JLH Headphone Amp

Attempt at my own PCB

Hi,
I would like to make my own etched PCB.
here is the power section

please give your suggestions , as i am totally new to this stuff.
 

Attachments

  • Schematic 18-03-2021.png
    Schematic 18-03-2021.png
    30.5 KB · Views: 577
  • PCB 18-03-2021.png
    PCB 18-03-2021.png
    43.4 KB · Views: 582
Thanks for reply,
as this will be diy type so will be using salvaged heatsink.
Can someone please check the schematic
is the schematic correct ?

also i have rearranged the pcb layout so can use old huge heatsink if needed
the previous one was attempt at layout used in mini psu by ESP.
 

Attachments

  • PCB 19-03-2021.png
    PCB 19-03-2021.png
    43.3 KB · Views: 522
I'm putting together one of these kits. The grounding and power distribution scheme on these things is petty bad.

Is there need/benefit to matching transistors, as there is with the JLH69 amp?

Most useful transistor substitutions? The 2N2907A's supplied with my kit are high-Hfe but poor linearity. Seems like a bad choice to drive the outputs.
 
Since no input was received, I tested and selected the transistors prior to assembly. BTW, the board in question is the "GZLOZONE ZERO ZONE V:2015 JLH1969H.P." board that has pretty much taken over the offerings on eBay and Aliexpress.

Before getting into the results, the grounding scheme needs looking at. There have been reports of a "ground loop" on the board that causes hum in some cases. There is, in fact, no ground loop, but just a very poor grounding scheme that can result in hum. (I think the reason these kits sell for so cheap is that the power and ground layout sucks and limits its performance, not something you want in a power amp...).

Worth mentioning, the fill on the component side of the board looks like a "ground plane", but it is not connected to ground anywhere. It's just a floating layer of copper.

Now see the attached pic of the solder side. The rectangle in the upper center is the "ground plane", just a pad, really. The xfmr center tap connects to it at the top via a short jumper. All of the post-rectifier smoothers and post-regulator filters are arrayed around the sides. At the bottom, three fairly thin traces connect separately to signal ground, the volume pot and filters, and bias networks. And then one thin trace fans out to the output and shunt feedback resistor on both channels. This is not a good idea. And it's about 100mOhms end-to-end - not a quality ground. That's the trace that is the problem.

Any impedance imbalance in the AC transformer and bank of capacitors generates a standing current on that ground pad, and develops a voltage across the impedance of the trace going to the feedback shunt R's. It's one of the worst grounding schemes I've encountered. There is no way it can "sound good" like this.

The fixes are marked in red.
: Cut the thin trace that connects input to output ground points on both sides.
: Cut all but a 1/4" of the ground "pad" right after the rectifier smoother caps, forcing the currents to mix before passing on to the regulators' end of the ground pad.
: Install separate ground jumpers on both channels from the output to the regulator side of the ground pad.

This puts line freq harmonics just a few dB above the noise floor, and makes the output dead-quiet.

It also helps to add jumpers from the ground terminal of each regulator to the regulator ground point on the pad, to lower their impedance. But that's not necessary for fixing the hum.

I'll go into the rest of the unit next post.
 

Attachments

  • GndMods.jpg
    GndMods.jpg
    258.6 KB · Views: 230
Except for the 2N2907A's, the supplied transistors were quite good. The TIP41C's all measured 119-120 Hfe at 150mA. The 2N2222's were around 170 Hfe and had linear gain curves. The 2907A's were high Hfe but very nonlinear gain. I went through some 50-60 2907A's I have on had (mostly pulls) and found four that had matched good gain and linearity.

The supplied 2200uF/25V lytics are just fine for smoothers, with ESR around 30mOhms. All other caps were tossed and replaced with Rubycon ZLH and UCC KYZ.

The supplied regulators are junk. And they're not fakes; the ST + and - 12V regulators are among the worst I've seen from an output impedance POV, even under load. I replaced them with a well-behaved set from TI and Fairchild.

