circuits and installation orientations. I expect my first shipment of these next week.
Most important, there are CCS devices on pin 20 that noticeably increase the musical detail and dynamics of the DDDAC sq.
I just ordered some 2SK246 JFETs which, according to Erno Borbely, have a 0 tempco point right around 0.4ma. He contrasts this with the 2SK170 which has a 0 tempco point around 10ma and a huge current variation with temperature changes at 0.4ma. The 2SK246 datasheet for linearity at a low Vds looks excellent too. Stay tuned.
I expect to receive some LT3092 current regulating chips on Monday. I will also test the BF862 because I already have them on hand for buffers. Also, a fellow diyAudio-er and I designed a CCS miniboard which can accommodate a variety of circuits and installation orientations. I expect my first shipment of these next week.
Most important, there are CCS devices on pin 20 that noticeably increase the musical detail and dynamics of the DDDAC sq.
Last edited:
Having just looked through the current threads, I am interested in the value of each tweak that can improve on the DDDAC or any other.
I like the constant current idea and a better clock but the I/V ouputs beeneift from board stacking. I know this can help for a number of reasons.
However the typically 9 mA out puts being multiplied up is a known winner. Is this saying that the chip out put current is a major weakness. The early TD5543 8 and 12 stack are example. Is this showing a fundamental flaw in modern DAC chip design.
I reckon if chip manufacturers made 100mA i/v outputs it would put all the other mods to shame. No wonder digital struggles to match the best analogue. What if chip manufacturers are keeping this up their sleeve.
Still prefer FM live radio concerts, but digital is cleaner in the treble, but lacks the real being there. Am I biased
I like the constant current idea and a better clock but the I/V ouputs beeneift from board stacking. I know this can help for a number of reasons.
However the typically 9 mA out puts being multiplied up is a known winner. Is this saying that the chip out put current is a major weakness. The early TD5543 8 and 12 stack are example. Is this showing a fundamental flaw in modern DAC chip design.
I reckon if chip manufacturers made 100mA i/v outputs it would put all the other mods to shame. No wonder digital struggles to match the best analogue. What if chip manufacturers are keeping this up their sleeve.
Still prefer FM live radio concerts, but digital is cleaner in the treble, but lacks the real being there. Am I biased
Well, thanks I guess, but as I say a lot, my knowledge of electronics and hifi is really limited compared to a lot of people here. I'm certainly no master of these things. I just like fiddling about and I don't mind taking pictures and explaining things when I can understand them myself. With almost everything I've done so far, I've either been following best practice and datasheets or getting help and guidance from others who have been kind enough to share their knowledge. I love that in this place and with this bit of kit, we can all share experiences and ideas and we can all gain together and end up with kit that's developed and real world tested in a way that no solely commercial product ever could do. If there's any trophies being given out, they need to be cut up into pieces and shared around with everyone here who's joined in 🙂
But anyway... This buffer stage is really for the benefit of people running a single dac deck and output transformers like the Cinemags as it steps up and beefs up the output signal so it can effectively drive the transformers properly. Adding a load of dac decks will also do this, but this buffer stage is simpler and cheaper.
My theory is that stacking dac decks improves the sound in 2 mains ways: by parallel processing and averaging the results, which will give a more accurate translation of the digital to analog signal, but also some of the benefit comes from what Boldname has just described whereby the current outputs are combined to give a beefier output signal.
I believe that adding shunt regs in improved locations and tweaking other components provides another way of getting a more accurate digital conversion, and it appears that a buffer stage is an alternative method of beefing up the output current.
How do these other methods weigh up against using a big stack of dac boards? Only a listening comparison test can truly show that. I'm game if others are interested? 🙂
I'm still trying to get my head around whether 100% of the extra detail I can now hear with the buffer stage switched in originates from the recording. I guess a scope will tell me for certain.
One thing's for sure, it's all sounding pretty amazing to my ears these days. I was really impressed with the DDDAC when I first got it, but it just keeps improving with every little tweak.
To those worried about pin20 current source solutions being complicated, I can say that if you get e301 diodes roughly close to 400uA, almost nothing can be simpler. But I don't know if I was just lucky to get an almost matching pair with a current at 125% of the rated value. I imagine if you need 11 pairs like that which all match, it becomes a much more unlikely game of chance....
I look forward to hearing people's thoughts and listening impressions on other reference current solutions and I'd love to hear if anyone has any theory on why the lm334 doesn't sound so nice...
Cheers,
James