• 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!

Matrix M-Stage HPA-3B mods

trl

Major Contributor
King of Mods
Joined
Feb 28, 2018
Messages
1,967
Likes
2,523
Location
Iasi, RO
Mod no. 1: I had replaced WIMA MKS 4 (Metallized Polyester) with JB JFX Premium Film Cap (Polypropylene Film) bought recently from TME.
I might be inclined to tell that trebles are a bit more upfront and detailed now, but to be honest without proper A/B testing it's quite hard to tell if any real difference occurred. Anyway, according to audiophile forums & resources Polypropylene is better than Polyester, so I'm pleased that I swapped WIMA MKS with JB MKP (I was tempted to use MIFLEX MKP10 or MKP11, but they were too big to fit properly). Caps links: 1, 2, 3, 4, 5, 6.

8247467.jpg

WIMA MKS input caps replaced with JB JFX Premium Filter MKP (however, I had them replaced all 4 of them later 4.7uF later)

Mod no. 2: I had installed G.O.S.S. (Grain Oriented Silicon Steel) to shield the toroidal transformer.
Initially, with my sensitive AKG K550 and without any source connected to XLR inputs, when using +20dB gain setting I was able to hear a bit of a hum noise when volume pot was at 3 o'clock or more, especially on the right channel. Now after partially shielding the transformer that noise is COMPLETELY gone!

8247470.jpg

G.O.S.S. shield installed

Mod no. 3: I got installed as well an EMI/RFI noise suppression filter, but this mod added a bit of hum noise again, because the AC lines from the EMI/RFI filter were too close to XLR inputs (also, PCB traces from XLR were right besides the noise suppression filter). So, I moved the filter from right to the left side, just above the 120/230V selector, as the red arrow shows (sorry for not providing a picture of the actual noise suppression filter positioning, but I was quite struggling with fitting the filter inside the case and I totally forgot to get another picture). Also, please ignore the big red wire, it was connected to the 2 Y-capacitors, but it's not really needed, so I removed it from lack of space (it also did a quite nice short-circuit with flame, so next time I'll be more careful).

8247484.jpg

G.O.S.S. sheet and the EMI/RFI filter

8247486.jpg

EMI/RFI filter on top of right of the PCB, just on top of the XLR input traces (had it moved on the left side later to remove the hum induced to the input traces)

Mod no. 1. is recommended and this will improve the sub-bass a bit. This is easy to be seen with square-waves on the scope more than listen to the ears.
Mod no. 2 is recommended only if with sensitive cans you can hear mains hum, otherwise there's risk of short-circuit, so needs care.
Mod. no. 3 is not recommended; instead an external EMI/RFI filter could be added later...perhaps one included in the power strip or something similar. There's too much risk with this filter to get installed inside the metallic case.
 
Last edited:
OP
trl

trl

Major Contributor
King of Mods
Joined
Feb 28, 2018
Messages
1,967
Likes
2,523
Location
Iasi, RO
Mod no. 4: I've replaced resistors R10, R1, R13, R14 of 10 KOhms with 270 Ohms resistors for further decrease amplifier's noise. These resistors are placed directly in signal path in series with the 50 kOhms potentiometer and for an inside temperature of 50C their noise would be 1.88881 uV RMS, that would be -114.5 dBV, same as stated in Matrix's datasheet (calculated with http://www.daycounter.com/Calculators/Thermal-Noise-Calculator.phtml). However, when volume is to the maximum, potentiometer's thermal noise covers the 10 kOhms resistors noise anyway.

The 270 ohms resistor are having only 0.310363 uV RMS of noise, that means -130 dBV. Difference is audible indeed, as I only modified one channel first and in my headphones one channel was having less noise (checked with sensitive cans like Solo2 and Superlux 16-ohms IEM).

Low_pass_filter_caps.jpg

Original SMD 10 KOhms resistors R10, R1, R13, R14 - directly in signal path

WP_20180321_21_19_50_Pro.jpg

Not-so-good-looking 270 Ohms resistors R10, R1, R13, R14 - directly in signal path (please use SMD metallic resistors here; 1 KOhm should do as well)

Mod no. 5: Also, in the picture from above, notice the 4 missing caps C29, C30, C31, C32 from the low-pass filter that f...ed up the trebles roll-off.