Most other changes have been well-covered in this thread: lowered gain to 4, eliminated the 1k/330pF RC in parallel with the feedback R; changed input C to 1uF; changed volume pot to 10k, the series R following it to 2k, and the shunt cap following it to 100pF, resulting in ~340kHz BW with the pot at normally-used positions; changed zener resistor to 3.3k to lower its noise. I added wire jumpers in parallel with existing traces from each regulator's output and ground pins to the filter caps and their section of the ground pad. The narrow pcb traces were adding 50-70mOhms of resistance to each of them; not good for a regulator. The result is a flat 44mOhms of impedance at the ground pad.

After correcting an assembly mistake (swapped resistor positions), I looked at waveforms, using load R's equivalent to the headphones I use. Square waves looked excellent with no issues on the edges into any of the loads. It takes around 20 minutes for the DC offset to stabilize, after which it is pretty stable.

So I did some listening before further testing. For reference, I used a TPA6120A-based amp I've tweaked out. The latter sounds excellent for higher-impedance 'phones (150 Ohms and up) but distortion rises sharply with low-Z phones. I was hoping the JLH would fill that need. Headphones used were Senn HD580 (300 Ohms), AKG 240DF (600 Ohms), Sen HD560 (300 Ohms), and AKG K245 (45 Ohms). (Impedances are minimums, as measured).

Impressions are very similar to the JLH amps I made for my electrostatic headphones. Smooth, slightly thick bass; detailed highs; surprisingly poor transparency; 2-dimensional. Like a soup/stew with a bit too much corn starch added. In short, the 6120 amp cleaned its clock, even into 44 Ohms. The fundamental richness and body of instruments and voices, and the rendering of space were so much better with the 6120. After having put several days into building and tweaking this thing, this came as a big disappointment.

So I did 1kHz distortion measurements with 10k, 600, 300, 150, and 44 Ohm loads (see attached). Levels were adjusted to reflect the difference in sensitivity of the headphones.

The noise floor is clean in all cases. The H2 difference between 10k and 600 Ohms @ 2.8VRMS was only a couple dB of H2, roughly -82dB. At 1.8V into 300 Ohms, it actually improved to -84. At 1.8V into 150 Ohms, it started to increase, to -80. And at 1V into 44 Ohms, it was only -70dB.

So, this headphone amp does not drive low-Z phones well. And it is not best-in-class by any stetch.

I've learned a lot in my two excursions into JLH land. I've come to agree with sgrossklass's assessment; that they were very good in their day, but have been outclassed since then.
 

Attachments

  • JLHP44.jpg
    JLHP44.jpg
    36.9 KB · Views: 90
  • JLHP15.jpg
    JLHP15.jpg
    37.2 KB · Views: 89
  • JLHP3.jpg
    JLHP3.jpg
    37 KB · Views: 96
  • JLHP6.jpg
    JLHP6.jpg
    36.7 KB · Views: 166
Last edited:
Since no input was received, I tested and selected the transistors prior to assembly. BTW, the board in question is the "GZLOZONE ZERO ZONE V:2015 JLH1969H.P." board that has pretty much taken over the offerings on eBay and Aliexpress.

Before getting into the results, the grounding scheme needs looking at. There have been reports of a "ground loop" on the board that causes hum in some cases. There is, in fact, no ground loop, but just a very poor grounding scheme that can result in hum. (I think the reason these kits sell for so cheap is that the power and ground layout sucks and limits its performance, not something you want in a power amp...).

Worth mentioning, the fill on the component side of the board looks like a "ground plane", but it is not connected to ground anywhere. It's just a floating layer of copper.

Now see the attached pic of the solder side. The rectangle in the upper center is the "ground plane", just a pad, really. The xfmr center tap connects to it at the top via a short jumper. All of the post-rectifier smoothers and post-regulator filters are arrayed around the sides. At the bottom, three fairly thin traces connect separately to signal ground, the volume pot and filters, and bias networks. And then one thin trace fans out to the output and shunt feedback resistor on both channels. This is not a good idea. And it's about 100mOhms end-to-end - not a quality ground. That's the trace that is the problem.

Any impedance imbalance in the AC transformer and bank of capacitors generates a standing current on that ground pad, and develops a voltage across the impedance of the trace going to the feedback shunt R's. It's one of the worst grounding schemes I've encountered. There is no way it can "sound good" like this.

