Upgrade Pro-Ject Head Box II headphone amp

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Dear all,

I just bought a second hand of these.
I found some good ideas here to improve the sound a bit.
Do you also notice a humming noise (50hz probably) when you touch the metal housing? The hum disappears when you als touch the ground of the in or output socket together with the housing, the body seems to act as an antenna I suppose..

Should the metal housing be connected to ground then?
Does anyone has the schematics?
And will it help to rectify with four diodes?

Thanks for sharing ideas
 
I greet the members of this beautiful forum.
After many years I returned to play with the iron.
Me too trying to improve the sound of this amplifier. It's not easy.
There are some places where it is possible to made changes: the choice of the Op Amp, the choice of the final transistors, the value of the feedback, the value of the cap in parallel with Feedback Resistor, the value of the emitter and output resistors (too high), the quality of the capacitors, and so on.
Some specimens have only one rectifier diode, others have the classic Graez bridge. I don't think it matters.
In few weeks I will probably post my mods here.
I cannot help about the humming noise: I’m working without the case, and I don’t have problems. But I remember humming with wrong network values through the feedback resistor.
I apologize for my inaccurate English.
Another greeting to everyone from Italy.
Here is the diagram (except possible errors).
 

Attachments

  • Pro-ject Head Box II.gif
    Pro-ject Head Box II.gif
    28.6 KB · Views: 371
Last edited:
The concept generally looks sensible enough (clearly designed by a veteran or at least someone familiar with tube circuits), bar a few things I'd take a closer look at:

1. The capacitance multiplier for bias / virtual ground supply looks like it may have too little voltage headroom for ripple IMO. Not sure why they chose a 100 ohm base bias resistor - there would be precious little voltage drop across that, and it makes the 47µF less effective than it could be (why use a cap multiplier when you might just as well go poassive instead?). I would propose the following mod:
a) replace 100 by 1k and add 12k in parallel to 47µF (you may be able to go even higher, like 2k2 + 27k or 4k7 + 56k/62k)
b) swap 10k and 9.1k R3 (I chose the resistor ratio in (a) specifically so you can do that, hopefully I didn't screw up the maff)
c) add reverse protection diode across NPN E-C (cathode to collector) - 1N4148 may do, 1N4001 would be better.

Honestly I'm not convinced that this setup (as shipped) even is more effective than some passive filtering. You could rip out the NPN, replace its B-E junction with 820 ohms and still add the protection diode (c), and I bet you'd be no worse off than stock.

Well what do you know, LTspice agrees with me. In fact, ripple rejection at 100 Hz goes up from 37 dB to 64 dB. How silly is that. (And no voltage headroom issue either!) In fact, I have to go with 2k2 / 27k in (a) before performance of the modified stock circuit becomes equally good, and with 10k / 120k for a marked advantage. (Attached simulation with 47k / 560k delivers 88 dB.) Let that sink in, you could have made this circuit simpler, more robust and a smidge cheaper with better performance, or squeezed out even better performance for the grand total of one more 1/4 W resistor and one more diode.

Hence why all the world uses passive R/RC filtering for such virtual ground bias and gets away with it.

2. Of the 3 output stage bias diodes, move two to touch BD139/140 transistor cases (e.g. thermal glue, epoxy). That should improve thermal stability, especially when the whole affair is pushed into AB operation (which it will handle but not particularly gracefully). Responsiveness may be improved further if one diode leg is moved directly to its corresponding transistor base leg (thermal conduction through metal is much faster than through plastic). With the bias diodes running warmer, the 22k resistor may have to be adjusted up a bit.
 

Attachments

  • headbox2-bias-1.png
    headbox2-bias-1.png
    31.5 KB · Views: 258
  • headbox2-bias-1.asc
    2.3 KB · Views: 49
  • headbox2-bias-3-mod.asc
    2.6 KB · Views: 34
Last edited:
Thank you sgrossklass for your interest.
I have never taken care of the power supply for the virtual earth, and perhaps I have even misrepresented the function of the transistor: I thought, with the 47uf capacitor, they wanted to delay the entry into operation of the circuit at power on. In fact, if in its place we put a 22uf cap, the starting time is halved..
Incidentally, I checked the resistor, and its value is actually 100 ohms.
I have no power rejection problems (I am convinced that the problems of willemhoekstra depend on mass loops - see the attached photo), but I am interested in your modp, in order to make the circuit more robust (in fact I have already fried two transistors in that position).
In the next few days I will try the change and publish the result here.
I also haven't understood another thing: why is the divider of the virtual earth asymmetrical? Why not two 10k resistors? And why is it necessary to swap resistors in your ModA?
Other than that, currently still have a slight 'tremble' of female voices at high volume: it seems that a 100 or 200 Hz oscillation is triggered. In male voices this effect seems less audible, perhaps masked by the different timbre.
Regarding the bias of the output stage, I can assure that it is absolutely stable, even at 35ma and with the circuit outside the case. I was the first to amaze myself with this, but it is so.
It should be added I replaced the 22 resistors with 1 ohm resistors (following NavGuy's advice), and the impedance of the headphones I use is only 28 ohm; also i like to listen rock at very high volume.
Therefore enough current flows: yet the bias is granite.
Again: once inside the case, the diodes would 'feel' the temperature even better, even if not glued on the outputs, with an effect similar to that adopted in Naim amplifiers.
Regarding the ground connection of the case, I found this photo on the net.
 

Attachments

  • headbox2.jpg
    headbox2.jpg
    164.7 KB · Views: 228
Last edited:
I have no power rejection problems (I am convinced that the problems of willemhoekstra depend on mass loops - see the attached photo), but I am interested in your modp, in order to make the circuit more robust (in fact I have already fried two transistors in that position)
Ouch. The protection diode may have helped avoid this, at least it's a proven approach for regulator pass transistors when sizeable capacitors are present on the output side (which is the case here). Otherwise said output capacitor may keep its voltage high for much longer than input voltage stays up, subjecting the poor transistor to substantial C-E reverse voltages up to the point of breakdown.

The bias circuitry does double as a soft power on, and that's not a bad thing in a single-supply design. Power-down is where things potentially get tricky, as the opamp input may easily exceed its common-mode input range, resulting in some nastiness. Again, a protection diode may save the day.
I also haven't understood another thing: why is the divider of the virtual earth asymmetrical? Why not two 10k resistors? And why is it necessary to swap resistors in your ModA?
In a single supply design, the supply voltage likes to sag under load, so that optimum virtual ground bias tends to be slightly below midpoint. We can assume that the designer started with 10k/10k and then tweaked values for maximum output in a given load impedance range.

The swapping resistors bit was just supposed to be for convenience, so the number of new parts is kept to a minimum. I figured that by swapping them around, I would gain some voltage headroom which could then add to the capacitance multiplier's, while leaving amplifier gain the same. That bit of interaction between bias resistor network and gain is somewhat annoying actually, as I would also like to have at least somewhat lower resistor values in the feedback network (not that much, maybe a factor of 2), but I suppose the provided input bootstrapping was worth it. Maybe.
Other than that, currently still have a slight 'tremble' of female voices at high volume: it seems that a 100 or 200 Hz oscillation is triggered. In male voices this effect seems less audible, perhaps masked by the different timbre.
I would want to hunt that one down a bit more methodically - doing so by ear tends to be tedious and unreliable. This approach would require the following steps:
1. Make a clean e.g. 400 + 600 Hz test tone (2x sine generator in Audacity would be fine), which could then be stored on a portable MP3 player. (All kinds of tests can be performed like this, including standard SMPTE IMD 60 Hz / lower-level 7 kHz and CCIF IMD 19+20 kHz, the same test tones you also be stepped in e.g. 10 dB volume steps.)
2. Use a couple resistors to make a ca. 30 ohm x2 headphone dummy load to mimic yours. It should dissipate at least 1 W per channel comfortably.
3. Unless you have a recording device available that'll take approx. +15 dBu of input, a passive L attenuator in parallel to the dummy load should be used to drop levels into safe terrain. For onboard audio you'd want approx. -12 dB, so maybe 3k3 + 1k2 resistors (1/4 W metal film)?
Regarding the bias of the output stage, I can assure that it is absolutely stable, even at 35ma and with the circuit outside the case. I was the first to amaze myself with this, but it is so.
It should be added I replaced the 22 resistors with 1 ohm resistors (following NavGuy's advice), and the impedance of the headphones I use is only 28 ohm; also i like to listen rock at very high volume.
Therefore enough current flows: yet the bias is granite.
The wonders of Class A operation, I guess. How did you manage to get down to 1 ohm emitter resistors - I suppose you either removed one bias diode or replaced it by a Schottky?
Regarding the ground connection of the case, I found this photo on the net.
So basically case front connects to grounding point on PCB, then the case screws together and is all grounded that way. I can only assume that willemhoekstra's unit might have been tampered with before or there is some sort of bad solder joint, loose connection or oxidation going on. With the picture for reference, it should be easy to find the issue.
 
Last edited:
Hi fellows,
Thanks for all the improvements that are possible.
About the humming noise, the photo with the red wire was very helpful. The unit I got had not any effectieve connection of the PCB ground the chassis. Just a wire from PCB ground to the chassis I tried but was not very helpful. Inspired by the photo I grounded the housing of the potmeter too. The Axis and housing were floating wrt ground and the case although it is fixed using a nut to it.
After soldering a connection of PCB ground to the chassis AND to the potmeter housing the humming disappeared. Thanks very much !!!! I use this amp to be able to listen using headphones on a SONOS.
 
Last edited:
Willemhoekstra, I am happy that the photo I found on the internet was useful to you: at least we have achieved a concrete result. It is true that it would be better to dedicate the summer (always more beautiful as the years go by) to other things, but I am in a fairly advanced experimentation phase, and I need little time to publish the updates here. Do not forget to check this discussion, and then, next winter, we will be able to compare our changes. I promise some news, fireworks, hopefully with little magic smoke ...
I must point out (also to Stephan sgrossklass) that the photo of the red wire does not belong to the stock, but it is already a modification by someone who had the same problem.

Quote sgrossklass: Ouch. The protection diode may have helped avoid this

Yes, after you wrote this here, I realized it could, like as it is recommended with Lm317.


Quote sgrossklass: ]the supply voltage likes to sag under load .... and then tweaked values for maximum output in a given load impedance range.

Very clear explanation, thanks. I would never have gotten there alone. However this design could be inherited from an unstabilized supply. Here it may not be necessary, as we have a robust 7812.
(Incidentally, I have provided the stabilizer with a heatsink which does not seem to be necessary (as in stock) if the alternate power supply voltage does not exceed 17v. It did not seem optimal to place the TO220 on the board, as it seems to have been originally designed, however this arrangement has not been made in the final version.)

Quote sgrossklass: I would want to hunt that one down a bit more methodically

It should be noted that I have no notions of electronics comparable to yours. This is perhaps why I prefer the use of the ear as a tool (very powerful, under certain conditions).
It seems to me, if I have not misunderstood, that the concept is this: to solicit the amplifier in the context of suspect frequencies, creating the conditions for unpleasant intermodulations, recording them in order to be able to analyze them. But isn't a good oscilloscope necessary for this? I do not have it.
The hypothesis that I made is however another: the problem would not happen in the context of the audible frequency, and intermittent signal saturations would cause an effect in the context of the audible.
It is a sneaky and subtle effect. It seems (but I'm not sure yet) that it occurs only in the presence of compressed recordings, with artificially declipped signals. However, it is more or less pronounced depending on how the compensation network is sized above the feedback resistor. I don't think I can accuse the recording of having a defect: in some cases I mean Chesky Records!!
Who knows if I can publish some few seconds samples here without violating copyright, or forum rules?

Quote sgrossklass: How did you manage to get down to 1 ohm emitter resistors - I suppose you either removed one bias diode or replaced it by a Schottky?

At first it seemed I had problems. I tried with green, white, red leds. But in the end it was enough to play with the resistor. It works perfectly, like I said.

But before further reflection, it is better that you publish the schematic I have adopted. It contains (with probable errors) a definitely unorthodox choice of components, and a 'trick' to tune the half-wave phase of the signal I have never seen, discussed or published (but maybe I haven't searched enough).