Screen Shot 2018-03-10 at 17.38.28 copy.png

Before changing the low-pass filter created R10, R1, R13, R14 and C29, C30, C31, C32

Screen Shot 2018-03-10 at 17.59.34 copy.png

After changing the low-pass filter created R10, R1, R13, R14 and C29, C30, C31, C32


Screen Shot 2018-03-10 at 17.32.42 copy.png

Before changing the low-pass filter created R10, R1, R13, R14 and C29, C30, C31, C32


Screen Shot 2018-03-10 at 18.00.27 copy.png

After changing the low-pass filter created R10, R1, R13, R14 and C29, C30, C31, C32

Notice: You might want to replace the original value for the capacitors C29, C30, C31, C32 with another value and not to actually remove them for good like I did, as this might interact with poor interconnect cables by adding possible noise or hum (most likely above the audio band)!

Note: Blue is the input audio signal from the generator, Red is the output audio signal from the 6.3mm jack. Only one channel was tested, the other one should be identical. Perfection means Blue superimpose with Red and become one sine.


Mod no. 1-bis: Upgrading the 4 input caps with bigger 4.7uF/250V.
WP_20180321_21_31_24_Pro.jpgWP_20180321_23_08_30_Pro.jpg
These are connected in signal path (DC blocking caps), so bass extensions will be better now.

WP_20180321_23_06_23_Pro.jpg

The bigger JB JSX 4.7uF/250V fits hardly into the case

Notice the blue resistor from right of the above picture; I've changed the 50 KOhms resistors from the inputs with 150 KOhms resistors; this increases the input impedance, but helps a bit with bass roll-off. This is not recommended, but worth a try.

Square_20Hz_before_High_Pass_mod copy.png

Original HPA-3B with 2.2uF input caps


Square_20Hz2.2uF150kOhm copy.png

Modified high-pass filter (changed input resistances from 50 KOhms to 150 KOhms) mod with 2.2uF caps


Screen_Shot_2018-03-22_at_21.07.15 copy.png

Modified high-pass filter mod with input 4.7uF caps


Screen Shot 2018-03-10 at 17.38.42 copy.png

Original HPA-3B with 2.2uF input caps (pretty good 20 Hz response)


Screen_Shot_2018-03-22_at_21.23.13 copy.png

Modified high-pass filter mod with input 4.7uF caps



Square_20Hz_after_cap copy.png

Audio signal applied after the input caps (or short-circuit input caps), so indeed: "No cap is the best cap":)


Note: Blue is the input audio signal from the generator, Red is the output audio signal from the 6.3mm jack. Only one channel was tested, the other one should be identical. Perfection means Blue superimpose with Red and become one sine.
 
Last edited:
OP
trl

trl

Major Contributor
King of Mods
Joined
Feb 28, 2018
Messages
1,967
Likes
2,523
Location
Iasi, RO
Thank you!

Most of the above mods are not quite necessary, so if your HPA-3B sounds perfectly, there is no need for any of these mods.
 
OP
trl

trl

Major Contributor
King of Mods
Joined
Feb 28, 2018
Messages
1,967
Likes
2,523
Location
Iasi, RO
Mod no. 6: Lowering HPA-3B's internal voltages from +/-18V to +/-15V to accommodate more op-amp swapping.

8477875.jpg8477879.jpg

I got rid of existing JRC7818/7918 by removing the 12 screws from both lateral black aluminium heat-sinks from power transistors, then with a small clipper/cutter I cut, one by one, all the legs for each power regulator. The new NJM7815/7915 were carefully soldered (legs were also carefully cut to accommodate with heat-sink holes, of course). I choose NJM manufacturer because they say noise from 10-100KHz would be 90uV/170uV RMS for 7815/7915 regulators, quite a decent value I'd say; bought them from PROFUSION UK, authorised JRC/NJM/MUSES dealer. However, SPARKOS regulators might be better here, feel free to give'em a try.

