Nitsch ECP Audio DSHA-3FN technical measurements

Nitsch ECP Audio DSHA-3FN 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 technical measurements posts.

The data presented were collected as follows:
1. PrismSound dScope III, picoscope 5243B, Keysight 34465A, Tektronix ADA400A balanced probe
2. Balanced XLR cables Belden 1800F with Neutrik 110R AES connectors (if used)
3. Single Ended cables Audioblast HQ-1 3 ft with Rean RCA connectors
4. 32 and 300 ohm loads used for measurements
5. dScope analyzer sample rate 48 KHz unless otherwise noted
6. 0dBu level used for testing unless otherwise noted
7. Amplifier input to output gain set for 0 dB unless otherwise noted
8. Audioquest Forest and Schiit Pyst USB cables used with measurement equipment
9. Vaunix Lab Brick USB hub
10. Shielded 14AWG and 16AWG power cables
11. ESD, EMI/RFI controlled lab bench and workspace

Measurements are made in accordance with AES17:2015

Sensitivity data for two headphones to keep in mind while viewing these measurements:
HD 650 impedance 300R, sensitivity 98 dB/mW
HE-500 impedance 38R, sensitivity 89 dB/mW

SPL levels for above headphones for reference:
0 dBu 300R 2.00 mW - 101 dBSPL @ 98dB/mW
0 dBu 30R 20.00 mW - 102 dBSPL @ 89dB/mW

All testing performed at 0 dBu unless otherwise noted.
This level is consistent with listening to headphones (referenced above) at 90 dBSPL average with peaks to 100 dBSPL, if the music has 10 dB Peak to Avg ratio. That is LOUD for long listening sessions.

DSHA-3FN
Measurements commenced after 1 hour of warmup each session.
Measurements were performed over a period of several months.

Index
Post 1 – Measurement setup description, highlights
Post 2 – Bal input 300 ohm load
Post 3 – Bal input 32 ohm load
Post 4 – SE input 300 ohm load
Post 5 – SE input 32 ohm load
Post 6 – Square wave response
Post 7 – 19+20 KHz IMD
Post 8 – reserved for corrections and / or additional data

Measurement setup picture:

Standard setup with DSHA-3FN for 300R and 32R loads
Tektronix ADA400A differential probe to the right of the DSHA-3FN

Listening evaluation picture:

DSHA-3FN inputs switched between Holo KTE Spring 2 and Spring 3 Bal & SE outputs. Listening evaluation amp volume set for 0 dB gain with Goldpoint SA2X and SA2 external stepped attenuators employed to set level.
There are plenty of impressions threads on the forum where more detailed and nuanced thoughts about the various components are discussed. For my preferences, either I like how a system renders music or I don’t. I enjoy this amp in this system, a lot.

WARNING: DSHA-3FN has balanced differential outputs. Only fully balanced connections between headphones, IEMs are to be used. 4 wire to 3 wire adapters WILL DESTROY the amplifier output. Don’t do this.

Notable highlights:
First, well done Doug and CeeTee! This is an incredible headphone amplifier.
Nearly perfect gain linearity spanning over 110 dB range in balanced input operation
± 1 dB gain linearity over 120 dB range in balanced input operation
SNR greater than 123 dB in balanced input operation
Excellent square wave response
Bandwidth: DC to greater than 168 KHz
Clarity and tube magic yet no tubes

Due to some very interesting attributes of this amp I will provide additional commentary and interpretation from which I would normally refrain.

Let's get this out of the way.
Generator level set to 0 dBu, Amp set for 0 dB gain with 300R load
Balanced input:

THD+N: 0.01297% Left channel (SINAD 77.7 dB)
THD+N: 0.01179% Right channel (SINAD 78.6 dB)

Those who worship THD+N / SINAD should stop at this point and dismiss this amp. The remainder will be a waste of time. However if you consider there is more to an amp than SINAD / THD+N please continue. This special sounding amp offers another opportunity to demonstrate how chasing low distortion numbers out of context is the very definition of foolishness.

Numerical distortion specification for DSHA-3FN:
THD+N < 0.015% at 1 KHz, 0 dBu, 300R load

This specification represents an extremely small bit of information with respect to the amplifier performance. Published THD+N has been around since the beginning of audio measurements. So it continues to be published out of tradition.

