You are right as for a sufficient rejection of images, @abraxalito. Nothing is wrong in your approach, keep a good work. If any doubts, Foobar2000 with SoX is giving free upsampling. With upsampling 44.1KHz source twice, we don't have to worry about imaging anymore. Most of the music I download now has a higher sampling rate anyway.
As for the preference of a digital filter brick wall, it is personal. It gives flexibility to the design, better measurements, but sonic advantage is at least questionable. When chosing between brick wall filter and upsampling, I will always do quality upsampling on a fast PC CPU.
At the moment I do enjoy a second-hand Audio GD R2R-11, I was lucky to buy locally. I got it only because of GrossDAC project delay, first a flu in China and now a complete lockdown of South Africa. I wouldn't be able to order anyhing during 21 days, while (more realistic) it will be extended to three months.
A-GD R2R-11 is a NOS design. I don't see any second-order LP filters, it would be some inductors in this no-feedback amp design. I don't see any, but it sounds fabulous, very real and natural presentation.
As for the preference of a digital filter brick wall, it is personal. It gives flexibility to the design, better measurements, but sonic advantage is at least questionable. When chosing between brick wall filter and upsampling, I will always do quality upsampling on a fast PC CPU.
At the moment I do enjoy a second-hand Audio GD R2R-11, I was lucky to buy locally. I got it only because of GrossDAC project delay, first a flu in China and now a complete lockdown of South Africa. I wouldn't be able to order anyhing during 21 days, while (more realistic) it will be extended to three months.
A-GD R2R-11 is a NOS design. I don't see any second-order LP filters, it would be some inductors in this no-feedback amp design. I don't see any, but it sounds fabulous, very real and natural presentation.
Last edited:
Since the PCBs are now in I've started building the first filter board and compared the capacitance readings I get on my LCR meter with those from the ZT219. They're reassuringly close - the ZT219 being about 0.2% away from the LCR. Which is plenty good enough in this application. But of course this is just my meter, no guarantees about anyone else's.
I'm not sure if 2% is good enough yet as I always aim to get better than 1% for my capacitance values. It would depend on how much passband ripple you think is acceptable - the kinks in the response which are just visible on the plots in post #716 will get bigger. But they're pretty small in comparison with the usual kinks in speaker response plots so I'm not sure they'll be audible.
Oh I noticed that the excellent accuracy of my meter doesn't extend to the lowest capacitance range (9.999nF) - on that range there seems to be an offset caused by some stray capacitance which isn't being nulled out. So it consistently reads higher than my LCR meter.
I'm pleased to report I have now gotten the first prototype singing. I found my first suspect AD829 in this build having soldered down a few dozen already. I may have to change one of the red LEDs to an IR LED to get the reference voltage into the right region, its at present too high with 3 red ones. I also need to figure out a way to make debugging a bit easier - with the filter in place the analog parts are too hard to probe. Some test points on the bottom might resolve this. Now to build up a few more and try a few tweaks.....
Oh I noticed that the excellent accuracy of my meter doesn't extend to the lowest capacitance range (9.999nF) - on that range there seems to be an offset caused by some stray capacitance which isn't being nulled out. So it consistently reads higher than my LCR meter.
I'm pleased to report I have now gotten the first prototype singing. I found my first suspect AD829 in this build having soldered down a few dozen already. I may have to change one of the red LEDs to an IR LED to get the reference voltage into the right region, its at present too high with 3 red ones. I also need to figure out a way to make debugging a bit easier - with the filter in place the analog parts are too hard to probe. Some test points on the bottom might resolve this. Now to build up a few more and try a few tweaks.....
Attachments
Hey Richard, I was gone for 2 weeks of insanity and you have been busy! This implementation does not look like it has as many capacitors, no hats, no pants etc..
How does it sound compared to the Phidac and the original Lingdac?
Also, when you are happy with the layout and have it finalized, put me down for 5 boards.
How does it sound compared to the Phidac and the original Lingdac?
Also, when you are happy with the layout and have it finalized, put me down for 5 boards.
thanks for the explanation abraxalito, It was confusing because I thought that P1-P4 were chips, i think they are are a connector. I also didn't know why there was a filter at the output, never heard of NOS droop, do all NOS dacs suffer this?
I have listened to parralel ad1955 and didn't like it as much as a single, however I don't really like this dac chip so much. I also didn't think parallel tda1543 was that good, but this is not a high res chip, musical and nice sounding, but misses alot of the detail.
