I'm curious as to the role of the digital filter in getting 'real music' sound out of a DAC. Certainly lingDAC with no digital filter sounds that way, but I've also tried it with 2X OS from my PC (Foobar with SoX resampler) and from 88k2 upsampled files created in Audacity and it loses none of its 'live sound' when fed in that manner. I conclude from these experiments that NOS sound isn't inherently the absence of a digital filter.
The PMD100 is an 8X upsampling device so it could be that PCM1702 (being R2R architecture) loses subjective dynamics when running at much higher speeds.
I do not think it was a matter of 8 x oversampling speed. The PMD100 was the best sounding oversampling filter we used in front of the PCM1702 regardless of number of times oversampling . And none of them sounded quite like the NOS version.
It is a strange thing, when you use a DSP to resample the bitstream, you do not loose the liveliness. I guess there is a fundamental difference. The digital filters (oversampling or not) used in DAC´s is there to remove aliasing , that is their main purpose, whereas upsampling algorithm do not have this as the main goal. I must say I am not enough into these problems to understand the difference completely, so I am only guessing.
You could try to look at the wave forms from the resampled files with no analog anti aliasing filter and see if the resampling process actually removes (reduces ) the aliasing artifacts (can you see the discrete steps). I would guess the filter used in the resampling process is quite different than the filters used in dedicated DAC filters?
I'd hazard that being as it was a couple of decades ago, there wasn't much opportunity to try out other degrees of oversampling (like 2X for instance)? Most commercial filter chips are 8X, there is SAA7220 at 4X which is rather an exception.
Do you mean by 'DSP' here that when you do the resampling remote from the DAC chip itself you don't lose the liveliness? Coz otherwise PMD200 is both a DSP and a digital filter chip, but then you didn't try that one. I rather suspect that NOS mods to older CD players involving removal of the SAA7220 may well benefit the most from the reduction in digital noise generated by that chip.
The digital filters (oversampling or not) used in DAC´s is there to remove aliasing , that is their main purpose, whereas upsampling algorithm do not have this as the main goal. I must say I am not enough into these problems to understand the difference completely, so I am only guessing.
A nit-picky point first - with DACs its imaging we want to remove rather than aliasing. No digital filter is able to remove imaging, that's inherent in any D/A converter. What the filter (whether local to the DAC chip or remote) is doing is making it easier for an analog filter to remove it by taking it further away from the wanted signal (music, up to 20k). So there's no difference in principle between a digital filter on or close to the DAC chip and a remote upsampler (say PC or DSP) - what's different is in the latter case hopefully there's less noise being generated nearby.
I can't see discrete steps coming out of lingDAC as the analog filter smooths those over. Filters used in dedicated chips are normally designed for absolute minimum silicon area so are typically 'half-band' type which reduces the number of multiplies by 50%. DSPs and PCs are not normally so constrained so can use better algorithms - for example ones which don't have a transition band which permits aliasing (as does a half-band filter).
The PMD 100 had options for 2x 4x and 8x oversampling, so that was easy to try , even decades ago.
No what I mean is that the software DSP in PC´s that has way more calculating power than standalone DSP. And there is still IMHO a difference between resampling and digital filters that is made to remove the imaging AND aliasing (sorry for not mentioning the imaging
)Still this is just my humble opinion and I have not made any research to prove anything.
I do not recall that I have said that I know how the ultimate NOS DAC should be designed. Only recomended iperv to listen to a NOS DAC if he never had heard one . The "well designed" was only mentioned for picking a DAC that have achieved some reputation and not just the cheapest Chinese offering (not saying that they all suck)
Oversampling ...
About the SAA7220.... see other DIYA threads about proper use of this device. It can deliver high-quality sound. Also note that when Philips designed their CDPs in the 80s, they were designed as a system of an "integrated chipset". I.e., the 7220 likely performs best with an SAA7210 upstream and a TDA154x downstream.
About examples of commercial NOS dacs that have garnered "critical" praise ... Zanden 5000 DAC. From ~2001 - present.
See; 6moons audio reviews: Zanden Audio Model 5000 MkII
For commercial audio products, the digital filter has to be oversampling.
About the SAA7220.... see other DIYA threads about proper use of this device. It can deliver high-quality sound. Also note that when Philips designed their CDPs in the 80s, they were designed as a system of an "integrated chipset". I.e., the 7220 likely performs best with an SAA7210 upstream and a TDA154x downstream.
About examples of commercial NOS dacs that have garnered "critical" praise ... Zanden 5000 DAC. From ~2001 - present.
