• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

PM-H2 headphone amplifier + distortion test files added

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,602
Likes
10,770
Location
Prague
PM-H2 headphone amplifier

PM-H2_sm.jpg


1. Introduction

PM-H2 is a kind of modification of my Dispre 2-JFET preamplifier. It was designed in 2011. The main difference is that PM-H2 uses one pair of Toshiba 2SK170 devices in the differential input stage, not the complementary-differential input quadruple that was used in the Dispre 2-JFET. The main reason of this change was a global lack of 2SJ74 Toshiba JFET devices, that became very hard to buy and genuine parts are almost unobtainable.

PM-H2 is basically used as a headphone amplifier.

PM-H2 has single differential JFET input topology biased by constant current source.

Output stage idle current is set at about 25mA, and it works in a class A up to 5V into 100 ohm load. For standard load 600 ohm, the output stage works in a class A for the whole range of output voltage up to clipping.

To remove output DC voltage component, DC servo is used.

PM-H2 has wide bandwidth and high slew rate.

2. Circuit description

pm-h2_1.png


Drawings of “PM-H2” show circuit diagrams of the preamp. Let us speak about left channel only, the right channel is identical and its part numbers are same as in the left channel +100.

X1 terminal block is an input of audio signal. It is followed by an input divider R1, R2. To get overall preamp gain of +1, please make R1=8k2, R2=2k7. In case you need voltage gain (up to 5x), please follow further text.

P1 potentiometer - it strongly recommend to use a good quality pot, as is the blue Alps.

Then there is an input RC filter (R3, C1), followed by input JFETs Q1, Q2. They constitute differential N-JFET input stage. Its operating current is set by multiple current mirror Q3, Q4, Q6 at 4mA per pair, i.e. 2mA per transistor.

Output current from this input stage is converted to voltage by Q5 loaded by active load Q6 – constant current source. This Q5, Q6 circuit has quite high output impedance, so it is followed by diamond buffer, Q7 – Q12, with very high input impedance and very low output impedance. Output impedance of the whole preamplifier is defined by R26 at 2.7 ohm.

R14 and R13 create quite strong global negative feedback, and they define gain of the preamp referred to input of R3.

Output voltage of the preamplifier is connected to both terminal blocks X2. The output is single-ended. Sheet 2 shows DC servo.

pm-h2_2.png

Comments on parts used and assembly

JFETs are used with Idss range from 6mA to 12mA for Q1, Q2 (BL grade). For best results, selected/matched are these 2 devices to have as same Idss as possible. I recommend matching at +/-0.2mA tolerance. No worse than +/-0.5mA, this would be the worst-case limit. Setup fixture for JFET Idss selection is described on my web page

http://web.telecom.cz/macura/testfet.pdf

No special adjustment or resistor selection is needed, except for input divider – see later.

Input divider

R1 and R2 (input divider) are selectable. The maximum gain is +5, with R1 shorted and R2 unused. The following table shows possible values of R1 and R2 and the resulting overall gain. The table is valid only for 10k potentiometer.

1614970274421.png



Output stage – diamond buffer

To check idle current value, voltage drop across R24 resistor is measured. Voltage drop (in mV) divided by 2.7 (R24 value) gives idle current in mA. To increase idle current, slightly decrease R18 value. To decrease idle current, increase R18. Idle current to be measured 5 minutes after power supply turn on. Before you set up idle current, please solder all the components including DC servo, and check that DC output voltage (X2-4 terminal) is close to 0V, no higher than several mA.

3. Further comments

Potentiometer P1


Recommended is Alps 27mm, type RK27112, 2 x 10k log.

Caution: body of Alps pot must be connected to PCB ground pin named as LSP1, located right from potentiometer. This can be done by short wire placed under potentiometer nut. The opposite end of this wire is to be soldered into LSP1 hole. If these directions are not followed, much worse inter-channel crosstalk will be obtained and the pot will be an antenna for RF/EMI signals.

DC servo (IC1, IC2)

OPA134, OP177, OP07 were verified. TL071 can be used in case of misery.

Heatsinks (KK1 – KK4)

Heatsinks used for BD139/BD140 transistors are SK 95 25 2 x M3 type, supplied by Fischer Elektronik. It is not absolutely necessary to use this type, but it is an optimum for mechanical fixing.

Power supply voltage

2 x 15 V stabilized. Current consumption per board is 2 x 100mA (standby), power supply should be able to deliver at least 2 x 200mA. The ZT15VA power supply with 70053K transformer is a good choice.