@stellarelephant - good to see you again on the thread Kevin. Yeah the larger board saves time in interconnection even though the form factor doesn't look quite as cute. The industrial-looking filter might turn out to be overkill, we shall see
@pelopidas I sure have been busy but that's for a good reason - not much else to do when everywhere outside goes into 'suspended animation' ! You don't reckon the filter-on-top counts as a 'hat' ? But no pants this time as that nut is cracked in a different way. As for sound, its clearer and with more 'air' than the PhiDAC. I've not listened to lingDAC in such a long time I can't come up with a direct comparison but I would guess it beats that in the bass, lingDAC's was a bit on the weak side with its stock PSU. A comparison of the HF of the two would be interesting. Thanks for your reserve order
@Luke All NOS DACs suffer from NOS droop yes, its baked into the math. I started my DAC journey with AD1955, I wonder if any of my original modded boards are still serviceable...
@pelopidas I sure have been busy but that's for a good reason - not much else to do when everywhere outside goes into 'suspended animation' ! You don't reckon the filter-on-top counts as a 'hat' ? But no pants this time as that nut is cracked in a different way. As for sound, its clearer and with more 'air' than the PhiDAC. I've not listened to lingDAC in such a long time I can't come up with a direct comparison but I would guess it beats that in the bass, lingDAC's was a bit on the weak side with its stock PSU. A comparison of the HF of the two would be interesting. Thanks for your reserve order
@Luke All NOS DACs suffer from NOS droop yes, its baked into the math. I started my DAC journey with AD1955, I wonder if any of my original modded boards are still serviceable...
At the start of my DAC journey I really didn't have any reference other than comparing a modded board to the original. I did find mods made a surprising amount of difference, mainly because the layout initially was so bad (the designer had ignored DS recommendations). I played around a lot with opamp I/V stages and filtering but eventually migrated to playing with TDA1543s which initially sounded rather dull to me in comparison. Is that what you meant by 'lacked detail' ?
Hello Richard,
(with reference to the above quote from post #709), would the SEbuff require modifictions in this case?
And if one's system comprises a preamplifier + amplifier, is the SEbuff still necessary to interface the PhiDAC/LingDAC (to the system)?
Hello abraxilito,
I recall the lingdac v2 design had space for non fit capacitors with regard to compensating variation of inductors and their effects to frequency response. The difficulty to non technical eventually led to the abandonment of the lingdac, however those with understanding can still progress.
The lack of base had also arisen from some in the lingdac, although I can’t say that I found/find that (simple battery power). The power supply improvement from subsequent development has led to better low frequency.
Is there any way to make improvements to the existing lingdac power supply? Or is it best just to move along with the new progress?
I recall the lingdac v2 design had space for non fit capacitors with regard to compensating variation of inductors and their effects to frequency response. The difficulty to non technical eventually led to the abandonment of the lingdac, however those with understanding can still progress.
The lack of base had also arisen from some in the lingdac, although I can’t say that I found/find that (simple battery power). The power supply improvement from subsequent development has led to better low frequency.
Is there any way to make improvements to the existing lingdac power supply? Or is it best just to move along with the new progress?
No mods required to the SEbuff but it runs at 7V so will need a regulator to reduce the 12V supplied to the DAC. Alternatively if the headphones being driven were higher impedance ones (say HD650 300R) then the SEbuff could be driven direct from 12V and the bias current reduced to maintain the same dissipation in the FETs.
No need for SEbuff if the load to be driven is a normal line-level one (above 5k say). The buffer's only needed to drive headphones i.e. high current loads.
Running on batteries - low impedance LiFePO4 would be ideal - wasn't the configuration when I remarked on the bass weakness of lingDAC. Rather I was running with the stock PSU board which has a few switched-cap voltage doublers and discrete series regs. Adding extra caps on the supply to the filter-I/V board improved matters in the bass with that set-up.
Run the filter-I/V board on the lowest impedance supply you can get, either with low ESR batteries or lots and lots of caps. I think I ran with over 100,000uF in my improved set-up - I discovered that my substantial supply of Panasonic 15,000uF 6.3V caps still worked fine at 9V
I eventually discovered that relative 'dullness' was more true to the source than the 'detail' which I had become accustomed to in AD1955. That finding was made when I started playing with low-pass filtering on the output of a DAC - by increasing the number of ferrite beads I used between DAC and I/V stage I was able to reduce the amount of 'detail' I got. Now I see 'detail' as an artifact which indicates something wrong in the DAC.
Of course some of the 'dullness' comes as a result of NOS droop so that needs fixing up.
Of course some of the 'dullness' comes as a result of NOS droop so that needs fixing up.
that's an interesting observation regarding detail being filtered out. I have a buffalo dac and a TDA1541 OS dac (pedja) and they both sound better to me. than tda1543 or AD1955, my biggest problem is my speakers are not good as the rest of my audio (kef Q900). I want to upgrade them at some stage, but they aren't too bad either.