Phono Frequency Response

I had a nagging to play vinyl on Saturday Night.  On top of some pile I can't remember buying I found Queen's Greatest Hits.  Having heard these same songs on digital, in some cases high resolution digital, this was different, having more presence, transparency, liquidity, and magic (and some noises).  But also, seemingly, less bass.

I think the less bass part is primarily a problem with this "Greatest Hits" pressing, which may not have the extended bass of the originals.  But also I suspected it could be my turntable system.  Often when there is less bass, we hear it as greater presence and transparency.

So I got out Hifi News and Record Review Analog Test LP, and played Track 7 Side 2 which is a sweep tone.  One of the very annoying things about this record is that each band is walled off from the others.  In order to play Track 7, you can't just cue to the space in between 6 and 7, because that just reverts back to 6.  You have to pass the wall at the beginning of Track 7, and cue precisely there.  It's hard to hit that point precisely, and when you do, the tonearm is still stabilizing at the beginning of the track.  That might, partly expaln the bass below 30 Hz in this recording shown here in Audacity.  There's also a matter of intermodulating with the tonearm and antiskate system resonances..

 The bump at the bottom where it more or less stabilizes is 30 Hz.  Up from there to the midrange level represents about a 1dB increase.  So it appears the bass below 100 Hz is gradually rolled off by 1dB at 30 Hz, then looks funky.  (The scale above is linear so it shows changes that look larger than they would on a logarithmic dB scale.)

I wouldn't think that 1dB loss would be that big a deal, I routinely make changes of that amount or more in the bass level, and in fact I tried adding 5dB (by changing the DAC output level from -15dB to -10dB) and the Greatest Hits recording still sounded like no bass, until the last track.

The instability below 30 Hz might be a problem, but possibly worse there is a peaky resonance just below 20khz which shouldn't be there.

Another measurement I made indirectly shows the instability below 30 Hz is not a matter of initial tonearm/table stabilization.  I happened to drop the needle at 30Hz instead of 20 Hz, and the reproduction is pretty clean, only showing the barest overlay of the needle drop stabilization, which disappears after 13 cycles of 30 Hz (just over 1/3 second).  That initial instability could also relate to the slight peak (small fraction of a dB) at 30Hz shown in the previous measurement.  This means that the instability shown above is caused by something else, presumably interaction with tonearm vertical and horizontal resonances that needs fixing for good bass below 30 Hz.  I'm showing higher magnification in the selected portion just after the needle drop (also shown) for clarity.

Digging through my files, I found a measurement with my Mitsubishi LT-3 linear tracking turntable, with damping added to tune the bass just as with the Linn (though it was an earlier job I didn't do as much or as well).  They both have a Dynavector 17D3 playing through an Emotive XSP-1.  I'm showing this somewhat aligned with the Linn result above.

The Mitsubishi is on top here.  I think it has slightly flatter bass below 120 Hz (perhaps 0.5dB loss down to 30 Hz vs 1.0dB for the Linn), it lacks a weird resonance in the Linn around 120 Hz, and the bass extends possibly a tad farther before it becomes unstable below 30 Hz--but they are so little different in that regard it's hard to tell, in both cases the bass becomes unstable just below 30 Hz.  They both seem to have a resonance just below 20kHz, I'm wondering if that's coming from the cartridge body or the screws.

I wonder if the resonance around 120 Hz in the Linn is the infamous flexible subchassis resonance and "flabby midbass."  From the day the LP12 was introduced people wondered why they were getting such a thin and flimsy subchassis with an expensive turntable, when even much cheaper turntables like the AR Turntable had very heavy gauge metal.  HP of Absolute Sound denounced both the subchassis resonance and the flabby midbass in his first review of the LP12.

Famously Linn doesn't have a subchassis upgrade except for the very expensive Cirkus bearing.  I figure this is because the original bearing has a noise that is absorbed by the flimsy subchassis.  The expensive Cirkus bearing eliminates that resonance, so you can even have specially cast and machined subchassis with it.  Such were never offered for the original bearing (except by outside vendors...and they were not universally loved).

The flimsy subchassis could have eliminated a bearing noise.  But the price may be that funny ripple in the frequency response around 120 Hz.  Remember the cartridges are the same model.