The fixes are marked in red.
: Cut the thin trace that connects input to output ground points on both sides.
: Cut all but a 1/4" of the ground "pad" right after the rectifier smoother caps, forcing the currents to mix before passing on to the regulators' end of the ground pad.
: Install separate ground jumpers on both channels from the output to the regulator side of the ground pad.

This puts line freq harmonics just a few dB above the noise floor, and makes the output dead-quiet.

It also helps to add jumpers from the ground terminal of each regulator to the regulator ground point on the pad, to lower their impedance. But that's not necessary for fixing the hum.

I'll go into the rest of the unit next post.


What an expert post - I wish I knew this when I tried to get rid of hum issues on my build for months on end (changing components, resoldering, metal shielding etc, etc). Succeeded to get it almost silent in the end, but not to this extent.




An externally hosted image should be here but it was not working when we last tested it.
 
Last edited:
I wish I knew this when I tried to get rid of hum issues on my build for months on end (changing components, resoldering, metal shielding etc, etc). Succeeded to get it almost silent in the end, but not to this extent.

Sorry to hear that so much time was lost to this. It seems to be the nature of the beast, at least in this form.

It would certainly be interesting if someone did a layout that optimizes the power and grounding schemes, i.e. like one would do with an actual power amp...
 
Sorry to hear that so much time was lost to this. It seems to be the nature of the beast, at least in this form.

It would certainly be interesting if someone did a layout that optimizes the power and grounding schemes, i.e. like one would do with an actual power amp...

Can you share the schematic (if any) and BOM which came with the kit. I may attempt at making PCB with your suggestions.
 
That would be great! The schematic was posted in #811:
JLH Headphone Amp

though a few parts values are different (the power supply 'lytics). A pic of the component side is in the prior post.

I should say: I don't think my fix is "optimum" from a design POV. It just fixes the hum problem on this board. I would rather see "signal ground" treated like a separate low-impedance trace that follows the signal flow from input to output, with nothing PS-related on it, with signal gnd tied to pwr gnd at the regulator side of ground only. Maybe use the ground plane for power gnd. This is a pretty simple circuit, so that should not be too difficult.

Also, IMO, the regulators should not be spread apart like they are here. They should be right next to each other.
 
In another attempt to salvage this board, I lowered the gain to 3.1x and used 470uF in the shunt feedback leg to move its cutoff freq way below 1 Hz. That resulted in a definite improvement. I then replaced the 2N2222A driver/phase splitter with a ZTX450. It has a bit lower Hfe (175 vs 270 for the 2222A) but even better linearity. The result was not good, a brighter thinner sound. I was going to try a BD139 but the ones I have here are even lower Hfe (130-ish). I recall DIYer grunf did lots of experiments with that transistor position in the 10W amp and found that high Hfe and good linearity was required for best results.

Running out of options...
 
Last edited:
Well wha'dya know - this thing is actually starting to sound pretty darn good!

First, I added more trace wires to lower the impedance of important traces and make them more equal (there were some large-ish differences between the two channels). While doing this, I realized it's probably best to just treat this board as a component holder and basically add P2P wiring to it.

Second, when swapping a ZTX450 in for the supplied KSC2222A made things sound worse, it made me think. The 450's current gain is a bit more linear than the 2222's, but it's Hfe was lower by 1/3. So Hfe has to be at least as high the originals. I went NPN hunting in the spares boxes. A lot of old junk in there.

I found a pair of 2SD1020-F's (pulls, from what I have no idea), found the data sheet online (atached), and it looked pretty interesting. "Designed for use in the output stage of tape recorders"... Hfe of the F grade is between 200 and 320... 170MHz GBW... sounds like they could be good drivers. So I measured them, and lucky me, their Hfe was 265 and near-perfectly matched. Put them on the trusty ol' curve tracer (a Polar T1200) and was greeted with flat, equally-spaced lines with no loop on the return trace. Better than the 2222A's.

So I put them in, warmed the amp up, and holy beans I'd say 75% of my previous complaints about the amp were solved. Still not up to the level of the 6120 amp, but it sounds really quite good, and the colorations now being "musical" with most recordings. Load drive into 44 Ohms was not improved, unfortunately.

So now I'm suspicious of the 2N2907A inputs and want to replace them with something more linear. Any suggestions?
 

Attachments

  • 2sd1020.pdf
    132.8 KB · Views: 116
Last edited: