JLH Headphone Amp

Hi Mayday,

Im listening to my JLH Class A amp as we speak, using a Sansa Clip+ and Seinnheiser HD565 Ovation headphones, sounds great with the mods discussed earlier in this thread. Maybe you are making this too complicated with output caps etc. Usually less is better in audio :). You should be able to adjust the output offset to less than 5mV and lose the huge caps. If nothing else, try to find out why there is a large offset, that is more likely a fault elsewhere. I will help too if you need it.

Hi,
Thanks for ordering your help :)
The only things I've done apart from swapping output resistors to 68R, is using film caps everywhere for everything under 1uF.
Changed the included 100uF caps for 180uF Elna RJH's (except the ones orienten incorrectly on the silkscreen where I used 470uF Rubycon ZL's).

U also use a 10K Dact type attentuator instead of the included vol pot.

Oh, swapped the 3V3 Zener's included to 2V7 Zener's as shown on the silkscreen.

The DC offset is identical between the channels with the trimpots turned fully CCW.

Should anything be very wrong I assume I'd get different offset between the channels? Or some other clue that something was off?

After swapping the 220K resistors connecting to the middle of the trimpots to 2x 100K in series to use the midpoint of these resistors as connection from DC-servo output when I incorporate in to the circuit.

After this operation I now get around 170mVdc offset.

That is when the circuit is hot.
Heatsinks can be touched and I can keep my finger on them without getting burned.

Maybe I should swap the series resistors to 2x 110K as the single 220K seemed to give lower and faster dropping DC offset.

From trimming offset before swapping the 220K resistors, I noticed turning the trimpots CCW lowered positive offset, and CW lowered negative offset.

Now getting 165mVdc on both channels after having the headamp powered on for around 30-45min, I haven't tried turning the trimpots from their fully CCW position.
 
Hi Mayday,

I would make sure your DC-servo is out of circuit and concentrate on getting your output DC offset as low as possible first. Take off any DC load/caps at output and turn your offset 10K pots mid-way. Check the DC offset at output, whatever value it is, check that turning the offset pot SLOWLY CCW or CW will reduce the offset to around 0mV. It may not be stable, but you should be able to alter it with the DC offset circuit as designed. I use 4 x 1N4148 diodes, 180uF filter cap, 10K resistor to -12V, 68K resistor to diode chain 10K adjustment pot - I can adjust the output above and below 0V and it is very stable once the output transistors are warm. You must have a 0.6 - 0.7V drop across R1/R2 = 5R1 to produce approx. 100mA bias current through the output pair
 
Hi Mayday,

I would make sure your DC-servo is out of circuit and concentrate on getting your output DC offset as low as possible first. Take off any DC load/caps at output and turn your offset 10K pots mid-way. Check the DC offset at output, whatever value it is, check that turning the offset pot SLOWLY CCW or CW will reduce the offset to around 0mV. It may not be stable, but you should be able to alter it with the DC offset circuit as designed. I use 4 x 1N4148 diodes, 180uF filter cap, 10K resistor to -12V, 68K resistor to diode chain 10K adjustment pot - I can adjust the output above and below 0V and it is very stable once the output transistors are warm. You must have a 0.6 - 0.7V drop across R1/R2 = 5R1 to produce approx. 100mA bias current through the output pair

Hi,
The DC-servo is not connected in any way, I've only soldered in the wires for in/out and power supply (not connected at the DC-servo end).

I had 4x 1N4148 in series (reversed orientation compared to zener) but the LED didn't light up.

I got to thinking that since going from 220K to 200K(2*100K) on the resistor going to midpoint of the trimpot made DC offset rise...maybe replacing them with 240K (2*120K) would lower DC offset?

Writing from my phone, so I don't have the schematic in front of me.
 
No, the 220K resistor only provides isolation between the 10K DC output adjustment pot and the diode bias network. Its value is not critical but lower values allow more neg 12V DC rail noise to be injected into the input Q5/6 transistor, as well as forming a low pass filter with C13/14. Lower values will also give a "stiffer" bias point setting. JHL amp builders have used values between 47K - 220K with no problems. The LED not lit means you don't have a -12V supply, unless LED wrong polarity. Try reducing the value until you can adjust output DC offset around 0mV (i.e. + and - readings)
 
I am tweaking my amp now, as it happens, to try to improve the bass response for my HD565 headphones. I increased R19/20 to 100K (from 68K), C13/14 to 1uF (from 0.68uF) and C23/24 to 560uF (from 220uF). The new values are fairly typical values chosen by many JLH amp builders so there is nothing very critical in the design of the input stage/dc offset adjustment. It sounds very good, bass is definitely a bit better but through headphones its never going to be ground-shaking :D. Photos attached.
 

