Nixie, would you kindly share a schematic so that we could also make and enjoy what you are hearing, I would be very keen on making a DAC that plays so much better than anything else.
The transformers are filtering!
Without those listening to such DACs would be torture to equipment and ears.
Without those listening to such DACs would be torture to equipment and ears.
Last edited:
I have devices both with and without that transformer. It's not torture at all, on the contrary. The DDDAC must have either that transformer or coupling capacitors or something else, because PCM1794A (mono mode) balanced outputs are at around 2.7VDC.
Last edited:
Here they are, all together:
DDDAC:
https://www.dddac.com/
The full experience is only with the tube I/V, that's what I'll deal with next.
Not so much, but the plan is to do the I/V conversion with the transformer first, tubes are just a gain stage. Similar to Audio Note DACs.
It's a big job for me, it will take time. 😎
It's a big job for me, it will take time. 😎
The transformer first is also filtering. So the filterless approach is not entirely filterless.
The transformers are the filters.
The transformers are the filters.
Yes, that's right. The transformer does the job partially, the signal after the transformer looks more like a sinusoid. But the transformer adds its own THD, colors the sound. All that together has little to do with measurement, just subjectively what one likes.
A digital signal has spectral copies (images) around all multiples of the sample rate. For example, a recording of a 1 kHz sine wave with 44.1 kHz sample rate has spectral peaks at 1 kHz, 43.1 kHz, 45.1 kHz, 87.2 kHz, 89.2 kHz and so on.
Normally a reconstruction filter or an interpolation filter and a reconstruction filter would suppress all the images, but when you only have a zero-order hold filter (that is, just keep the value constant until the next sample arrives), the suppression of the images is quite limited. Hence the weird waveforms.
When you can't hear the images, your ears effectively form the reconstruction filter. A cat with good hearing can hear up to 85 kHz, hence my statement that subjecting a cat to such signals would be animal abuse (unless the sample rate is 192 kHz or more, of course). In fact, I once did an experiment where I converted audio to 11.025 kHz sample rate and repeated each sample four times, to get an impression what 44.1 kHz-sample-rate audio should sound like to a cat through one of these DACs. It was pretty awful.
A side-effect of zero-order hold filtering is that with low sample rates, such as 44.1 kHz, the response already rolls off by a few decibels below 20 kHz. That may be one of the reasons for its popularity. There are also no intersample overshoot or pre- and post-echo issues with such a set-up.
It's hard to find a speaker that plays anything at 40kHz, so it's listenable without a filter. A piezo driver maybe, but that's rarely used. A classic dynamic driver cannot reproduce this.
I'm sure the response of your loudspeakers rolls off, but probably much too slowly to be usable as a reconstruction filter. 20 kHz has an image at 24.1 kHz at 44.1 kHz sample rate.
I have a ribbon tweeter that can reproduce a lot above 20k, and most dome tweeters are like a brick wall filter even earlier.
It is the basics of the sampling theorem. What goes in to (analog) and comes out of (analog) an ADC-DAC chain must not contain anything above fs/2 = 22,05 kHz for 44,1 CD quality. This is where almost every single DAC fails, it's just down 10 dB or so at fs/2 when it should be -150dB at least. -150 is my requirement - but the theorem requires infinite attenuation. For info on this ->
http://www.tonmeister.ca/wordpress/2024/10/01/aliasing-is-weird-part-1/
http://www.tonmeister.ca/wordpress/2024/10/01/aliasing-is-weird-part-2/
//
Not sure i agree. For the record i use and enjoy both OS and NOS, completely unfiltered dacs. With a significant bias towards the latter types since Miro's monster thread pushed me to build an 1862 dac.
It took some effort to get the sound to my liking. Initially it felt like HF sparkle and air were a little lacking. Interestingly, no analogue filters/nos frequency equalisers had any positive effect upon this. Eventually, the HF deficiencies were resolved mostly through intangible methods: different rectification, different PS caps, different i/v stage.
In the end, it's highly enjoyable and most definitely not a torture to my ears.
My power amps are followers with no voltage gain or loop nfb - heavily paralleled diamond buffer running in class A, or SE SIT followers. They can't be intimidated by slew rate. Voltage gain stages differ between tubes and discrete high voltage opamp types. They don't seem to be perturbed in any way by the unfiltered output.
Amplifiers with questionable stability, built upon prodigious amounts of loop feedback may react differently, i wouldn't know.
Finally, if out of band noise was indeed a serious issue, higher sampling rates, whether native or the result of upsampling would sound very differently. But they don't
The idea that a tube stage, by virtue of employing a tube, would offer a limited bandwidth and any meaningful amount of filtering is a myth. Bandwidth of a tube line stage can easily exceed 200kHz.
The other idea, that coupling transformers necessarily serve as low pass filters is also a myth. While such transformers certainly exist, the majority of decent transformers seem to err on the other side - they often have a resonant peak in the 50 - 100kHz range in order to extend the bandwidth.
Last edited:
Nah. For starters amplifiers have higher measured distortion levels than DAC. Amplifier distortion can be well into the audible range. Also you're not doing blind listening tests when you change OPAs (pretty hard to do actually as you have to power off to change them) so you are subject to the placebo effect and confirmation bias and all those psychological effects which are well documented but you keep choosing to ignore.
What is a little funny is how people are ok with having their beloved class D output 300 - 1000kHz of RF in their speaker wires but 30kHz from an unfiltered dac is a huge problem 😎
The problem with topics like this is that it is not defined what a good measurement is in terms of a good sounding device. Is it THD 0.00001% or THD 1% with nice harmonic decay and dominant 2nd harmonic? Or something else? The engineering approach is the minimum of any distortions, but I'm not convinced that it is the right way to evaluate.
300 kHz is so high even my cat won't hear it.
- Status
- Not open for further replies.