ECP Audio DSHA-3FN Distortion at 1 KHz, 0 dBu, 300R load

In this measurement suite the THD+N measures as expected verifying the specification. However more metrics have been added along with the L and R channel FFTs. Observe that 3rd harmonic is the major contributor to THD+N at 0 dBu. Also note how 60 Hz AC Hum is observed to be -127 dBu in the L ch and -124 in the R ch, and aggregate noise is approximately -108 dBu in both channels yielding a very low noise background, essentially black.

ECP Audio DSHA-3FN Distortion at 1 KHz, 0 dBu, 300R load L ch

In this graphic observe how the numerical specification denotes a singular operating point performance. Very limited information.

Consider expansion of the performance profile by sweeping over a range of levels and decomposing the THD+N into a few specific constituent measures.

ECP Audio DSHA-3FN Distortion vs amplitude 1KHz 300R -40 to +10 dBu L ch

This expanded measurement reveals DSHA-3FN to have a much more complex behavior which cannot possibly be inferred by the numerical specification. In the graph 3rd harmonic distortion, D3 varies over level and is the dominant contributor to THD+N above approximately -17 dBu. Below that level PS Noise becomes the dominant contributor to the THD+N metric.

What happens above +10 dBu with a 300R load?
ECP Audio DSHA-3FN Distortion vs amplitude 1KHz 300R above +10 dBu L ch

At these high levels, way too loud for anything but very short moments of listening, D3 is the dominant contributor to THD+N until +15 dBu where D2 becomes the dominant contributor up to the onset of clipping distortion. PS Noise contribution is insignificant.

While the graph below is for a 32R load, the behavior of the PS Noise is representative for a range of loads.
ECP Audio DSHA-3FN Distortion vs amplitude 1KHz 32R -40 to +10 dBu L ch

Trace ID:
Yel: THD+N
Grn: 2nd harmonic distortion (D2)
Brn: 3rd harmonic distortion (D3)
Red: 4+HD+N (crap factor including AC mains noise)
Blu: FFT 1 KHz -20 dBu 32R load

In the graph 3rd harmonic distortion, D3 varies over level and is the dominant contributor to THD+N above approximately -16 dBu. Below that level PS Noise becomes the dominant contributor to the THD+N metric. At approximately -17 dBu D3 drops below D2 and becomes significantly lower.

PS Noise is the greatest contributor to THD+N below approximately -16 dBu, observe the PS Noise comprises of 60, 120, 180, 420 and 660 Hz. It is not a single large spike but spread out over the fundamental and 4 harmonics. THD+N captures the aggregate of these PS Noise frequencies but doesn’t provide sufficient information as to audibility. All those noise spikes are at or below -119 dBu which will be inaudible on even the most sensitive IEMs.

Note about threshold of AC Mains noise perception
In my acoustic lab when listening between 65 and 75 dB SPL on headphones with sensitivity of 95 to 100 dB/mw, mains noise of -75 dBu is at the threshold of perception. At this level I perceive the blackground compromised and a certain sourness to the listening experience that would mostly disappear when listening level was increased to 75 to 85 dB SPL. Mains hum between tracks is still evident. In normal home environments ambient noise masking is likely sufficient to hide this issue if AC mains noise is at or below -75 dBu. If AC mains noise is at or below -85 dBu then I no longer perceive it with the headphones of the sensitivity mentioned in my very quiet acoustic lab. Most listeners will be undisturbed by this issue to which I am so sensitive. DSHA-3FN PS Noise spikes are at or below -119 dBu which is 44 dB lower than the threshold of audibility for most headphones.


ECP Audio DSHA-3FN Distortion vs Frequency Bal input 0 dBu level 300R load

Trace ID:
Yel: THD+N
Grn: 2nd harmonic distortion (D2)
Blu: 3rd harmonic distortion (D3)
Red: 4+HD+N (crap factor including AC mains noise)
Distortion vs frequency also demonstrates a far more complex behavior for this amp than possible from numerical THD+N spec.

ECP Audio DSHA-3FN Dynamic Range Bal input 300R load

Impressive performance! When a system is gain staged correctly this covers threshold of hearing to threshold of discomfort, rock concert levels.