After the above mode I've installed 2 x SS V5-dual and first thing I noticed was a lower noise than the original LME49860, which is quite an improvement I'd say, because with max. gain of +20dB and sensitive 32 ohms headphones we all know that there is some noise out there, especially when tested with volume to the max (RCA shorted, no music).
smily_headphones1.gif


Noise estimation was done by changing the opamp from one channel only, then comparing left channel with right channel with sensitive AKG K550 headphones and my own ears. Right channel had installed the default LME49860 op-amp and left channel had installed the below op-amps, in order of background noise estimated by my own ears:
- BURSON V5-dual: about 30-40% lower noise than original LME49860 (a bit hot to touch, some pops&clicks when switching gain, might need a more airy case perhaps, DC-output >10mV/channel when high-gain)
- MUSES8920: about 25% lower noise than LME49860 (I like the very low DC-output, only few mV)
- LM4562: about the same noise as MUSES8920
- AD8599: about 3 times more noise of LME49860 (have no idea why, I thought this op-amp is low-noise, based on it's datasheet).
- OPA1652: about 40-50% lower noise than original LME49860 (no pops & clicks when switching gain on power button ON/OFF)
Headphones used for these tests: AKG K701 (used for scene presentation and layering), BEYERS DT880/600 (used for details, saxophone and women voices) and AKG K550 (used to get background noise and overall musicality).

To accommodate V5 op-amps, besides replacing both power regulators, both DIP8 sockets need to get removed and to solder V5 opamps directly onto the PCB, because HPA-3B case can't be perfectly closed otherwise; BURSON V5 is about 2mm taller than HPA-3B can support! However, this will not apply to V6, because its outer case can be easily removed and the 2mm difference is gained this way.

WP_20171029_16_17_09_Pro.jpg WP_20171029_18_10_25_Pro.jpg

Power drops after the above mod, but not a lot:
MATRIX_HPA-3B_15V_60Ohms_power.png vs. MATRIX_HPA-3B_18V_60Ohms_power.jpg
Single ended (60 Ohms): 1.1W vs. 1.5W (8.042V RMS vs. 9.402V RMS)

MATRIX_HPA-3B_15V_60Ohms_balanced.png vs. MATRIX_HPA-3B_18V_60Ohms_balanced.jpg
Balanced (60 Ohms): 3.5W vs. 4.5W (14.55V RMS vs. 16.38V RMS)

For 32 Ohms output power is about 30% higher than the above 60 Ohms test, so quite a beefy headamp.
 
Last edited:
OP
trl

trl

Major Contributor
King of Mods
Joined
Feb 28, 2018
Messages
1,967
Likes
2,523
Location
Iasi, RO
Mod. no. 7: Cutting PCB traces and joining the ALPS potentiometer with the XLR plugs via shielded wires. This mod cleans up a little bit the background noise when highest gain setting (+20dB) is used and volume pot is above 3 o'clock (especially the right channel).

Input_shielded_wires.jpg

Some manufacturers are doing some tricks to minimise the interferences from potentiometer wiring:
- Sennheiser is using a metallic rod to minimise the lengths of the wires: HDV 820
- Zen is using the same metallic rod style: http://electronics-diy.com/schematics/1114/zen-headphone-amplifier.jpg
- Dr. Meier Corda is using PCB shield (ground) to provide EMI/RFI protection: https://www.innerfidelity.com/images/MeierAudio_CordaROCK_Photo_InsideTop.jpg
- Objective2, one of the least noisy headamp ever created, has external PSU and analogue input nearby the potentiometer: http://lh4.ggpht.com/-AuQyUd8hJFA/TlK7jR4sXqI/AAAAAAAABOI/0P6dlJhJXZ4/s1600-h/o2%2520v11%2520board%255B4%255D.jpg

Perhaps the best way to place an analogue potentiometer would be nearby the input plugs (XLR/RCA), so metallic rod is probably the best approach, but a good shielding will do as well. However, Matrix was using inside HPA-3B simple PCB traces without any shield around, nor on top/bottom and mains hum could be heard with IEM's (at least in my setup).
 
Last edited:
OP
trl

trl

Major Contributor
King of Mods
Joined
Feb 28, 2018
Messages
1,967
Likes
2,523
Location
Iasi, RO
Mod no. 8: Replaced SMD transistors with ThroughTheHole BC550/BC560 to further decrease the background noise of the amplifier. Also, it will look like the Audeze Deckard, with TO-92 transistors instead of SMD.

