Goldpoint SA1X and SA2X passive attenuator technical measurements

Discussion in 'Portable and Other Gear Measurements' started by atomicbob, Jan 6, 2019.

  1. atomicbob

    atomicbob dScope Yoda

    Pyrate BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    18,653
    Trophy Points:
    113
    Location:
    On planet
    Goldpoint SA1X and SA2X passive attenuator technical measurements

    Standard Prologue
    If you are unfamiliar with audio measurements please use a search engine with the query:
    "audio measurements" or "audio measurement handbook"
    Look for publications by Richard C. Cabot and also by Bob Metzler, both from Audio Precision. There are other useful publications as well. These will provide basic knowledge.
    Interpretation of the following measurements is beyond the scope of this post.

    The data presented were collected as follows:

    1. PrismSound dScope III, picoscope 5243B, Atten ATF-20B signal generator, Keysight 34465A DMM
    2. Balanced cables Canare L-4E6S starquad and Belden 1800F with Neutrik XLR connectors
    3. Single Ended cables Audioblast HQ-1 1.5 ft with Rean RCA connectors or RG6 with BNC connectors
    4. 75R RCA to BNC adapters as necessary
    5. Vaunix Lab Brick USB hub for measurement equipment
    6. Audioquest Forest and Schiit Pyst USB cables used for measurement equipment
    7. 14 and 16 AWG sheilded power cables used

    All testing performed at 0 dBu input and 30 dB attenuation unless otherwise noted.


    Highlights
    It should be no surprise the internet abounds with armchair engineers spewing incredible quantities of misleading bovine excrement. This measurement set will hopefully offer some enlightenment.
    A passive attenuator is simply a potentiometer or stepped attenuator forming a resistor divider to reduce the input signal by some desired amount.

    Several passive attenuator myths will be addressed:
    1. reduces audio band frequency response
    2. reduces audio transient response
    3. increases distortion
    4. reduces dynamic range
    5. increases noise
    6. introduces channel imbalance - this is matter of cost / performance tradeoff. Low cost potentiometers can introduce channel imbalances especially at greatest attenuation (lowest volume settings.) Well designed potentiometers and stepped attenuators will not have this issue.

    Picture of the Audio Lab setup for Goldpoint SA1X and SA2X measurements
    00 20181117 SA1X - SA2X measurement seteup.jpg

    Passive Attenuator part 1 of 3

    SA1X amplitude - phase - gain (A01) 100K load
    01 20181117 SA1X A01 amplitude - phase - gain 30 dB atten.png
    From this measurement it appears a well designed passive attenuator achieves less than approximately 0.005 dB channel imbalance. Myth #6 busted for well designed potentiometers and stepped attenuators (read: not cheap.)


    SA1X THD+N THD nth harmonic distortion (A04) FFT 0dBu + 30dBu attenuation 100K load
    02 20181117 SA1X A04 THD+N  THD  nth-HD FFT 30 dB atten.png
    Y axis range has been lowered to include -140 dBu in the FFT. Distortion measurements demonstrate exceedingly low distortion, at the same level as that of interconnecting wire. Myth #3 busted.


    SA2X signal to noise (A07) 100K load
    03 20181117 SA1X A07 SNR 30 dB atten.png
    Signal to Noise is also at the same level as interconnecting wire. Myth #4 busted.


    SA2X Crosstalk (A08 A09) 100K load
    04 20181117 SA1X A08 A09 Crosstalk 30 dB atten.png
    Two cables run parallel will exhibit approximately the same level of crosstalk. This again is exceedingly low.


    SA1X 1 KHz spectrum Left and Right 0dBu + 30dBu attenuation 100K load
    05 20181117 SA1X 1 KHz THD+N THD FFT 0dBu 30 dB atten.png
    Audio band residuals are below -130 dBu. No distortion products visible.