See; 6moons audio reviews: Zanden Audio Model 5000 MkII
Hi
I saw the schematic. A simple common base + current mirror, interesting
circuit. Did you compare it to Rbroer´s by any chance?
Less simple I/V for TDA1543
That one is a folded cascode, which Jocko Homo recommended (I believe they were in touch via PM/email). Richard recently used something similar, but his circuit (Richard´s) is unique in that the filtering takes place before the active stage.
Do you prefer the servo to coupling caps and transformers? I currently prefer transformers. Even if I have to use a coupling cap before the trafo. With tda1543 single ended for instance. The trafo filtering function is useful for sure, with double speed (88.2). Which I prefer (to non-oversampling).
Thanks
Alex
Last edited:
I have no tried Jocko Homo´s I/V stage, sorry.
The DC servo I use in my I/V is a little special, as it is controlling the current in the CCS , as opposed to injecting the error signal into a input. I have used this form for DC servo in different circuits, and yo my ears, it does not have the same ill-effects as normal DC servos, so , yes, I prefer it to coupling capacitors. I have not much experience with transformers , though.
But this type of circuit seems to be very tolerant to the high freq. noise that comes out of an unfiltered DAC (first time I designed it was with PCM1702) and I believe this is one of the reasons op-amps are performing poorly as I/V in NOS dac´s.
After quick glance on Rbroers I/V , it looks very much like my version, but the DC servo is here injected to the input, and I do believe you get more isolation from the DC servo and its faults (there always are unwanted artifacts) by injecting it to the CCS (T1 in Rbroers circuit).
Interesting.
About transformers: they can slightly soften the sound in a way that is similar to analog tape (which is a pretty much a transformer isn´t it?). So it isn´t the last word in resolution, but to my sensitive ears they are the ticket as of lately. And the low pass function is useful.
I am curious about the differences between your I/V (with current mirror) and the more typical ones which usually have a CCS at the top. Did you try any other IVs?
I have seen the mirror before, in Eric Juaneda´s schematics for example.
Thanks,
Alex
About transformers: they can slightly soften the sound in a way that is similar to analog tape (which is a pretty much a transformer isn´t it?). So it isn´t the last word in resolution, but to my sensitive ears they are the ticket as of lately. And the low pass function is useful.
I am curious about the differences between your I/V (with current mirror) and the more typical ones which usually have a CCS at the top. Did you try any other IVs?
I have seen the mirror before, in Eric Juaneda´s schematics for example.
Thanks,
Alex
I like current mirrors, they are pretty easy get to perform well.
Current mirrors are used very often in power amplifiers at the bottom of the VAS stage (goldmund as example). The trick in the I/V stage is it acts a little like a cascode, in reducing the Miller effect in the input transistor (not that important in a common base circuit) and the voltage gain is performed by the second transistor in the mirror, depending on the impedance it looks into.
I have tried a lot of opamp based I/V and passive solutions with high gain stages after. The 1543 IMHO sounds better with minimal voltage swing at the output of the DAC chip.
Thanks for explaining. I have tried a few different IVs myself, but didn´t try the current mirror yet.
My preference depends on what I´m listening to! Some things work very well with passive I/V. It is a sort of distortion that can "disappear", you hear it at first (it thickens the sound a bit; 2nd harmonic I suppose) but then forget about it and I can definitely enjoy the sound this way.
I would agree that active I/V with minimal voltage swing at the DAC sounds cleaner. More definition too. It can sound thin depending on what I´m listening to. My amp and speakers are fairly clean sounding.
Thanks
Alex
I don´t think it´s off topic, but in any case. BTW what I wrote below is about Abraxalito (Richard) and the interesting I/V he presented here for the tda1387.
Thanks
-Alex
Thanks
-Alex
Ha, welcome to the thread and thanks for the bump. BOMs are just awaiting final revisions with last-minute circuit changes, I'll endeavour to put them up before the weekend's out as I have a little unexpected free time
For the power supply I do have a 5*5 PCB with just one error on the gerbers. This is using an array of LM2662s as voltage doublers to create the rails necessary for the 3 boards from a single 5V supply at 250mA. I will go on a hunt for its schematic, the BOM will take a little bit longer.
For the power supply I do have a 5*5 PCB with just one error on the gerbers. This is using an array of LM2662s as voltage doublers to create the rails necessary for the 3 boards from a single 5V supply at 250mA. I will go on a hunt for its schematic, the BOM will take a little bit longer.