O2 headamp output booster PCB

Thanks Ut looks like I may be able to salvage it. It's so strange how the amp was working perfectly fine and all of a sudden developed a large dc offset. It was only a week or two ago that I posted the pics of my meters reading zero when I measured the DC offset and I haven't changed anything since!
 
Hi agdr

Just want to add my 5 cents into the distortion discussion.
Opamp, especially JFET input ones, will have some additional distortion while working in non-inverting configuration (as a result have significant common mode signal) from a relatively high impedance source and input impedances are not matched.
I guess this is exact situation you have. One input is connected to high impedance potentiometer, another to something like 1k feedback resistor.
You can do a few things:
Use lower pot value as much as you can (5k?)
Match feedback resistance to signal source resistance. May not be really an option since the pot impedance changes.
Go to an inverting configuration. Can be an interesting option because it will allow to use linear potentiometer. They usually have better channel matching and a little bit cheaper.
 
Hey Sergey888,

Thanks for the suggestions!

The circuit below is what the booster board is plugging into in the existing O2 headphone amplifier. The fellow that designed it has an NJM2068 based gain stage driving a 10K audio pot, with a 2.2uF film cap on the pot wiper to keep DC out of the wiper and any DC on the amp input from shifting the amp output level. Then a 40.2K to ground for the input bias current return of the NJM4556A chips it used to power.

One of the design goals with the booster board has been to change as little as possible on the O2 board it is plugging into, to make things plug-and-play, and to allow folks to switch back easily. So that would probably eliminate changing the pot, although I agree with you that a lower value, and a linear taper, is a good idea.

In fact that is exactly how things wound up on my O2 desktop amp project (ODA). I used a 1K pot for lower Johnson noise, along with a LME49990 to drive it which can drive 600 ohms. Then I recently switched from audio to linear because the linear had twice the power dissipation rating. I didn't know about the better channel-to-channel tracking though with the linear pot! Interesting. That makes sense. It would probably be much harder for the manufacturer to get two log resistance traces to match.

But for the O2 amp we are stuck with the 10K audio pot. Also from a driving standpoint. The datasheet THD+N graph for the NJM2068 is done with a 2K load. The O2 designer is already using 1.5K for the NJM2068 feedback resistor and the pot would wind up in parallel with it. Then the coupling caps - when I went down to a 1K pot in the ODA I had to go up to four 4.7uF coupling caps to keep the low end of the frequency response the same.

Good thoughts about raising the booster board feedback resistor. I've been thinking about that too. Maybe 5K (4.99K) for mid-range on that O2 amp 10K pot would be better. Or 3.3K for 1/3 of the way up. I would guess that most folks will run their pot 1/3 - 2/3 of the rotation before flipping the gain switch to the next level.

I'll have to ponder the inverting configuration a bit. To keep from forming a voltage divider with that 40.2K resistor he has to ground the input resistor for an inverting stage would probably have to be 10x, or somewhere around 390K. Then same for the inverting feedback resistors to keep the overall stage gain -1. Those higher resistor value may introduce a lot of additional Johnson noise.

Study the chip datasheets a bit more this weekend the OPA827 may not be that much of a THD+N gain over the OPA140 after all. Mainly just a small decrease in voltage noise. The bipolar LME49990 should be though, at the expense of a larger DC output offset voltage.
 
agdr

There is a trick with loading a pot with a resistor. If you use a linear pot and load it with a resistor with value several times lower than the pot, you get pretty reasonable approximation of an exponential characteristic. But it still involves changing a pot.

If you are going to change a resistor in the feedback circuit of the OPA140 + LME, don't forget to scale the other components values as well. Otherwise you will significantly change frequency compensation.

Also mid range of a 10k pot will be 2.5k, not 5k
 
Last edited:
I've decided to start over with a new O2 board. Luckily I bought spares of all the parts when I built the first one so all I need is the pcb and a few jacks and switches. I figured it would probably be easier to build a new one then to try and troubleshoot the mystery voltage that suddenly appeared on pin3/u3.
 