    SA1X 50 Hz + 7000 Hz spectrum Left and Right 0dBu + 30dBu attenuation 100K load
    06 20181117 SA1X 50+7000Hz 0dBu 30 dB atten - 256K FFT.png
    Y-axis changed to include -150 dBu. Output of anything other than 50 and 7000 Hz in the audio band is less than approximately -138 dBu, which is 97.6 nV or less than 100 billionths of a volt.


    SA1X 600 Hz + 1700 Hz spectrum Left and Right 0dBu + 30dBu attenuation 100K load
    07 20181117 SA1X 600+1700Hz 0dBu 30dB atten.png
    Y-axis changed to include -150 dBu. Output of anything other than 600 and 1700 Hz in the audio band is less than approximately -138 dBu.


    SA1X IMD spectrum Left and Right -16dBu + 30dBu attenuation 100K load
    08 20181117 SA1X 19+20KHz -16dBu 30 dB atten.png
    Should be a familiar theme by now. IMD is at the threshold of wire.


    SA1X residual noise spectrum Left and Right A=0dB + 30dBu attenuation 100K load
    09 20181117 SA1X residual noise 30dB atten.png
    What noise? there isn't any more than a wire would produce in the same measurement. Myth #5 busted.
     
  2. atomicbob

    atomicbob dScope Yoda

    Pyrate BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    18,653
    Trophy Points:
    113
    Location:
    On planet
    Passive Attenuator part 2 of 3

    Picture of attenuator gain accuracy setup
    10 20181119 SA2X_F20_small_v2.jpg

    SA2X Gain Linearity vs Volume Knob position
    11 20181119 SA2X gain linearity.png
    Very precised, repeatable attenuation levels for the SA2X 47 level stepped attenuator.

    Frequency Response SPICE Modeling (predictions) with a 48K load
    (Liquid Platinum input impedance):

    Frequency Response model for Canare L-4E6S
    Electrical parameters (per foot):
    46.0 pF
    0.45 uH
    0.0299 ohms

    SA2X frequency response SPICE model 20 dB attenuation Canare L-4E6S 2ft 48K load
    12 20190106 SA2X Canare L-4E6S 2ft Liquid Platinum load - 20 dB attenuation - overview.png
    20 dB attenuation predicted as expected.

    SA2X frequency response SPICE model 20 dB attenuation Canare L-4E6S 2ft 48K load
    Extreme y-axis zoom
    13 20190106 SA2X Canare L-4E6S 2ft Liquid Platinum load - 20 dB attenuation - zoom.png
    At this extreme y-axis zoom the attenuated signal is only 0.001 dB lower than target at 30 KHz and 0.003 dB lower than target at 52 KHz. There is no frequency response degradation. Myth #1 busted.


    Frequency Response model for Belden 1800F
    Electrical parameters (per foot):
    12.0 pF
    0.18 uH
    0.0237 ohms

    SA2X frequency response SPICE model 20 dB attenuation Belden 1800F 2ft 48K load
    14 20190106 SA2X Belden 1800F 2ft Liquid Platinum load - 20 dB attenuation - overview.png
    Again 20 dB attenuation predicted, as expected.

    SA2X frequency response SPICE model 20 dB attenuation Belden 1800F 2ft 48K load
    Extreme y-axis zoom
    15 20190106 SA2X Belden 1800F 2ft Liquid Platinum load - 20 dB attenuation - zoom.png
    At this extreme y-axis zoom the attenuated signal is only 0.001 dB lower than target at 116 KHz and 0.003 dB lower than target at 200 KHz. There is no frequency response degradation. Myth #1 really busted. Digital audio cables with a 110 ohm impedance are quite useful for balanced interconnects and are not expensive. Far better than most boutique interconnects.