Attachments

  • RIMG0564.JPG
    RIMG0564.JPG
    659.2 KB · Views: 614
  • RIMG0565.JPG
    RIMG0565.JPG
    646.8 KB · Views: 593
  • RIMG0566.JPG
    RIMG0566.JPG
    656.1 KB · Views: 601
Last edited:
No, the 220K resistor only provides isolation between the 10K DC output adjustment pot and the diode bias network. Its value is not critical but lower values allow more neg 12V DC rail noise to be injected into the input Q5/6 transistor, as well as forming a low pass filter with C13/14. Lower values will also give a "stiffer" bias point setting. JHL amp builders have used values between 47K - 220K with no problems. The LED not lit means you don't have a -12V supply, unless LED wrong polarity. Try reducing the value until you can adjust output DC offset around 0mV (i.e. + and - readings)
The LED lights up fine now, it didn' with 4x1N4148 in series instead of the zeners

I could adjust around 0mV offset with 220K, with 2x100K in series for 200K offset with trimpot fully CCW eventually got down to around +155mVdc.
This with 1100uF "bipolar" caps on the output and 110R dummy loads.

I am tweaking my amp now, as it happens, to try to improve the bass response for my HD565 headphones. I increased R19/20 to 100K (from 68K), C13/14 to 1uF (from 0.68uF) and C23/24 to 560uF (from 220uF). The new values are fairly typical values chosen by many JLH amp builders so there is nothing very critical in the design of the input stage/dc offset adjustment. It sounds very good, bass is definitely a bit better but through headphones its never going to be ground-shaking :D. Photos attached.

I've used the included 470nF + 1uF wima parallell on the bottom of the PCB from the get go for C13/C14.
C23/C24, are those the ones with the wrong orientation on the silkscreen? If so, I've had 470uF Rubycon ZL there since I populated the PCB.

I have the inputs open if that makes any difference?
I know some designs requires them loaded or shorted to gnd for proper dc offset adjustment/stability.

I'll go over the board again today and see if there's any solderjoints that has been effected by me soldering wires to component legs etc.

I will also see if I can find 2pcs 110K axial resistors (I've got litteraly thousands of 0805 resistors as I mainly use those when building on perf board, but regular through hole parts, not as many).
 
C23/24 need to be installed with opposite polarity shown on silkscreen. Inputs and outputs should be unloaded. The value of R19/20 is not critical but I would stick to 100K or 110K or less, just to get it working properly.

The diode or original zener implementation only produce a fixed negative VREF voltage (about 2.5 - 3V) for the dc offset trimmer pot. Diodes are better because they allow some degree of temperature tracking of the output dc offset within the amplifier case. AFAIK nobody has calculated it empirically but 4 x 1N4148 diodes seem to work fine. Each diode has a tempco of around -2mV/C and this seems to compensate for the combined thermal drift of the transistors. But don't quote me on it;) The only down-side is the thermal stabilisation takes a few minutes while the output transistors warm up but I don't think that is a big deal.

If you cannot adjust the output dc offset to approx. 0mV (+5 to -5 mV), something is wrong and there is no point going any further. Adjusting the dc trimmer causes Q5/6 to conduct more or less, adjusting the DC balance point between Q3/9 and Q4/10 i.e. the output. Easy things to check, 0.5 - 0.6V drops across R1/2 and R25/26. Check NPN and PNP are in correct place, even with pre-assembled boards. When you get a nice stable 0mV on your output, you wont need big output caps or DC-servo circuits to pollute the sound:)
 
Last edited:
Re-heated just about all solder joints to be sure.
Set the trimpots about 10turns CW (using 23 turn pots).
Startup 15:14 this afternoon
An externally hosted image should be here but it was not working when we last tested it.


15:55 this afternoon
An externally hosted image should be here but it was not working when we last tested it.


16:10 this afternoon
An externally hosted image should be here but it was not working when we last tested it.


Offset is virtually identical between L & R channels.