I've decided to start over with a new O2 board. Luckily I bought spares of all the parts when I built the first one so all I need is the pcb and a few jacks and switches. I figured it would probably be easier to build a new one then to try and troubleshoot the mystery voltage that suddenly appeared on pin3/u3.

Good that you bought some spares! One of these days as time permits you might want to mess with the old one some more and see if you can track that voltage down.
 
Good that you bought some spares! One of these days as time permits you might want to mess with the old one some more and see if you can track that voltage down.

I was able to install a new socket for u3 and the O2 is back up and running. I connected the booster board and no more turn on/off thump and the dc offset is down to acceptable levels. It is unmeasurable on the left channel and .8mv on the right channel which leads me to believe there is still something not right with that channel. I'm gonna try giving the board a good cleaning with a toothbrush and some alcohol and see if it helps. Either way I plan to to build a new O2 since I already ordered a new pcb.
 
Hi agdr
Just want to add my 5 cents into the distortion discussion.
Opamp, especially JFET input ones, will have some additional distortion while working in non-inverting configuration (as a result have significant common mode signal) from a relatively high impedance source and input impedances are not matched.

This is somewhat true but there are special cases where it is not.

It is due to input capacitances of the JFET input pairs and these are usually highly unlinear which basically means that in non inverting configurations the distortion rises when the source impedance increases.

The solution is the use a JFET Opamp with input cascodes.

Take for example the OPA627, it has input cascodes which nullifies the rising distortion with increasing source impedance in non inverting configurations.

Another option is the newer OPA827, it does not have input cascodes, but instead it is using TI's SiGe BiCOM3 process which, among other things, minimizes the input capacitance to neglible levels.

So yes, Opamps with JFET inputs do in general have rising distortion with increasing soure impedance in non inverting configurations, but there are exceptions.

The OPA627 is in my opinion too expensive for what you get, but the OPA827 is a lot more reasonably priced and with better overall specifications than the OPA627.

An unity gain buffer with BUF634 or LME49600 enclosed in the feedback loop of an OPA827 is a great low noise, low distortion solution. Add an input gain stage with a low noise, low distortion opamp like OPA1611 or LME49990 and you can safely use a 10K pot efore the unity gain buffer without adding noise or distortion. To make it all DC coupled, use a DC Servo with an OPA140 JFET Opamp and add it to the unity gain buffer.

You end up with a low distortion, low noise, low DC offset, high quality DC coupled headphone amplifier that will beat the crap out of many a "highend" "Audiophile" consumer headphone amplifier.
 
Last edited:
Neutrality.....

Are there any examples of this type of design in a DIY format or in a commercial head amp?

thanks
Alex
Adydula,

To my knowledge there is no such design available, either DIY or commercial, if we are talking a 100% copy of what I have described. But there probably are variations of it. There is more than one way to skin a cat :)

I am working on a variation with Balanced inputs and single ended but dual mono outputs.

To give you an idea :

Balanced in>SE>Volume>Unity gain output buffer(with DC servo).

Components I am working with :

OPA1612>OPA1611>Relay attenuator>OPA8278+2xBUF634 with OPA140 DC servo.

All supplied by floating LT3080 regulators.

Very low noise, very low distortion and very low DC offet should be possible to attain with this design. Resistor values optimized for low noise and high linearity, 4-Layer PCB etc. Majority of components being SMD.

But this is still a work in progress.

It will get its own DiyA thread when the time comes.
 
Another option is the newer OPA827, it does not have input cascodes, but instead it is using TI's SiGe BiCOM3 process which, among other things, minimizes the input capacitance to neglible levels

Very interesting! Thanks for the input. I've been wondering what some of the internal differences are between the older OPA627 and the OPA827. I was kind of wondering if there was a fabrication process difference involved. Well this is really good news! There may be some hope after all for better THD+N measurements with OPA827's in the O2 booster board rather than the OPA140s.