    SA1X square wave 20 Hz 2 Vpp 10 mS / div 10K load 100 KHz measurement BW 26 dB attenuation
    16 20181117 SigGen SA1X 20 Hz square 2000mVpp input 26 dB atten 10mS div 200KHz filter.png

    SA1X square wave 20 Hz 2 Vpp 200 nS / div 10K load 5 MHz measurement BW 26 dB attenuation
    17 20181117 SigGen SA1X 20 Hz square 2000mVpp input 26 dB atten 200nS div 10MHz filter - BW calc.png
    Bandwidth estimation: BW (MHz) = 0.35 / RT (mS)
    Where RT = 10 to 90% Rise Time
    0.35 / 0.239 nS = 1.464 MHz

    Myth #2 busted big time. There is no transient response degradation.
     
    Last edited: Jan 6, 2019
  3. atomicbob

    atomicbob dScope Yoda

    Pyrate BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    18,653
    Trophy Points:
    113
    Location:
    On planet
    Real world examples part 3 of 3

    The need:
    Dangerous Music Convert-2 DAC outputs +22 dBu at 0 dBFS when run with Cal of -18 dBFS and output trim OFF.

    Liquid Platinum with volume knob at 1:00 o'clock runs unity gain. To obtain a 0 dBu level, approximately 100 dB SPL peak for HD650, requires 22 dB of attenuation from the Convert-2 output.

    QSC RMX 1450 has 32 dB of voltage gain when input trims are set for 0 dB attenuation. A 0 dBu input will result in 32 dBu output. Used with speakers having sensitivity of 90 dB/W/meter and desiring 100 dB SPL with 110 dB SPL peaks, assuming 10 dB peak to average in the music mix, requires only 31.24 dBu at the output, less than 0 dBu input. 32 dB of attenuation from the Conver-2 output will be necessary to achieve slightly below the desired output and avoid overdriving the QSC amp.


    SA2X 22 dB attenuation measurements with DM Convert-2 DAC set for +22 dBu output

    Convert-2 + SA2X Bal A04 THD THD+N spectrum SA2X set to 0 dB attenuation
    20 20190103-04 Convert-2 Bal A04 THD THD+N spectrum - AES Dante 88K 22dBu - L ch.PNG

    Convert-2 + SA2X Bal A04 THD THD+N spectrum SA2X set to 22 dB attenuation
    21 20190103-03 Convert-2 Bal A04 THD THD+N spectrum - AES Dante 88K 0dBu - L ch.PNG
    There is essentially no difference between those two measurements. Myth #3 busted again.


    Convert-2 + SA2X Bal dynamic range SA2X set to 0 dB attenuation
    22 20190104-14 Convert-2 Bal A04 dynamic range - AES Dante 88K 22dBu.PNG

    Convert-2 + SA2X Bal dynamic range SA2X set to 22 dB attenuation
    23 20190104-13 Convert-2 Bal A04 dynamic range - AES Dante 88K 0dBu.PNG
    Again, there is essentially no difference between those two measurements. Myth #4 busted again.


    Conclusion
    Used intelligently a passive attenuator offers high quality functionality with extremely minimal signal degradation. Don't connect 30 meters of Canare L-4E6S starquad between the attenuator output and amplifier input. Keep it short, preferably less than 2 meters (6.5 feet). Enjoy music and dispense with the nervosa.
     
    • Like Like x 30
    • Epic Epic x 1
    • List
    Last edited: Jan 6, 2019
  4. atomicbob

    atomicbob dScope Yoda

    Pyrate BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    18,653
    Trophy Points:
    113
    Location:
    On planet
    Real world example part 2.

    Dangerous Music Convert-2 represents a first class pro audio DAC. It isn't cheap. Determining optimal configuration for best performance is expected in pro audio settings. Here is a small sampling of the many measurements I have made on the Convert-2 (abbreviated C2 hereafter.)

    Configuration for these measurements:
    AES into C2 AES input 1
    88 KHz sample rate 24 bit
    C2 calibration set for -18 dBFS
    Sescom Belden 1800F 2 ft digital audio cables with Neutrik connectors, both analog and digital connections
    Output trim (internal and external) adjusted to produce close to 0 dBu (typically -0.487 dBu)

    Comparison of C2 internal output trim with Goldpoint SA2X external attenuation.

