When measurements take a backseat

August 5, 2019
 by Paul McGowan

As engineers, we always start a new design with the best of intentions. We’re hell-bent on making sure our new creations perform to their best. One of the ways we do that is by measurement which always precedes listening.

When PS Audio’s analog engineer Daren Myers set out to design the new Stellar Phono Preamplifier his first stop was the measurement suite in PS Audio’s lab. “Extremely low distortion? Check. Vanishingly low noise? Check.” Then, it’s on to the music room to see how all that fine measuring equipment sounds. And that’s sometimes where confident hopes turn to disappointed scowls.

If you’d like to get in on the design process with a first-hand view, watch this short 2-minute video we put together of Darren explaining his journey from hope, to failure, to success. It’ll be worth the few moments it takes to get the inside scoop on this new product that’s sure to set the high-end audio world buzzing.

Subscribe to Paul's Posts

23 comments on “When measurements take a backseat”

  1. I strongly hope Darren made a cross check of the circuit modifications (resulting from the voicing in music room one) in his own listening room with his own stereo system and got identical sound quality improvements! Did he?

    1. Design is doomed when the concept it is based on is incompatible, that is insufficient to achieve your goal even when its ultimate potential is achieved.
      Measurements are doomed when you are measuring the wrong things, using a flawed method, have omitted critical measurements that would tell you vital information you need to know.

      More obvious flaws are invalid procedure, miscalibrated test equipment, the wrong test equipment, defective test equipment.

      You can’t measure the thickness of a human hair with a ruler. Taking my temperature will not tell you anything about my blood pressure where the real problem might be.

      At a trade show the chief engineer for Maxell magnetic tape had a setup with a Nakamichi Dragon tape deck and a lot of test equipment. He set out to prove his best tape was better than TDK’s. That’s how it looked but I pointed out that he’d have to optimize the bias and equalization for the TDK tape first and then show me better results than when it was optimized for his tape. He offered me a job on the spot.

      One of the most important measurements for a woofer is harmonic distortion as a function of frequency and output level. How many manufacturers give you a specification? How many test reports measure it and give you the results of that measuement?

      Tests based on legacy bench test methods like measuring the frequency response of an amplifier at one watt output with a resistor for a load won’t tell you it’s frequency response when you hook it up to a loudspeaker. The test conditions aren’t remotely close to the conditions of actual use.

      If your goal is to produce something that pleases you then measurements are irrelevant. On the other hand if you are trying to achieve the sound of music from a recording that closely resembles the sound of live music then the first things you have to learn to understand is sound, hearing, how to model them, and how to measure all of the parameters that correlate to what you hear. Short of that your efforts are hopeless because they measure all of the wrong things. When you measure that no one can hear above 20 khz but you produce equipment that has as it’s goal flat frequency response to 40, 50, 100 khz you are wasting time and effort. Your darts not only won’t score a bullseye, they won’t come close to hitting the dart board. When the most dynamic music has a dynamic range of about 80 db at a live concert building a system with a dynamic range of 144 db is another dart thrown in the wrong direction. This is why so called high resolution or high definition audio even when it actually works is a total waste of time, effort, and money. It does nothing to make usable improvements to performance.

      Study sound. If you ever understand it you’ll at least have an idea of what you’re up against trying to duplicate it and why all the best efforts you can buy have failed. You don’t listen to individual components. You listen to an entire system starting with the recording process and ending whit the sound that reaches your ears after you room’s acoustics have altered what come out of the loudspeakers. So the industry is still at square one looking in a lot of directions except the right ones. Pathetic.

  2. Massive negative feedback is, in principle, a Very Good Thing. In practice it is pretty good too, but there is a caveat. It assumes that there is no phase shifting within the amp, something which will not be the case, even when using a direct coupled circuit. As soon as you have frequency dependent phase shifts the negative feedback, whilst reducing overall distortion, will actually introduce new low level distortion. This can be very complex and is normally regarded as increasing the noise level. If the phase shifts are significant the amp will probably become an oscillator! Designers often make their major feedback loop frequency dependent to stop the risk of oscillation. I thus have no problem with the notion that a designer, by reducing the amount of feedback, can produce an amp which sounds better.

