Here’s a thought

October 6, 2015
 by Paul McGowan

Here’s another thought on the continuing discussion of why a preamp matters in the chain. Perhaps it’s the DAC’s volume control itself.

One thing is obvious. A DAC without an integral volume control cannot be used without a preamplifier; either separate or integrated. Thus, when we think of why preamps matter with DACs, we’re only discussing those DACs with built in volume controls. And there’s a suspicion that it is the volume control itself that might be the culprit. So let’s take a look.

Using our DACs as examples, we start with the PerfectWave, Mark 1 or 2. The digital volume control on this instrument was lossless at any setting above 50. In other words, from 50 or higher, no change in resolution took place at any volume setting. A free ride. Our newest DAC, DirectStream, takes this a notch higher. Designer Ted Smith figured out a volume control that has zero loss at any level setting–a major achievement. So, how could a control with zero resolution be suspect? That’s a question I have been mulling on for some time now.

One piece of the puzzle seems obvious. We know that tiny changes in the way internal FPGA process are organized make significant differences in sound. Even changes to the display affect sound quality. It’s a delicate process when jitter, power supply and the tiniest of changes can be heard and must be attended to. What’s to stop us from believing that different level settings have different sound qualities–despite the fact there are no measurable resolution losses?

Were it to be found true much would be explained. For instance, we know not all preamps sound better than DACs directly into power amps. In fact, most don’t. This observation lends credibility to the explanation that it is not preamps that make DACs sound better, rather, it is preamps helping DACs not sound worse. This theory can only be true if the preamp is of sufficient quality to add less degradation than using the DAC’s volume control. That all kind of makes sense.

But, just because something makes sense, doesn’t mean it’s true.

How do we make this determination to see if this theory holds water? The first thought that comes to mind; use the preamp as a baseline and try the DAC at different volume settings. Only, there’s a fly in that ointment. A preamp’s performance differs at each level position on the volume control–turn the DAC’s volume down and the preamp’s up, to compensate, and we have changed too many variables.

But, for now, we’ll keep this idea at the forefront of our thoughts and research and revisit this subject when progress is made.

Subscribe to Paul's Posts

15 comments on “Here’s a thought”

  1. Could it be possible, Paul, that different designs of a gain or an attenuation stages (volume control) on the other hand handle different frequencies with different gain or attenuation factors respectively???

      1. Beside the double blind test issue every stereo system consist of a vast number of coupled oscillator circuits characterized by their resonances. Thus each tiniest modification in this most complex system will change the spectrum of the resonance frequencies. And thus the sound. Having no clue about the sound quality of the original recording or the mix created by the sound engineer with the help of his own speakers it is not impossible that the changes in the resonant frequency spectrum might yield in a sound (!) that seems better than even the originally mixed sound now heard by a different pair of speakers in a different listening room. I would suggest to make more valid and less subjective comparisons using a reference recording mixed with the help of the same headphones/ headphone amp that are finally used in the home stereo system.

      2. Have you tried measuring time delays? (phase). This is one area where audio metrology is usually missing. A resolution of a microsecond is recommended, several double blind tests have shown humans hear to at least 3 microseconds for IATD, which is not the most subtle perceptible time effect.

        Another test that is rare is high resolution IMD. Humans can easily hear .01% distortion if it is anharmonic, like difference frequencies arising from the slightest of non-linearities.

        Jitter falls in the area where the frequency and time domains intersect – picosecond time variations shifting frequencies.

        The difference between volume control in the same box and splitting the functions can be related to coupling between the digital domain and the analog.
        A complete enclosure induces ring currents in the case from internal and external RFI, and these can bleed around internal baffles even with extensive suppression on the conductors (GHz caps and beads). I have seen results improve going from a conventional circuit in a six sided box to minimized path length (~1″ for a power amp) on a three sided ground plane.

        A well respected mic preamp designer told me that he keeps the signal separated from ferromagnetic materials, an effect that is greater at lower signal levels because of hysteresis. Epoxy circuit boards have measurable non-linear dielectric absorbtion that may affect lower level signals more. I wonder if your preamp has point-to-point wiring?