Signal and supply connection

INPUT

X1-1 left channel live
X1-2 left channel ground
X1-3 right channel ground
X1-4 right channel live

OUTPUT

X2-4 left channel live
X2-3 left channel ground
X2-2 right channel ground
X2-1 right channel live

SUPPLY

X3-1 +15V
X3-2 0V (ground)
X3-3 -15V

Technical parameters
  • Output voltage max. 27 Vp-p / 600 ohm (9.54 Vrms)
  • Output voltage max. 8.54 Vrms / 50 ohm (P = 1.46W)
  • Load impedance - can drive even 24 ohm headphone
  • S/N at 50 mV / 600 ohm 80 dB (BW = 24kHz)
  • Freq range (-3dB) 0.1 Hz – 170 kHz
  • THD at 1Vrms/50 ohm - less than 0.002% through audio range
  • CCIF IMD 19+20kHz at 1Vrms/50 ohm - less than 0.002%
  • Slew rate >30 V/us
  • Input RC filter prevents from slew rate limitation
  • Gain - adjustable from 1x (0 dB) to 5x (14 dB) by resistors in a voltage divider
  • Input impedance – 10 kohm
  • Output impedance – 2.7 ohm
  • Supply/consumption - 2 x 15V stabilized, 2 x 100 mA
  • Power supply requirements - 2 x 15V stabilized, at least 2 x 200 mA

Measurements

PM-H2_50R__swing.jpg

Voltage swing into 50 ohm load

PM-H2_thdampl.png

THD(+N) vs. output voltage at kHz into 50 ohm and 20 kohm load. Measurement limited to 1V because of not enough precision of my system.

PM-H2 step response.png

Full swing step response (square wave response)

1614972913113.png

Response to 40 Hz square wave
 
Last edited:
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,602
Likes
10,770
Location
Prague
My attempt to measure distortion vs. output voltage and load resistance. We can see no issues with 33 ohm load and higher. With 15 ohm load, distortion gets higher and clipping is at 4.1 Vrms. Due to limited source voltage, I could not reach clipping with load 33 ohm and higher.

PM-H2_thdampl_load6.png
 
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,602
Likes
10,770
Location
Prague
In a combo with Topping D10s DAC. This DAC does not have a headphone output and with PM-H2 providing such output it sounds great. Fantastic resolution and clarity.

P1040791-2.JPG
 
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,602
Likes
10,770
Location
Prague
Hello everyone,

I have decided to prepare test files to check audibility of harmonic distortion on your headphones and headamps. The files can be downloaded from


The files are monophonic, 44.1kHz/24bit, with 1kHz fundamental frequency. The level is -2dBFS to prevent digital clipping. 2nd harmonic distortion is added with level -20dB, -40dB, -60dB and -80dB referred to 0dBFS. The resulting distortions are pure tone, 8%, 0.8%, 0.08% and 0.008%. Please feel free to find you threshold of hearing of H2 distortion with your system and your ears.
The files were created with Paul's @pkane Distort software - thanks for your efforts, Paul!
The test is best performed in foobar ABX and it would give a test report like this

foo_abx 2.0.2 report foobar2000 v1.4.8 2022-03-09 08:22:07 File A: 1kHz_orig_24.wav SHA1: c09dbc2f4aa3c793dda2cc643ffb310a42862817 File B: 1kHz_H2-40dB.wav SHA1: 27e0530be259a9f084722b0b301eb14e5a60b83c Output: ASIO : Topping USB Audio Device Crossfading: NO 08:22:07 : Test started. 08:22:16 : 01/01 08:22:21 : 02/02 08:22:25 : 03/03 08:22:30 : 04/04 08:22:35 : 05/05 08:22:39 : 06/06 08:22:44 : 07/07 08:22:48 : 08/08 08:22:53 : 09/09 08:22:57 : 10/10 08:23:00 : 11/11 08:23:05 : 12/12 08:23:09 : 13/13 08:23:14 : 14/14 08:23:19 : 15/15 08:23:23 : 16/16 08:23:23 : Test finished. ---------- Total: 16/16Probability that you were guessing: 0.0% -- signature -- 1e8239e981ca0c1ec0b959408f2f3da5cb4b64e

My result with the H2-60dB (THD = 0.08%) file is 15/16, still excellent and also surprising to me. In this case, all the components in the audio chain, i.e. DAC, headphone amp, headphones and ears do matter.

foo_abx 2.0.2 report foobar2000 v1.4.8 2022-03-09 08:45:33 File A: 1kHz_orig_24.wav SHA1: c09dbc2f4aa3c793dda2cc643ffb310a42862817 File B: 1kHz_H2-60dB.wav SHA1: 5685ff73c46eb1d09a89eb3536c3a19099852efe Output: ASIO : Topping USB Audio Device Crossfading: NO 08:45:33 : Test started. 08:46:22 : 01/01 08:46:31 : 02/02 08:46:40 : 03/03 08:46:54 : 04/04 08:47:11 : 05/05 08:47:19 : 06/06 08:47:26 : 07/07 08:47:34 : 07/08 08:47:41 : 08/09 08:47:55 : 09/10 08:48:02 : 10/11 08:48:08 : 11/12 08:48:16 : 12/13 08:48:28 : 13/14 08:48:39 : 14/15 08:48:55 : 15/16 08:48:55 : Test finished. ---------- Total: 15/16 Probability that you were guessing: 0.0% -- signature -- f93822e1e74026430c1c72e08019a56f97f609a1
 
Last edited:

pkane

Master Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
5,699
Likes
10,386
Location
North-East
Hello everyone,