I've not tried trimming DC offset down to zero as I wanted to see where it landed with the pots set about in the middle of their range.

The DC-servo board is not connected in any way yet, just mounted in the chassis.
 
Those limit values seem OK though I had something similar with full CCW or CW. Make sure you actually have 2.5 - 3.0V on your dc trimmer pot. Now SLOWLY turn DC trimmer CCW or CW half-turn at a time to reduce offset. You should see it drop from these readings. It might "spring" back up a bit but that's ok. It should never return to these initial values. Take care when (and if!) you get down to around 0mV, turn pot VERY SLOWLY otherwise you will over-shoot into the negative offset range.
 
Last edited:
Mayday, just noticed your marked-up schematic. Have you made any of the changes recommended in the previous posts on this thread? If not, probably explains why you are having difficulty setting up and why we are not seeing the same things.

I have 68R output resistor instead of 2R2, different caps, but that's about it really. Apart from using a very nice 10K stepped attentuator.

I don't even remember all the modifications tried in this thread as it's been a while since I read the entire thing.
 
Forgetting the cap changes for now, the general consensus was:

R9/10 change from 2K2 to 3K3
R21/22 change from 1K to 10K (important because its in dc offset circuit)
R19/20 change to less than 220K, around 100K
Replace the 2V7 zener with 3 or 4 1N4148 diodes in series

I've located the resistors you're mentioning in the post.
Should be easy enough to swap those.

R19/R20 however I use two resistors in series as I'll connect the DC-servo out to the midpoint.
Strangely when dropping the value for them from 220K to 200K I got more pos DC offset. Maybe the other changes will compensate for that?

As for getting to around 100K for R19/R20, I have 47K and 68K that I know of from memory(every value in between as well could I use 0805's), so my choice there is 94K or 136K.

Zeners vs 1N4148, I guess as I'm using 2V7 zener as marked on the silkscreen, I'd use 3x1N4148? IIRC they are supposed to be oriented in the opposite direction compared to the zeners?

About thermal drift with diodes/zeners, wouldn't mounting them on the bottom of the PCB be beneficial?
 
We are just trying to "standardise" your build to be similar to the 100's of other JLH amps built by happy people on this thread. I would use a single 68K resistor for R19/20 for now. Mount the 4 x 1N4148 diodes in same place as original zener stacked on top of board as you did before, but observe reverse polarity to silkscreen. The thermal tracking is only an approximation, as long as they are sensing the internal case temperature. With those changes, you should get to the 0mV (approximate) starting point. Then you might want to listen to it and make any further "sonic" changes before you start on your other changes (output caps or DC-servo etc)
 
We are just trying to "standardise" your build to be similar to the 100's of other JLH amps built by happy people on this thread. I would use a single 68K resistor for R19/20 for now. Mount the 4 x 1N4148 diodes in same place as original zener stacked on top of board as you did before, but observe reverse polarity to silkscreen. The thermal tracking is only an approximation, as long as they are sensing the internal case temperature. With those changes, you should get to the 0mV (approximate) starting point. Then you might want to listen to it and make any further "sonic" changes before you start on your other changes (output caps or DC-servo etc)

I'll get on it tomorrow, right now I think my fiance feels a bit negleckted.

I've made notes of what to do, that's always helpful..having a list to "tick off" as you go :)
 
We are just trying to "standardise" your build to be similar to the 100's of other JLH amps built by happy people on this thread. I would use a single 68K resistor for R19/20 for now. Mount the 4 x 1N4148 diodes in same place as original zener stacked on top of board as you did before, but observe reverse polarity to silkscreen. The thermal tracking is only an approximation, as long as they are sensing the internal case temperature. With those changes, you should get to the 0mV (approximate) starting point. Then you might want to listen to it and make any further "sonic" changes before you start on your other changes (output caps or DC-servo etc)
I made the suggested changes, though stuck with 2*47K resistors R19/20, and I found the DC offset much easier to trim down.
Half a turn on the trimpots made alot more difference than before.

After about 30-40min I started trimming DC offset, I was in a bit of a hurry, but got the DC offset to around +/-4mVdc for both channels.

It seems the DC offset started dropping faster as well and didn't drift as much.

I swapped 2K2 to 3K3, 1K to 10K, 2V7 to 4x1N4148 and 2*100K (220K spöt) to 2x47K.


The heatsinks didn't get as hot as before these changes I think.
 