Let's see if you are able understand what it is, regardless of whether it really works…

See you in a few days…
 
Here is the schematic of the amplifier how it is working right now.
The choice of the final oversized tranasistors is not part of the 'recipe': they are the ones I managed to combine according to my idea (it is possible that it is not new, however).
In any case, in my opinion, the BD139-140 of current manufacture are not suitable for authoritatively driving headphones, and they give a flat sound.
Someone said a long time ago, in the seventies: "NPN, PNP transistors are as truly matched as men and women of the same height and weight".
So I matched them following another theory.
 

Attachments

  • Pro-ject Head Box II MOD.gif
    Pro-ject Head Box II MOD.gif
    31.1 KB · Views: 264
I followed the suggestion of sgrossklass:
Honestly I'm not convinced that this setup (as shipped) even is more effective than some passive filtering. You could rip out the NPN, replace its B-E junction with 820 ohms and still add the protection diode (c), and I bet you'd be no worse off than stock.
I can't say for sure that it's better, but it's certainly not worse. It is certainly more robust.
The amplifier thus described sounds quite good. We will see if we can understand the nature of the problem that occurs in some passages of the female voices.
It must be remembered that there is a review on Head-Fi entitled "Pro-ject Headbox II: First impressions" which, unfortunately, is perfectly in agreement with my own impressions.
The final transistors are matched of HFE and VBE, but also of a third characteristic, which is my 'trick' (or my illusion).
I want to quote John Linsley Hood's 'Audio Electronics' book:
“In many ways the Shaw/Baxandall quasi-complementary output transistor circuit is preferable to the use of fully complementary output transistors, since, to quote J. Vereker of Naim Audio, ‘NPN and PNP power transistors are only really as equivalent as a man and a woman of the same weight and height’ – the problem being that the different distribution of the N- and P- doped layers leads to significant differences in the HF performance of the devices. Thus, although the circuit may have a good symmetry at low frequencies, it becomes progressively less symmetrical with increasing operating frequency.
With modern transistor types, having a higher current gain transition frequency (that frequency at which the current gain decreases to unity), the HF symmetry of fully complementary output stages is improved, but it is still less good than desired, so the relative frequency/phase characteristics of each half of the driver circuit may need to be adjusted to obtain optimum performance.”
 

Attachments

  • Pro-ject Head Box II MOD - resistor 820.gif
    Pro-ject Head Box II MOD - resistor 820.gif
    33.1 KB · Views: 226
Pro-ject Head Box II - las mods (maybe)

Hi folks!
Perhaps the time has come to close the case: the final changes should be that of the attached diagram.
There are two important changes compared to the latest version.
I have definitively adopted the OPA1652 (18 Mhz) instead of the OPA1642 (11 Mhz).
In this configuration, for me, it is by far the best compared to many tested.
Here is the list:
JRC2068D (stock), JRC4562D, JRC5532D (good, better than NE5532), LM4562 (National), NE5562 (TI), OPA1612 (here it oscillates heavily), OPA1692, OPA1656, OPA1642, OPA2209.
The replacement of OPA1942 only occurred when I replaced the ouputs with the following 'married couple' (remember Vereker's phrase): MJE15028 (30 Mhz) + 2SB1185 (Rohm, 70 Mhz).
I paired them based on HFE (measured at 6ma, having only a Mega328), and the VBE, unfortunately, is adapted with a 5% deviation. This is probably the limit to be respected.
I think I understand (this is the trick I was talking about) that the most important thing is to synchronize the phase of the two half-waves of the signal, as happens in the famous Naim amplifiers with the adoption of two compensation networks. Here another method is used to achieve the same result.
Now the amplifier sounds 'good', balanced throughout the sound range, with deep basses. I can listen it for hours without fatigue.
Thanks again to Sgrossklass for the suggested modification (a resistor instead of a transistor), which proved to be perfectly effective.
Regards
 

Attachments

  • Pro-ject Head Box II MOD - DEF.gif
    Pro-ject Head Box II MOD - DEF.gif
    35.9 KB · Views: 216
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.