    Discussed as pairs.

    -90 dBFS Convert-2 internal output trim
    01 20190107-18 convert2 Bal 1 KHz -90 dBFS 24 bit - AES -18dBFS ref 0 dBu out - int atten.PNG

    -90 dBFS Convert-2 output trim off - SA2X external attenuation
    02 20190107-18 convert2 Bal 1 KHz -90 dBFS 24 bit - AES -18dBFS ref 0 dBu out - ext atten.PNG
    Low level resolution has improved with the SA2X, notice less noise on the sinusoid in the second graph.


    Gain Linearity Convert-2 internal output trim
    03 20190107-11 convert2 Bal 1 KHz gain linearity - AES -18dBFS ref 0 dBu out - int atten.PNG

    Gain Linearity Convert-2 output trim off - SA2X external attenuation
    04 20190107-11 convert2 Bal 1 KHz gain linearity - AES -18dBFS ref 0 dBu out - ext atten.PNG
    Gain linearity is better behaved to a lower level in the second graph and also notice channel imbalance displayed in the first graph eliminated in the second.


    Dynamic Range Convert-2 internal output trim
    05 20190107-14 convert2 Bal dynamic range - AES -18dBFS ref 0 dBu out - int atten.PNG

    Dynamic Range Convert-2 output trim off - SA2X external attenuation
    06 20190107-14 convert2 Bal dynamic range - AES -18dBFS ref 0 dBu out - ext atten.PNG
    As noise floor drops dynamic range improves by 6 dB with the SA2X over Convert-2 internal output trim.


    Noise and Interference Convert-2 internal output trim
    07 20190107-20 convert2 Bal noise and interference - AES -18dBFS ref 0 dBu out - int atten.PNG

    Noise and Interference output trim off - SA2X external attenuation
    08 20190107-20 convert2 Bal noise and interference - AES -18dBFS ref 0 dBu out - ext atten.PNG
    Further demonstration of noise floor dropping with an external passive attenuator as compared to the Convert-2 internal output trim.

    Conclusion
    In every comparison above, external passive attenuation performs better than using the Convert-2 internal output trim. Passive attenation, well-implemented, will have exceedingly minimal impact on audio signal quality.

    Additional comments

    Several passive attenuators are employed in my acoustic lab. An incomplete list:
    Coleman Audio M3 monitor controller
    Goldpoint SA1, SA2, SA1X, SA2X
    Schiit Saga, a great passive controller, period. No qualifying statements necessary.
    Audio Technica AT8202

    The selection of the Goldpoint SA2X for this particular set of measurements was one of convenience ... mine. It was sitting on the lab shelf ready for use. Other passive attenuators may provide similar performance. I am not suggesting everyone needs an SA2X. But for a pro audio lab with a collection of expensive, balanced DACs the SA2X makes sense in lab use.




     
    Last edited: Jan 7, 2019
  5. damaged-goods

    damaged-goods Acquaintance

    Joined:
    Feb 8, 2017
    Likes Received:
    77
    Trophy Points:
    28
    Location:
    Europe
    Nice writeup, thanks very much for another excellent set of measurements!

    What is the output impedance the amp sees at 30db attentuation?
     
  6. Psalmanazar

    Psalmanazar Most improved member; A+

    Pyrate Slaytanic Cliff Clavin
    Joined:
    Sep 27, 2015
    Likes Received:
    5,345
    Trophy Points:
    113
    Thanks atomicbob! Using digital cables sounds like a good idea. May you post measurements sometime comparing 110 ohm digital cable to more typical interconnect cables sometime?
     