DSHA-3FN Output Impedance

Output impedance is 4.9 Ω

 

Bal input 300 ohm load

DSHA-3FN Distortion FFT 300 ohm load Bal input


DSHA-3FN Distortion vs Amplitude 300 ohm load Bal input


DSHA-3FN Distortion vs Frequency 300 ohm load Bal input


DSHA-3FN 50 + 7000 Hz 300 ohm load Bal input - Left Channel


DSHA-3FN Gain Linearity 300 ohm load Bal input - Left Channel


DSHA-3FN THD+N vs Frequency 300 ohm load Bal input - Left Channel


DSHA-3FN Residual Noise 300 ohm load Bal input - Left Channel


DSHA-3FN Multitone 300R 0 dBu Bal input - Left Channel


DSHA-3FN Multitone 300R -10 dBu Bal input - Left Channel

When measuring at a level consistent with approximately 90 dB SPL for many popular headphones observe the significant drop in distortion spikes between the source tones. Very clean performance.

Complete Bal input 300 ohm load measurement report pdf attached

 

Bal input 32 ohm load

DSHA-3FN Distortion FFT 32 ohm load Bal input


DSHA-3FN Distortion vs Amplitude 32 ohm load Bal input


DSHA-3FN Distortion vs Frequency 32 ohm load Bal input


DSHA-3FN 50 + 7000 Hz 32 ohm load Bal input - Left Channel


DSHA-3FN Gain Linearity 32 ohm load Bal input - Left Channel


DSHA-3FN THD+N vs Frequency 32 ohm load Bal input - Left Channel


DSHA-3FN Residual Noise 32 ohm load Bal input - Left Channel


DSHA-3FN Multitone 32R 0 dBu Bal input - Left Channel


DSHA-3FN Multitone 32R -10 dBu Bal input - Left Channel

When measuring at a level consistent with approximately 90 dB SPL for many popular headphones observe the significant drop in distortion spikes between the source tones.

Complete Bal input 32 ohm load measurement report pdf attached

 

SE input 300 ohm load

Single ended input will lose the advantages a Balanced input topology provides. Noise floor will be higher, gain linearity range reduced, dynamic range reduced, etc. However, transient response, frequency response, multitone performance will remain relatively the same. Given the exceptionally high performance experienced using Balanced input, the convenience of SE input should result in hardly any audible difference.

DSHA-3FN Distortion FFT 300 ohm load SE input


DSHA-3FN Distortion vs Amplitude 300 ohm load SE input


DSHA-3FN Distortion vs Frequency 300 ohm load SE input


DSHA-3FN 50 + 7000 Hz 300 ohm load SE input - Left Channel


DSHA-3FN Gain Linearity 300 ohm load SE input - Left Channel


DSHA-3FN THD+N vs Frequency 300 ohm load SE input - Left Channel


DSHA-3FN Residual Noise 300 ohm load SE input - Left Channel


DSHA-3FN Multitone 300R 0 dBu SE input - Left Channel


DSHA-3FN Multitone 300R -10 dBu SE input - Left Channel

When measuring at a level consistent with approximately 90 dB SPL for many popular headphones observe the significant drop in distortion spikes between the source tones.

Complete SE input 300 ohm load measurement report pdf attached

 

SE input 32 ohm load

DSHA-3FN Distortion FFT 32 ohm load SE input


DSHA-3FN Distortion vs Amplitude 32 ohm load SE input


DSHA-3FN Distortion vs Frequency 32 ohm load SE input


DSHA-3FN 50 + 7000 Hz 32 ohm load SE input - Left Channel


DSHA-3FN Gain Linearity 32 ohm load SE input - Left Channel


DSHA-3FN THD+N vs Frequency 32 ohm load SE input - Left Channel


DSHA-3FN Residual Noise 32 ohm load SE input - Left Channel


DSHA-3FN Multitone 32R 0 dBu SE input - Left Channel


DSHA-3FN Multitone 32R -10 dBu SE input - Left Channel

When measuring at a level consistent with approximately 90 dB SPL for many popular headphones observe the significant drop in distortion spikes between the source tones.

Complete SE input 32 ohm load measurement report pdf attached

 

Square wave response

Amplifier is adjusted to produce 0 dB gain with a 1 KHz sine wave for 300 ohm load measurements. This calibration is performed again for 32 ohm load measurements. Measurements are made with at 20 Hz square wave. SG = Signal Generator

Bandwidth Estimation formula:
BW (MHz) = 0.35 / RT (mS)
Where RT = 10 to 90% Rise Time

DSHA-3FN SG 20Hz sqr 2Vpp 500nS div 5MHz fltr 300R Bal

0.35 / 2.079 uS = 168.4 KHz

Transient response is exceptional. Percussive sounds benefit from this behavior.


There is a large overshoot with an exceptionally short ring time which will be damped by headphones due to limited ability to respond above 40 KHz.