Using an OPA140 servo loop around the (bipolar) LME49990 + LME49600 combination might be interesting. The arrangement may result in the lower THD+N of the bipolar chip with some of the same low DC offset of the OPA140 + LME49990 loop. As far as NwAvGuy's O2 headamp the booster board plugs into, it wouldn't care one way or the other, FET or bipolar input, since he has a coupling cap on the pot wiper to keep DC current out. I just went with FET to keep the current through his 40.2K ground return resistor low, and hence keep the IR drop on that resistor low, so it wouldn't reflect back to the output as DC offset. But with the servo it wouldn't matter.
 
Is a dc servo really even necessary with these lme49600+opamp headphone amps? The Wire for example is dc coupled and uses the lme49600 in the feedback loop of the lme49990 and dc offset is only .3-.4mV.

In most cases, no.

But imagine you have a DC coupled amplifier and attach a source to it that has a lot of DC offset, that DC offset is then amplified by your gainstage.

Say you have a gain of X4 and you attach a source with say, 5mV DC offset, you now have 20mV of DC offset on the output of your amplifier, not good.

I personally prefer to minimize DC offset as much as possible.
 
Last edited:
In most cases, no.

But imagine you have a DC coupled amplifier and attach a source to it that has a lot of DC offset, that DC offset is then amplified by your gainstage.

Say you have a gain of X4 and you attach a source with say, 5mV DC offset, you now have 20mV of DC offset on the output of your amplifier, not good.

I personally prefer to minimize DC offset as much as possible.

I never thought of it that way thanks for the explanation:)
 
I got my new O2 board in the mail today so I should have it built in the next few days and hopefully I won't have any mystery voltages on this one:D. AGDR I remember in your O2 modifications thread you had mentioned something about the bass being better with the O2 on AC vs Battery power because of the impedance of the regulators being lower then the batteries. Would using a polymer capacitor for C8 and C9 help at all? I was thinking something along the lines of these 16SEPC270MX+S Panasonic | Mouser
 
I got my new O2 board in the mail today so I should have it built in the next few days and hopefully I won't have any mystery voltages on this one:D. AGDR I remember in your O2 modifications thread you had mentioned something about the bass being better with the O2 on AC vs Battery power because of the impedance of the regulators being lower then the batteries. Would using a polymer capacitor for C8 and C9 help at all?

You are right, I thought that at one time based on the math. But it turned out that I was wrong about that one.

I did some actual testing after that and found the battery impedance didn't make any difference at all. Kind of surprised me. :) I ran several low frequency sine waves into the O2 and measured the output voltage on AC and on batteries (the output voltage was setto be low enough to work on either batteries or AC). The net result: no difference at all in output voltage either way. Not even a tiny difference, I measured pretty carefully. What that must be saying is that the NiMH battery impedance is much lower than I thought it was from the graphs I found via Googling that day.

So it is OK to build your O2 up the standard way.
 
I would not recommend using the LME49990 if you have a pot in front of it. High frequency distortion increases rapidly with increasing source resistance.

I would not use it with source resistance much above 500R if you want to keep the high frequency distortion in check. Otherwise you are wasting the potential of the LME49990.

LME49990 used for its low distortion properties and in an non inverting configuration absolutely needs low source impedance or it is a waste of time using it.
 
agdr

There are a few things you should not forget about bjt input opams.
1. There are a lot of them with input bias current compensation, so you may get an appropriate bias voltage with given DC impedances.
2. They have higher current noise, especially those with lower voltage noise, as their input stage runs on a higher current. Using them even with 10k pot may end up with current noise dominating. LME49990 is a good example.

Have a look at OPA(2)209. It has nice combination of current and voltage noises and low bias current. On the down side GBW is not so great, but not too far for OPA140 ect. It is quite linear, especially on a lite load.
Also large phase margin can be quite forgiving.