  7. pedalhead

    pedalhead Friend

    Pyrate
    Joined:
    Oct 2, 2015
    Likes Received:
    2,046
    Trophy Points:
    93
    Location:
    Oxfordshire, UK
    Excellent stuff, cheers @atomicbob . Audiophile-dom needs more myth-busting like this & what Archimago has been doing for a while.
     
  8. Cspirou

    Cspirou They call me Sparky

    Pyrate
    Joined:
    Sep 27, 2015
    Likes Received:
    8,200
    Trophy Points:
    113
    Location:
    Northwest France
    Awesome measurements. If you have time I'd like to know what impedance mismatch looks like. Like a 100k pot with a 10k load
     
  9. Skyline

    Skyline Double-blindly done with this hobby

    Pyrate
    Joined:
    Sep 28, 2015
    Likes Received:
    3,426
    Trophy Points:
    113
    With a Liquid Platinum on the way, I'm VERY interested in this. @atomicbob , have you tried any more budget-friendly options? Balanced input/output is a must.
     
  10. msommers

    msommers High on Epipens

    Pyrate
    Joined:
    Nov 6, 2016
    Likes Received:
    2,750
    Trophy Points:
    113
    Location:
    Calgary, Alberta
    Home Page:
    First, thank you so much for this work!!

    Regarding the Convert-2 + SA2X, 0 vs 22 dB, there does seem to be some differences from 100 Hz and below. Would you characterize these differences as very minor and inaudible?
     
  11. atomicbob

    atomicbob dScope Yoda

    Pyrate BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    18,653
    Trophy Points:
    113
    Location:
    On planet
    Additional measurement data put into post #4 above.

    It depends on several variables such as the value of passive attenuator used (10K, 25K, 50K, 100K, etc) and the amount of attenuation being applied.

    A project in the works for some time:
    20181126 cable testing - sig gen - oscilloscope - LCR meter - small.jpg
    Completion is TBD. So many other tasks in the queue.

    I suggest using the search terms "passive preamp" and "passive attenuator" with combinations of the words audio and balanced.

    I believe you are referring to the Dynamic Range graphs. Less than -120 dBu below 100 Hz is likely not to be a noticeable difference given ISO Contours of Equal Loudness (Fletcher-Munson updated).
     
  12. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    I would not worry about impedance mismatch. For most cable runs and at the frequencies of interest, this should not behave like a transmission line (reflections). Furthermore, we are not concerned about maximizing power delivery. The pot should be thought as part of an electronics network (following electric circuit analysis rules).

    However, usually the pot is 10k and the load is 100k. I believe this is to keep volume control more "linear". Note audio pots are logaritmic so this may not be a big deal. But if you want a quick and dirtly linear plot of voltage delivery for both situations, here you go:

    100k pot w 10k load

    100kpot.png

    10k pot w 100k load

    10kpot.png

    "pot value" is the pot position. Depending on how you look at it it sweeps from 0 to x ohms or the other way around. At 0 ohm "pot value", the pot presents max resistance. I guess I could have done it the otherway around presentation wize, but the results are the same.

    Here is what was done to show interactions between the pot and the load impedances:

    Vin = 10;
    x1 = 0:100:Rpot;
    x2 = Rpot - x1;
    x3 = x2*Rload ./ (x2+Rload);
    Vout = Vin*x3 ./ (x3+x1);

    The load is in parallel with the "active" part of the pot, while there is a voltage divider between the load and pot parallel impedance and the pot "inactive" part, or whatever folks usually call those legs or pot things.

    This maybe why it's easier to find 10k pots than 100k ones. In some ways you could say they make more sense (at least based on the quick pot sweep I did). But dunno.

    This affects volume control, and things are still monotonic. So again, proly not a big deal. Again, specially considering pots are logarithmic anyways.

    I would not OCD on this BTW. Pots and stuff like this border in cable audio nervosa IMO.
     