  3. Simply compare the transient response of a pre-amp with lots of negative feedback to one with less, chalk and cheese!
    Why is his head nodding to the music, is it to prove that he is “getting it”? 😉

  4. Darren seems like a cool guy. Plus, he’s listening in close field to Harbeth P3’s which are great monitors. I’m sure this will be a wonderful preamp for those who still listen to vinyl.

  5. “Why is his head nodding to the music, is it to prove that he is “getting it”?”
    According to the Blue Man Group nodding is always good, even a must when you get it.

  6. Shame this wasn’t around 6 or 7 years ago, I ended up having a phono amp modified to have a pair of mm and mc inputs with variable gain (on mc only). It’s odd how few phono amps allow you to connect two tonearms. I now use ADC, which solves a lot of problems and annoys purists, although PSA did have ADC in their phono converter. To my surprise there are some relatively new premium phono amps knocking about, most recently one from Rega, so it seems there is still a market for phono amps.

  7. “We don’t need no stinkin’ measurements”
    Refreshing to read and listen to someone not hung up on measurements. There are many, MANY people out there digging their heels in because focus solely on what they can only see.
    I have tubes and people tell me “Buddy you may think it sounds good but what about the DISTORTION”?
    I tell them give me more.
    Thanks for sharing of this video and the Stellar Phono Amplifier is a looker.

  8. I remember building a breadboard of a 120v 60HZ source with 48v DC input (500 watt output) in the 1970’s for military use. The unit had a 1% regulation spec and the frequency had to stay within 0.5% of 60Hz from -40 to +65C at any load condition. This was going to power a satellite tracking unit for the air force. It was about half discrete and half IC parts, the control portion was on prototype board but everything else was nailed on piece of plywood. We did it by modulating a 18kz carrier with a 60 Hz sine wave clock (555’s and opamps), this was before any switch mode IC’s were available.

    It took a lot work to that working right and in the end the military paid $20k each for the two dozen units (6″ high 19″ rack0 we built. The acceptance test for each unit took 4 hours with a government inspector witnessing all of it. The reason for this was they were using a commercial tape machine and it had an AC motor in it, a DC motor would have been a lot cheaper.

  9. Drummers today have access to several kinds of measuring devices for precise and consistent drum head tuning. That way, all drummers can end up with the same generic sound that will sound better than someone who does not know how to tune. But, when someone has the ear and heart … and feel for how drums should sound to him? Nothing can touch it. Its acoustic. But, its audio… ….. Likewise… A great designer will give his component “a soul.”

  10. Negative feedback is an error correction system which reduces distortion by comparing the input and output wave form, looking at the difference and adding this difference out of phase to the input in precisely the right proportion. The idea is to null out the distortion by predistorting the input signal in precisely the opposite way and by the exact magnitude the circuit will distort the signal. If you add too much, you will be adding distortion and if you add enough of it you will create even more distortion than you started with. The idea is simple enough, the optimal execution not quite so simple. In fact the math can get so complicated that the textbook I used which was written by two USAF generals was targeted at senior undergraduates in electrical engineering and graduate students. There’s more calculus in the text than there is in a calculus textbook. So, if you’re a tyro, don’t monkey around with what you don’t understand.

  11. Important question though – the version that measured well but he wasn’t happy with the sound, and the final version he was happy with – surely the next step was to measure the difference between these two in every single way possible (and then follow up with a “null-between-the-two” test).
    This would give a better idea what to measure the next time something like this is being developed, or at the very least to document something would be extremely useful to the wider HiFi community.

    Don’t get me wrong, listening is the final arbiter and should be, nevertheless this seems the obvious thing to do, surely?