Well done Mayday, great news:). Now you are at a good starting point to continue. The lower value R19/20 makes it easier to stabilise, i.e. it forms a stiffer bias for Q5/6. It may take you a few adjustments to get the dc offset in desired range but it should be quite stable. The long-term stability depends on the local environment/size of your amplifier case but since it is a headphone amp, that should be similar for most amplifiers. Have you listened to the amp? The only other change is to replace C23/24 with something in range 470 - 680uF to reduce the low frequency bandwidth and improve bass response. When you are happy with the sound, you can implement your dc servo as others have done to reduce the dc offset completely. Good luck with that and please post your circuit in case others want to implement this improvement in their own build.
 
I made the suggested changes, though stuck with 2*47K resistors R19/20, and I found the DC offset much easier to trim down.
Half a turn on the trimpots made alot more difference than before.

After about 30-40min I started trimming DC offset, I was in a bit of a hurry, but got the DC offset to around +/-4mVdc for both channels.

It seems the DC offset started dropping faster as well and didn't drift as much.

I swapped 2K2 to 3K3, 1K to 10K, 2V7 to 4x1N4148 and 2*100K (220K spöt) to 2x47K.


The heatsinks didn't get as hot as before these changes I think.

I would seem my phones swedish autocorrect was active in this post as well, I usually catch it but I was in a bit of a hurry :eek:
 
Well done Mayday, great news:). Now you are at a good starting point to continue. The lower value R19/20 makes it easier to stabilise, i.e. it forms a stiffer bias for Q5/6. It may take you a few adjustments to get the dc offset in desired range but it should be quite stable. The long-term stability depends on the local environment/size of your amplifier case but since it is a headphone amp, that should be similar for most amplifiers. Have you listened to the amp? The only other change is to replace C23/24 with something in range 470 - 680uF to reduce the low frequency bandwidth and improve bass response. When you are happy with the sound, you can implement your dc servo as others have done to reduce the dc offset completely. Good luck with that and please post your circuit in case others want to implement this improvement in their own build.

Thank you for all your help, it is grately appreciated :)

C23/24 is that the ones with the silkscreen error?
If so, I've had Rubycon ZL 470uF there from the start.
All other 100uF spots are populated with 180uF Elna RJH's.
No, I have not yet had the courage to hook up my Phidelio X2's to this headamp...

My Changes from the original kit so far:
All caps exept the the larger ones replaced with Wima film caps.
470nF input caps parallelled with 1uF Wima (on the bottom of the PCB)

2V7 (or as my kit came with, 3V3) zeners replaced with 4x1N4148 in series. OPPOSITE direction from how the zeners are marked on the silkscreen.

R19/20 changed from 220K to 2x47K in series (I will use the midpoint of these resistors to "inject" signal from the DC-servo).

R21/22 changed 1K to 10K resistors.

R9/10 changed 2K2 to 3K3

Probably not necessary but I matched all of the changed resistors.

Where I had already used the zero ohm resistors, I parallelled with tinned Cu wire on the bottom of the PCB, where I had not yet populated them I just used tinned Cu wire.

The so called zero ohm resistors measures from 0,5 to over 1ohm so they are not the greatest quality.

Not yet done:
Remove onboard voltage regs and use LT1083's (requires true dual secondary transformer ie 0-15VAC 0-15VAC, CAN NOT be used with center tapped transformer).

Hook up the DC-servo. All wiring done already.
A big thanks to Mooly for helping with the DC-servo amongst many other things!

Install UPC1237 based HP protection circuit?
I have a kit on the way and from my understanding, the cut off for the relay can be altered to the desired offset voltage by changing resistors(I have not yet read up on this).
 
Always happy to help. I would recommend you actually listen to the JLH headphone amp, after all, that's what its for;). Check your dc offsets are below 10mV and plug in your cans. This will give you a sonic starting point before you make any further changes that might actually degrade the sound. In fact, I would listen to it for several days before making any other changes. If you plan to use LT1083 low drop-out regulators with 20VDC supplies shown on your schematic, I recommend you fit 4R7 - 10R 3W resistors in the supply lines before the on-board psu filter caps. I fitted 3R3 resistors conveniently into supply lines using the vacant bridge diode positions, feed from an external separate 18V dual supply. The resistors reduce the heating of your regulators but more importantly provide welcome additional mains ripple rejection on the supply.