    Last edited: Jan 7, 2019
  13. HotRatSalad

    HotRatSalad Friend

    Pyrate
    Joined:
    May 4, 2018
    Likes Received:
    1,684
    Trophy Points:
    93
    Location:
    NH
    How does the Saga compare to the Goldpoint head to head ?
    I referred to this
    https://www.superbestaudiofriends.o...aga-preamp-technical-measurements.3780/page-2

    also

    "Additional comments
    Several passive attenuators are employed in my acoustic lab. An incomplete list:
    Coleman Audio M3 monitor controller
    Goldpoint SA1, SA2, SA1X, SA2X
    Schiit Saga, a great passive controller, period. No qualifying statements necessary.
    Audio Technica AT8202"

    IS it worth worth grabbing a goldpoint for balanced alone if you have a Saga ?
     
  14. atomicbob

    atomicbob dScope Yoda

    Pyrate BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    18,653
    Trophy Points:
    113
    Location:
    On planet
    If you have a Saga in your system now then you already have one of the best sounding preamps available. If you suffer from FOMO or balanced nervosa then you may want to explore the Polarity Reverse Trick here and possibly enhance your experience rather than needlessly spend money.
     
  15. msommers

    msommers High on Epipens

    Pyrate
    Joined:
    Nov 6, 2016
    Likes Received:
    2,750
    Trophy Points:
    113
    Location:
    Calgary, Alberta
    Home Page:
    Aside from the obvious taming of volume, would either of these work to tame the hot outputs from pro dacs for hifi devices? Or is a 4:1 transformer required?
     
  16. skem

    skem Friend

    Pyrate
    Joined:
    Nov 1, 2017
    Likes Received:
    1,911
    Trophy Points:
    93
    Location:
    Charles River
    Strongly suggest a transformer if you really need to cut the voltage. The loaner pro DACs, however, have very clean volume controls that work perfectly well to attenuate the output. The Convert-2 volume control is crystal clear. I felt maybe the Solaris volume control might have been adding compression, but I can’t be certain since the DAC had also warmed up a day by the time I got balanced volume control to compare it to.
     
  17. atomicbob

    atomicbob dScope Yoda

    Pyrate BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    18,653
    Trophy Points:
    113
    Location:
    On planet
    Taming volume, taming hot outputs. Same problem. It's just voltage levels. Passive attenuators are one option for reducing voltage. The Convert-2 output trim is a potentiometer acting as a passive attenuator. It has channel imbalance as demonstrated above in the measurements but otherwise works fine. Precision stepped attenuators can be designed with nearly perfect channel balance as shown here .
     
  18. skem

    skem Friend

    Pyrate
    Joined:
    Nov 1, 2017
    Likes Received:
    1,911
    Trophy Points:
    93
    Location:
    Charles River
    Just to add, many people feel resistors used to cut voltage with inherently capacitive outboard lines do result in a loss of perceptual quality. This is why most pre-amps have buffers and why passives are not popular. So the obvious question is why?

    AtomicBob:What does your SPICE simulation of the LRC circuit show for the phase response over the audioband?
     
    Last edited: Feb 10, 2019
  19. atomicbob

    atomicbob dScope Yoda

    Pyrate BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    18,653
    Trophy Points:
    113
    Location:
    On planet
    Please look here for actual measurements demonstrating that a 10K passive attenuator will not have high frequency loss unless using minimal attenuation combined with long, high capacitance lines. Two foot low capacitance cables such as Blue Jeans LC-1 at 6 dB attenuation (worst case) will incur only 0.2 dB loss at 20 KHz. Typical higher attenuation settings will be less than 0.05 dB at 20 KHz.
     
  20. atomicbob

    atomicbob dScope Yoda

    Pyrate BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    18,653
    Trophy Points:
    113
    Location:
    On planet
    Less than 0.16 degrees at 20KHz with 20 dB attenuation into a Liquid Platinum using two feet of Belden 1800F. 20190210 SA2X Belden 1800F 2ft Liquid Platinum load - 20 dB attenuation - phase to 20 KHz.png
     

Share This Page