  2. Paul
    If a DAC sounds perfect – no matter what the level is set to – a pre-amp should help the DAC not to sound worse?
    That is a strange – not to say funny – kind of logic.
    What is the comparative heightening to ‘perfect’ or – as I would say – ‘immaculate’?

    1. It is a very difficult question. When we “hear” a DAC it is through a specific chain – and we make a judgement based on what we hear. But the sound is influenced by the very chain we are required to listen through – so we can never truly “hear” the pure DAC. My supposition is good preamps allow a DAC’s volume control to be set at 100 and out of the picture – something we cannot do or try without too much volume.

  3. Let’s stay on this topic for a while. It is a very divisive one…to pre or not to pre. I like that Paul is being open-minded and willing to toss theory aside and find, well, better theory.

    From the post above…here’s a question to ponder…does “perfect” sound always sound better than “non-perfect”? I’m not so sure. The portable FPGA Dac I use ( a Hugo) sounds much better when run through a high end tube preamp. And I always thought active line-stage preamps sounded better than the “short signal path” of a passive preamp…if the active line stage was well-designed. That said, I’ve met plenty of high end preamps that I didn’t care for when placed behind a digital chain.

    I think Paul is on to something when he looks at the FPGA programming; there are endless combinations and permutations that may, or may not, add synergy to the chain when a preamp become part of the equation. Or does the preamp somehow simulate a rounding of the digital steps in the waveform, hence smoothing things out, or adding body and warmth?

    For me, I’ve come closest to analogue with a BAT 51se (tube preamp) in front of an FPGA DAC. If anyone has, or has tried a BAT (or other high end tube pre) in front of some of Paul’s gear (DirectStream)…I’d be be anxious to know if they liked what they heard.

    Paul, I would think the only way to really test some of your theory is to somehow replicate the volume control design in one of the better preamps that you liked and add only that to the DAC chain before the amp. Then you eliminate the other aspects of the pre that may (or may not) be contributing. Though I know that impedance matching is also important, I suspect that even if it were the same, the DAC would sound better with the “whole” preamp design in front of it, and not just the volume control from said design.

    In sum, I’ve found that many non-tube DACs sound better with a tubed preamp. – Lorne

    I strongly suspect that putting tubes behind a FPGA DAC or other SS DAC is just something that works well. I don’t know if it is accurate or measures well, but using the highest resolution of devices that I can find (with maybe the exception of the bass), Stax 009 electrostatic headphones, I find this combination sounds less fatiguing than no preamp, but still maintains accuracy, speed, micro-details and resolution that FPGA’s are getting so good at.

    1. The problem with this well thought out suggestion is that a well designed preamp attenuator, one that actually sounds right and helps DAC not sound worse, requires an output buffer at the least, a preamp stage in any case. In other words, a preamp is nothing more than an input switch, an attenuator and a line stage. You could eliminate the input switch, you cannot eliminate the attenuator or output stage to get the results we’re discussing–as in an earlier post where I described the problems of passive attenuators.

      This is a very interesting problem, indeed. And when I feel I have more to contribute to the question, we’ll reopen the discussion.

  4. I’ve tried to use the PWD (Mk.2) as a preamp, connected with XLR to the amp.
    The digital volume control most of the time set in between 50 and 60.
    But I did not like the sound too much. A little bit dense, not tranparent.
    Volume set to 60 or 70 didn’t fix the problem, and that was already waaaay too loud.
    So 90 or more could/would have been the answer, but I couldn’t risk a war with the rest of the neighborhood…
    And that, for me, has always been the problem when using a DAC as a preamp.
    So I purchased a topclass preamp and set the DAC volume control 100.
    I tried, just for fun, to set it to 80 or so, but even then the sound became a little bit less open.
    At least in my imagination.
    Bottom line : I still don’t like digital volume controls, no matter what the experts say about it.
    An old fashioned analog control still does the job best. Al least for me.

  5. This is not exactly about this discussion but perhaps it’s an ‘analog’ of it. I normally hate the term musical. It allows one to call any sound good, especially in out audio universe defined by such a complex combination of values.