    1. Strange isn’t it? You’d think an engineer would want to measure what sounded better so that he could understand why. Maybe he’s too lazy….. or …… maybe he doesn’t want to know. Anyway bench testing in isolation of actual use is futile because bench tests are not the same conditions as operational conditions. Connecting equipment to inputs and outputs can change the way they perform. And frankly if that is the case, to the degree that a change does occur….. it’s a weak overall design since my conditions of operation will not be the same as his. This is especially true for preamplifier circuits that have a relatively high output impedance connected to a load that has a relatively low input impedance like many solid state amplifiers. I still don’t understand why these circuits continue to use the legacy emitter follower and cathode follower output stages when a class AB low power push pull design would have a far lower output impedance. When it comes to power amplifiers, connecting them to real loudspeakers is a whole different kettle of fish from connecting them to a resistor. Not only can speakers be highly reactive, they can generate reverse voltage destabalizing the quiescent operating point of tubes and transistors. This is one reason why a high power amplifier will sound better than a low power amplifier of the same design operating at the same output level under the same conditions. It has to have a better power supply.

      1. Hmmm. “Lazy” seems harsh, I AM lazy – which is why I rarely finish anything I start designing 😉
        I suspect wanting to leave that last bit of “making it sound better” to, er, “tweaking” rather than strict maths (for want of better terms) is when it genuinely does become an art, rather than a science. If the end results are good then objective achieved, further investigation costs time and hence money.

        I take your point about real-world operating conditions, I am afraid I always optimistically assume that the very finest audio measurements will take this sort of thing into account, presenting dynamic source and load conditions and tracking the changes in characteristics they invoke, including highly adverse conditions. It seems that isn’t the case.

        I am sometimes glad that I can’t afford any of this stuff (which encourages soldering iron adventures) – nevertheless I have found this site an inspiration rather than dispiriting, as so many other “high-end” audio “websites” (that are little more than animated glossy brochures) end up being 🙂
        .

    2. I know. It’s tempting to believe that once you’ve made such a change you can simply measure the difference. The null test would show exactly that difference. Right? Well, yes, of course it would, but it would also be meaningless.

      What you get from a null test is a look at the difference between the two either on a meter or scope. You can derive a tiny bit of info by further taking apart the difference data, but that data doesn’t actually tell you anything other than it’s there.

      For those of you owning a Power Plant, you can see this easily. Go to the front panel screen and using the scope first look at the incoming AC, then the outgoing AC, and then the difference. That difference screen is the null test. Note the results. A lot of undecipherable “wiggles” and detritus on the screen.

      Now, imagine we did the same with the phono stage, recorded that, and tried to apply that to our next design. Meaningless. All it shows is that there is a difference, not what the difference is or where it is coming from our what to do about it. Worse, it would mean the next design would have to be identical to keep the null down. That doesn’t help us design better sounding products, it only helps make identical sounding products.

      Hope that makes sense.

      1. Hmmm, fair response, thank you 🙂
        I wouldn’t suggest trying to replicate a sound as such, just thought it would be good to know what changes in the sound with what changes in the circuit, however it occurs that:
        – components are NEVER going to be exactly as their “model” describes, even ignoring…
        – components have tolerances, even the very best closest tolerances, and…
        – more than a very small number of components and the number of variables quickly multiplies up to be unmanageable.
        As I said to Mr. Soundmind above – I always imagine test equipment to be much more able than it is, but ultimately I guess the finest resolution test instrument we have IS our ears, it just lacks a certain consistency, reliability, and a snapshotting / pen-recording feature 😀

  12. It’s always interesting (for a non ee person) that such good things can arise inspite of a certain time schedule in the back, especially as things seem to be no matter of just zero‘s and one‘s in electronic engineering but often a matter of luck, chance and the unexplainable (paired with the necessary know how, musical sense and experience).

Leave a Reply

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

4865 Sterling Dr.
Boulder, CO 80301
1-800-PSAUDIO

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