HPA-3B_SMD_transistors.jpg

Original SMD transistors, 16 of them (4x4 with heatsinks above)

HPA-3B_SMD_transistors_2.jpg

image006.jpg


image001.jpg image003.jpg
 
OP
trl

trl

Major Contributor
King of Mods
Joined
Feb 28, 2018
Messages
1,967
Likes
2,523
Location
Iasi, RO
Mod no. 9: Adding few more decoupling caps on the backside of the PCB, nearby the two dual op-amps (4 x WIMA MKP 0.1uF/100V in parallel with V+/GND & V-/GND for both OPAMPs, 4 x WIMA MKP 0.1uF/100V in parallel with the 4 x 330uF Nichicon capacitor from the middle of the board and 2 x AVX 10uF/35V tantalums in parallel with the 2 x 330uF capacitors from output of power regulators).

8126783.jpg


This not really necessary, but most op-amp's datasheets are recommending 0.1uF decoupling caps, along with 1...10uF, depending on the application (in HPA-3B's case there are 1uF caps already installed).
 
OP
trl

trl

Major Contributor
King of Mods
Joined
Feb 28, 2018
Messages
1,967
Likes
2,523
Location
Iasi, RO
I was able to measure the above modded HPA-3B @5V RMS (dummy resistor at around 600 Ohms).
2nd harmonic is somewhere around -94dB, pretty good actually. The 3rd harmonic is at around -98dB.

HPA-3B_1KHz_spectrum.png




Dynamic is 108dB and the gain was set to HIGH. However, with my 16 Ohms/103dB sensitivity IEM's connected, on the HIGH gain the background noise becomes audible. Lowering the gain to MID or to LOW makes noise to disappear completely, so I assume dynamic might increase a bit on MID or LOW gains (will probably check this sometimes too).

RMAA_5V_RMS_HPA-3B.png
 
Last edited:

maxxevv

Major Contributor
Joined
Apr 12, 2018
Messages
1,872
Likes
1,964
That's a really solid performance if the measurements are correct !

I wonder how does the unmodified version measure though ?
 
  • Like
Reactions: trl
OP
trl

trl

Major Contributor
King of Mods
Joined
Feb 28, 2018
Messages
1,967
Likes
2,523
Location
Iasi, RO
I wonder too...I actually made a post yesterday on Head-Fi, just in case someone who owns this amp can do an ARTA and/or RMAA test.
 
OP
trl

trl

Major Contributor
King of Mods
Joined
Feb 28, 2018
Messages
1,967
Likes
2,523
Location
Iasi, RO
On max. volume and max. gain (in single ended operation) the background noise of HPA-3B when using IEMs is lower than CV2's headamp; tested with A/B, although it's not needed the A/B test, because difference is big enough (CV2 has also a bit higher gain, so it's probably not only amp's internal noise).
 

maxxevv

Major Contributor
Joined
Apr 12, 2018
Messages
1,872
Likes
1,964
I wonder too...I actually made a post yesterday on Head-Fi, just in case someone who owns this amp can do an ARTA and/or RMAA test.

Originally, I had volunteered my Massdrop set to be dropshipped to Amir. But Massdrop shipped it out 2 weeks earlier than expected, as such I couldn't change my shipping address in time... :rolleyes:

That said, the original set is already awesomely good . On balanced, on lowest gain, I can drive both the HD6XX and HE4XX to levels beyond what I would comfortably listen to before I even turn the knob 50%.

The best part is that its so linear there is practically no distortion to be discerned. Such that, if I turned the knob blindly , the volume keeps going and its still sounding so super smooth. You barely notice the volume increase until you take it off the head or look at the knob ! :D

Its a seriously under-rated, too under-the-radar headphone amp in my opinion.

Perhaps you can take some measurements of the balanced output at lowest gain and 50~75% knob volume setting. I suspect the performance would be really, really good from what I hear with mine.
 