    I have a set of small, 2 way speakers. The designer, a friend, wanted to add the last octave and added an 18 inch bass reflex woofer with passive crossover designed just for this speaker(there are only 3 examples). He was worried the woofer would over load his room so he designed it so the port could be stuffed to produce a good closed box alignment. The price in bass was about 10 Hz. I listened both ways and chose the closed alignment because it had better definition but it was drier. We spoke and both agreed on what we heard. And there’s no question that a reflex alignment is less accurate; it has more overhang. He knows that, of course, but he chose the reflex format because it was richer and more musical, even more real sounding. And I know exactly what he’s saying.

    Could you, Paul, be discussing something like this?

  6. “How do we make this determination to see if this theory holds water? ”

    I’m always up for a good puzzle, especially an electrical puzzle. How do we determine if a potentiometer causes audible distortion. What method would tell us? I’m going to assume that the distortion is of a nature or so small that no test instrument known can measure it. It can only be heard by audiophiles. This in the pursuit of a never ending crusade for truth, justice, and the rooting out and extermination of every last trace of distortion no matter how insignificant. I’m reminded of something my father used to say. (My parents were very strange, unusual, and interesting people. How do you think I got this way? 🙂 )

    One day I sat upon the stair
    And saw a man who wasn’t there
    He wasn’t there again today
    Oh god I wish he’d go away!

    Okay, my method is based on one of my usual tricks for this kind of test, a technique audiophiles don’t like. Comparing the element in question with a shunt. Fortunately Paul, you have all of the resources you’ll need to set up this test. An small army of clever and ingenious engineers, technicians to build breadboards, and the place to conduct the test.

    You’ll need to first design your sound system with no volume controls at all….anywhere. The system will have a fixed gain!!! The gain will have to be designed to be not too loud and not too soft given the sensitivity of your speakers. Hey, audio engineers have designed systems with every other control taken out, why not this one too? Even your on off switch doesn’t do a whole lot. So you have a system with no potentiometers. It can include a D/A converter, a preamplifier, or both. Now install volume control potentionmeters or whatever else you use in your D/A converter with a bypass or shunt switch. Set the gain with the volume control in the circuit of each so that the gain of each unit is the same with the volume control in or out. To test the volume control at different settings you’ll have redesign the system components to have different individual gains but the same overall gain. In other words you’ll have to play the relative gains off of the D/A, preamp, and power amp so that different gains and therefore different volume control settings for each will yield the same overall system gain.

    There are some disturbing things you’ve said.

    “We know that tiny changes in the way internal FPGA process are organized make significant differences in sound.”

    If there are changes that are level dependent but not due to the volume control the FPGA is not linear.

    “Even changes to the display affect sound quality.”

    This suggests that there is either a power supply issue or there is coupling between wires to the display and signal wires. A separate power supply and better internal isolation could reduce or solve this problem.

    I can’t tell you how happy I am that my hearing is not this sensitive or that I just never noticed it and it wouldn’t matter to me if I did. I still have tons of adjustments to compensate for such a problem.

  7. Are these tests done with the same output levels? I would suspect pre-amp/ DAC combination volume tests would include speakers responding differently to output levels affecting their magnetic flux. And if testing perceptions of quality in volume changes, wouldn’t our ear/ brain combination play an important role? The Fletcher–Munson curves came to mind as an example.

      1. Gain matching is a tricky issue. Two speakers, one bright sounding with strong treble and weak bass and another with strong bass but weak treble are arbitrarily matched in loudness at one frequency, say 1000 hz. Which one will sound louder? It depends on the recording being played. Either might sound louder than the other in this example. Often speakers that sound bright seem to be louder. This is an ear catcher for tyros in showrooms, especially acoustically dead ones but eventually prove tiresome to listen to in the long term at home. Solution, buy wires with high inductance and/or capacitance that roll off the high end….or throw a blanket over it.

Leave a Reply

Stop by for a tour:
Mon-Fri, 8:30am-5pm MST

4865 Sterling Dr.
Boulder, CO 80301

Join the hi-fi family

Stop by for a tour:
4865 Sterling Dr.
Boulder, CO 80301

Join the hi-fi family

linkedin facebook pinterest youtube rss twitter instagram facebook-blank rss-blank linkedin-blank pinterest youtube twitter instagram