I have decided to prepare test files to check audibility of harmonic distortion on your headphones and headamps. The files can be downloaded from


The files are monophonic, 44.1kHz/24bit, with 1kHz fundamental frequency. The level is -2dBFS to prevent digital clipping. 2nd harmonic distortion is added with level -20dB, -40dB, -60dB and -80dB referred to 0dBFS. The resulting distortions are pure tone, 8%, 0.8%, 0.08% and 0.008%. Please feel free to find you threshold of hearing of H2 distortion with your system and your ears.
The files were created with Paul's @pkane Distort software - thanks for your efforts, Paul!
The test is best performed in foobar ABX and it would give a test report like this

foo_abx 2.0.2 report foobar2000 v1.4.8 2022-03-09 08:22:07 File A: 1kHz_orig_24.wav SHA1: c09dbc2f4aa3c793dda2cc643ffb310a42862817 File B: 1kHz_H2-40dB.wav SHA1: 27e0530be259a9f084722b0b301eb14e5a60b83c Output: ASIO : Topping USB Audio Device Crossfading: NO 08:22:07 : Test started. 08:22:16 : 01/01 08:22:21 : 02/02 08:22:25 : 03/03 08:22:30 : 04/04 08:22:35 : 05/05 08:22:39 : 06/06 08:22:44 : 07/07 08:22:48 : 08/08 08:22:53 : 09/09 08:22:57 : 10/10 08:23:00 : 11/11 08:23:05 : 12/12 08:23:09 : 13/13 08:23:14 : 14/14 08:23:19 : 15/15 08:23:23 : 16/16 08:23:23 : Test finished. ---------- Total: 16/16Probability that you were guessing: 0.0% -- signature -- 1e8239e981ca0c1ec0b959408f2f3da5cb4b64e

My result with the H2-60dB (THD = 0.08%) file is 15/16, still excellent and also surprising to me. In this case, all the components in the audio chain, i.e. DAC, headphone amp, headphones and ears do matter.

foo_abx 2.0.2 report foobar2000 v1.4.8 2022-03-09 08:45:33 File A: 1kHz_orig_24.wav SHA1: c09dbc2f4aa3c793dda2cc643ffb310a42862817 File B: 1kHz_H2-60dB.wav SHA1: 5685ff73c46eb1d09a89eb3536c3a19099852efe Output: ASIO : Topping USB Audio Device Crossfading: NO 08:45:33 : Test started. 08:46:22 : 01/01 08:46:31 : 02/02 08:46:40 : 03/03 08:46:54 : 04/04 08:47:11 : 05/05 08:47:19 : 06/06 08:47:26 : 07/07 08:47:34 : 07/08 08:47:41 : 08/09 08:47:55 : 09/10 08:48:02 : 10/11 08:48:08 : 11/12 08:48:16 : 12/13 08:48:28 : 13/14 08:48:39 : 14/15 08:48:55 : 15/16 08:48:55 : Test finished. ---------- Total: 15/16 Probability that you were guessing: 0.0% -- signature -- f93822e1e74026430c1c72e08019a56f97f609a1

Good job, Pavel! Hearing distortion at -60dB is not that easy, as you can see in a similar (preference, not audibility) test by Archimago using DISTORT:


I can't claim any sort of "golden ears" title, but when Archi sent me the files for preview before he published them, unexpectedly I ranked them perfectly by the amount of harmonic distortion in a blind test. But that was only on four tracks and wasn't an audibility comparison, just preference ranking ;)
 
  • Like
Reactions: pma
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,602
Likes
10,770
Location
Prague
Good job, Pavel! Hearing distortion at -60dB is not that easy, as you can see in a similar (preference, not audibility) test by Archimago using DISTORT:

Hi Paul, thank you, in fact I am able to resolve -60dB H2 distortion only on a sine wave and also depending on headphones and DAC and headamp used. Though (or just because) I am old (67 this year), I know what level to set in such test :). This has to be tested at quite low SPL, where the ear masking allows to tell such low distortion. I wonder if the current auditory masking curves would be valid forever, or if an update should be done based on much better circuits and drivers we have available now.
 
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,602
Likes
10,770
Location
Prague
Though I have improved my measuring system, I am still unable to measure a think like this, unless it is pushed to the limits with load below 33 ohm. With 600 ohm, I measure rather a system limit than the headphone amplifier. I would need an AP for this ....

PM-H2_THD_level_1kHz.png

THD vs. level at 1kHz
 
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,602
Likes
10,770
Location
Prague
I have some more measurements with my "improved" but still insufficient measuring system. Without autoranger, it is difficult.

THD from 20mV to 1V
PM-H2_1kHz_15R-300R_1.png


THD from 500mV to the limit set by DAC output voltage
PM-H2_1kHz_3R9-300R_2.png


THD spectra
PM-H2_THD_1kHz_850mV.png


PM-H2_THD_1kHz_47R.png


CCIF IMD (distortion affected by ADC parameters)
PM-H2_CCIF_19+20kHz.png


PM-H2_CCIF_19+20kHz_47R.png
 
Top Bottom