Although the Mitsubishi turntable yields better frequency response, I don't think it sounds as good.  It has larger and more undamped arm/cartridge resonances and/or motor and bearing noises.  The movement of the linear tracking tonearm subtly affects tonality.  I could possibly fix the larger arm/cartridge resonances, with the application of more damping material, and the electronic motor is probably due for a tune-up.  My Linn Valhalla board has been replaced at least twice, the Mitsubishi has it's original electronics and it's a wonder they're even still working.

Testing the M50

In May 2012, I decided to reward myself for something or other, and bought the M50 I have long planned to buy.

The Earthworks M50 is the king of measurement microphones nowadays, with calibrated (and quite flat) response to 50kHz.  It costs far less in real terms than the standards of old, such as B&K's with their custom amplifiers and such.

I think John Atkinson of Stereophile uses an M50, either that or an M25.

Until now, I've had no proof that my system even has frequency response above 20kHz or so, something I've invested thousands of dollars and endless fine tuning to achieve.  I have no idea what the proper calibration for my previous custom 25kHz calibrated Dayton Audio above 25kHz.  Given the different calibration curves for different angles, it's also hard to know which of several calibrations to use.  I didn't really trust it to 20kHz let alone 25kHz (but in fact I was wrong...it matches the M50 amazingly closely!  Perhaps I needn't have bothered.  Nah.  Anyway I plan to use my old mike for general purposes and the M50 for calibration and "final" measurements.)

For the measurements here, I recall I put the microphones straight forward.

Here is my system response with the 25kHz Dayton Audio microphone:

Here is the response with the M50, which I finally figured out how to calibrate correctly (the calibration they supply cuts off below 700 Hz..for ARTA I had to add in fake points at that exact same level).

The M50 shows a slightly cleaner HF peak at 22kHz, and about 3dB more output at 40khz, and that's about it.  If I had an accurate calibration above 22kHz for my Dayton I could probably make them match even better.

I think the speaker system response, shown in much higher resolution that you usually see above 100 Hz (and less below, thanks to FFT techniques) is pretty good.  It slopes nicely downward from 1kHz (though perhaps a bit too early and there is more loss than desirable above 8kHz or so.  The inaudible 22kHz peak merely restores the highest treble to the midrange level (and less than the deep bass), and also makes possible the response extension to 40kHz where there may still be more output than at 14kHz.  Many highly esteemed tweeters have much larger ultrasonic peaks from fancy construction.  My 21mm Dynaudio D21AF's have factory response curve showing flat from 1200Hz to 39kHz measured with close microphone but they don't account for a lot of variables here, including the polar response dilution of a cloth dome tweeter (which is good enough to even make 40kHz response possible--which can't be done with many materials including of course my electrostatic speakers).

Anyway, there clearly IS ultrasonic response at useable level at 40kHz, which has been my goal, and now proven with a top calibrated microphone.

Instead of huge peaks and valleys in the bass in the unequalized bass response, or where the wall bounce isn't properly corrected (I dodge that with my 4 foot wall spacing and crossing subs at 125hz with 8th order linear phase crossover) it's pretty flat 80-800 Hz (amazingly flat in my building and measuring experience), with just a soft rise below 80 Hz (which I may have reduced with subsequent adjustments as I'm writing this months after these measurements were recorded.)  It sounds natural.  1/6 octave omni measurements on my phone show less difference below and above 80Hz than these FFT measurements.

Some measurements last year were made before I figured out how to run the supertweeter miniDSP at 96kHz.  So it was running only at 48kHz like my other DSP's (the 48kHz plug in is more flexible than the 96kHz...but that flexibility is not an issue for my supertweeter...so I run only the supertweeter crossover at 96kHz where the highest sampling rate matters).  So measurements through the middle of last year showed a steep dive at 24kHz, the Nyquist Frequency of 48kHz sampling.  It took me awhile even to figure out that sampling rate was a problem, and until I did I greatly feared that I was basically getting zero response above 24kHz.  I remembered that fear into the present.  By that time I was thinking the microphone calibration explained the rolloff above 22kHz...but in fact the peak there, and the general response, shown by the Dayton...extending the highest calibration to all higher frequencies...is very close to that shown by the M50, with hard-to-see difference at 40kHz where useable response is ending with a precipitous roll off.  Also the Dayton curve shows a little notch where the calibration ends, whereas the M50 shows a smooth peak at 22kHz.