OP
trl

trl

Major Contributor
King of Mods
Joined
Feb 28, 2018
Messages
1,967
Likes
2,523
Location
Iasi, RO
I think I can measure the balanced outputs, of course...I can do that, just need to look for an adapter from XLR to 3.5mm jack. I'm not sure it should measure any better than single-ended, after all there are same 4 identical amplifiers inside and while playing music in single-ended only two of them are actually operational and referenced to the ground.

When getting the HPA-3B into balanced-operation, then each amplifier will amplify V+ and V- for both stereo channels, without referencing the ground. For power hungry cans balanced will be better; Meier Corda explains this very well here: http://www.meier-audio.homepage.t-online.de/ (active GND driving should measure about the same as balanced driving).

My HPA-3B sounded very powerful and great from the beginning. What I didn't liked was it's higher heat dissipation, a little bit of mains-hum (only with sensitive cans) and maybe some background noise with IEMs, but I'm usually too picky on the background noise. Also, trebles were a bit soft, but seems that correcting the phase-shifts did resolved that (the high-pass filter actually).

Also, in my square-waves tests that got the HPA-3B until clipping, the resistors from around the power transistors got brownies and a bit smoked (smell appeared too), so I needed to change them for more powerful ones. This didn't happened with my Burson CV2 headamp where the resistors were warm-to-hot, but no brownies in there. I think Audeze Deckard (about the same schematic and components values, but only 2 amplifiers inside) is using more powerful resistors, at least by looking at Internet pics, but I'm not 100% sure.
 

actionmask

Member
Joined
Jan 31, 2019
Messages
19
Likes
28
Location
San Jose, CA
Nice work!
May I ask a question
Is the 270 ohm resistor in the signal path necessary?
Since there is 50 kOhms potentiometer.
Thanks.
 
OP
trl

trl

Major Contributor
King of Mods
Joined
Feb 28, 2018
Messages
1,967
Likes
2,523
Location
Iasi, RO
Thank you!

This resistor is in signal path, indeed, and in my case it was lowering channel imbalance a bit (definitely these resistors were not installed for this matter). Of course, its role is most likely as source resistance or to work as low-pass filter (in conjunction with capacitors).

However, I've seen that LME49720+LME49600 headamp has no input resistors, so I'm sure in most designs these input resistors could have a value of zero.

I personally lowered to about zero these input resistors just to lower the background noise, and to my ears and basic measurements too the noise decreased (with pot knob higher than 12 o'clock).
 

actionmask

Member
Joined
Jan 31, 2019
Messages
19
Likes
28
Location
San Jose, CA
Thanks. The famous JDS Atom amp has a similar implementation. JDS use JRC2068 instead of LME49720. As you can see they use a 270ohm SMD resistor (R7 and R20) for 2068's input.
BTY, have you bought Atom? If you do, you may try to replace the 2068 op-with better op-amps.
I'm not good with SMD components.
Maybe it will give an even better performance.
JDS Labs Atom Headphone Amplifier Teardown PCB.jpg
 
OP
trl

trl

Major Contributor
King of Mods
Joined
Feb 28, 2018
Messages
1,967
Likes
2,523
Location
Iasi, RO
I don't have ATOM yet, thanks.

Swapping 2068 opamp might worth the trouble only if the internal gain is higher than 5X, which is probably not the case with ATOM.
 
OP
trl

trl

Major Contributor
King of Mods
Joined
Feb 28, 2018
Messages
1,967
Likes
2,523
Location
Iasi, RO

JohnYang1997

Master Contributor
Technical Expert
Audio Company
Joined
Dec 28, 2018
Messages
7,175
Likes
18,292
Location
China
Thanks. The famous JDS Atom amp has a similar implementation. JDS use JRC2068 instead of LME49720. As you can see they use a 270ohm SMD resistor (R7 and R20) for 2068's input.
BTY, have you bought Atom? If you do, you may try to replace the 2068 op-with better op-amps.
I'm not good with SMD components.
Maybe it will give an even better performance.
View attachment 28334
I did. No improvements can be made. Single lme49600 reached it limit.
I did opa2156 opa1612 opa2211 opa2189 etc and changed feedback net work for higher open loop gain. None. No improvements.
All tests are done in unity gain.
 
Top Bottom