The following square wave measurements demonstrate essentially the same performance as for Balanced input 300R load shown above.

DSHA-3FN SG 20Hz sqr 2Vpp 500nS div 5MHz fltr 32R Bal

0.35 / 2.127 uS = 164.6 KHz



DSHA-3FN SG 20Hz sqr 2Vpp 500nS div 5MHz fltr 300R SE

0.35 / 2.053 uS = 170.5 KHz



DSHA-3FN SG 20Hz sqr 2Vpp 500nS div 5MHz fltr 32R SE

0.35 / 2.085 uS = 167.9 KHz

 

19 + 20 KHz IMD CCIF

DSHA-3FN 19+20KHz IMD CCIF 300R Bal input

IMD is exceptionally low. Insignificant distortion.

DSHA-3FN 19+20KHz IMD CCIF 300R Bal input



DSHA-3FN 19+20KHz IMD CCIF 32R Bal input



DSHA-3FN 19+20KHz IMD CCIF 300R SE input



DSHA-3FN 19+20KHz IMD CCIF 32R SE input

 

Clarification about DSHA-3FN Balanced vs unBalanced (SE) input

It isn't the Amp (nor source) but the connection method that has different CMRR immunity. Bal inherently can achieve 120 dB while unBal often only makes 100 dB, 20 dB less. Loss of measurement performance is due to the connection method, not the Amp or source. Residual noise levels being discussed are approximately 775 nV for Bal vs 7.75 uV for SE. The measurements are in dBu which is an absolute level. Most headphones will only produce between 1 and 10 dB SPL at 7.75 uV and only a few IEMs might produce around 30 to 40 dB SPL at such levels. Also given the noise is mostly 660 Hz and below, it is doubtful even listeners with the highest sensitivity IEMs would detect any PS Noise thanks to basilar membrane performance as depicted on ISO Contours of equal loudness.



In the diagram above unBalanced has an ambient E field noise impinged on the signal. DSHA-3F operating in SE mode simply connects the input transformer appropriate for a SE connection. It is the connection method that has less noise immunity, not the amp.

Same E field noise impinging on the Balanced connection is common to both 0° and 180° signals. In full differential mode the noise effectively cancels.

 

added a clarification about DSHA-3FN Balanced vs unBalanced (SE) input to post 8 above.

 

What is the maximum power output of this amp in ma at the 32 ohm and 300 ohm settings?

 

DSHA-3FN power discussion.

How much power does this amp produce at various loads?
How much power does a given headphone need to produce a specific SPL?
What SPL does a listener really need?

Clipping begins typically where distortion measurements indicate a sudden, sharp rise. Specifications often follow the rise to 1% or 10% and note the power at those levels. For sake of avoiding clipping the knee in the distortion curve denotes clipping onset.

The following two charts appear in the original post:

With a 300 Ω load clipping onsets at +26 dBu.


With a 32 Ω load clipping onsets at +6 dBu.

The following equations compute transformation of dBu to Vrms, then Vrms for a given load R into Power:

Using the above equations Vrms is calculated to be 15.46 for a 300 Ω load. This computes to power of 797 mW
In a like manner Vrms is calculated to be 1.55 for a 32 Ω load. This computes to power of 75 mW


Using the two power levels calculated and assuming a linear correlation, the graph above was created. Remember that the DSHA-3FN performs essentially the same as a DSHA-3F and the DSHA-3F was originally intended to bring out the best performance for Focal headphones, such as Elex, Clear, Utopia.

As an example Focal Clear have a nominal impedance of 55 Ω. From the chart it may be inferred a Pmax of 140 mW.


Using the chart above Focal Clear require 0.4 mW to reach 100 dB SPL at 1 KHz. The DSHA-3FN can provide a Pmax of 140 mW to the Clears. 350 times more power than necessary to cleanly deafen the listener. DSHA-3FN will produce more than enough clean power for most of the headphones and IEMs on the chart above.

HiFiMan HE-6 needs 50 mW to achieve 100 dB SPL at 1 KHz. DSHA-FN can produce 120 mW into 50 Ω, about 2.4x necessary. Headroom is lower so for very loud listening HE-6 users may want more power. Personally I listen at 70 to 80 dB SPL which is more than adequately powered by DSHA-3FN. But some listeners like their sound VERY LOUD and this amp might be perceived to be inadequate. Keep in mind OSHA and NIOSH recommended time exposures to loud sound levels to remain in safe operating areas or risk hearing damage.

 

Thank you for the detailed answer Atomic Bob!