Differential and DC offset

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
In some writings, it is said that one of the advantage of having differential for input stage is that it can hold DC offset. It makes me think. Designs like JLH or NAD3020 power amp doesnt use differential at front end. They just use single transistor.
How do these amps hold their DC, without servo? (and if we dont use cap in the output). Will their DC wonders around when the amp gets hot?

Differential also cancels 2nd harmonic. What is the goal of the designer not to use differential. Is it to get overwheling 2nd order harmonic?
 
Differential also cancels 2nd harmonic. What is the goal of the designer not to use differential. Is it to get overwheling 2nd order harmonic?

I suppose it is possible to make a differential deliberately distort but this is not the conventional reason for it. Rather, a differential pair, used correctly, provides one of the most accurate ways to subtract two voltages.
 
Disabled Account
Joined 2003
I have to say that my jlh1969 has been very stable, DC-wise. After it thermal-stablizes (in 30 seconds or so), the DC on the output cap doesn't wonder more than 60mv - on a pair of mosfets. the circuit seems to be more stable with BJT output devices.

The single transistor input is fast (faster than ltp). But it does lose some DC stability, tho I doubt it matters that much.
 
This is an interesting example of differential and DC offset. In this X-amp schematic, R19 and R29 forms a voltage divider more than to hold offset. I think the offset is set by the low-value R1/4 and R44/45 at the output.
In this particular schematic, can the amplifier maintain DC offset without R19 and R29?
 

Attachments

  • axschematics11.gif
    axschematics11.gif
    45.3 KB · Views: 848
lumanauw said:

It is interesting. How can I use differential front end, but have big 2nd harmonic? In what part of differential should be modified?

So why those designers use single transistor for front end?

By adding a simple single-ended open-loop pre-amp to any power amplifier you can increase 2nd harmonic distortion.

The use of single transistor input stages is mostly an outdated practice and it's rare to see anything like this in a modern (post 80's design). The ltp beats it in every way without adding complexity.
 
Disabled Account
Joined 2003
thanh said:

Why? Can ltp faster than single transistor?

the set-up of those single transistor input stages is a "current-feedback" amp - voltage feedback is applied to the emitter of the input transistor (the NAD and JLH are such examples). such a set-up tends to have faster bandwidth: I have simulated quite a few of those JLH variants and they go to 20-25khz open loop, vs. a few khz for the LTP using the same transistors.

You will also notice that new high-speed opamps tend to be current-feedback amps (those they do use LTP in a different way than the LTP we are talking about here).

Another thing about those single transistor input stage is that they are so stable there is no miller cap on the VAS: you can still put one in but normally they are not required. for LTPs, if you don't put a miller cap in, they usually oscilate.

the downside, if that can be called one, is that for stability reasons, you cannot use large feedback resistors in a current feedback amp. But that reduces resistor noise as well.
 
Re: Re: Differential and DC offset

millwood said:
The single transistor input is fast (faster than ltp)...........

Indeed there is no basis for this assertion in fact....

Richard C said:


The use of single transistor input stages is mostly an outdated practice and it's rare to see anything like this in a modern (post 80's design). The ltp beats it in every way without adding complexity.

True :nod:
 
millwood said:


the set-up of those single transistor input stages is a "current-feedback" amp - voltage feedback is applied to the emitter of the input transistor (the NAD and JLH are such examples). such a set-up tends to have faster bandwidth: I have simulated quite a few of those JLH variants and they go to 20-25khz open loop, vs. a few khz for the LTP using the same transistors.

This is all rather misleading......a two-stage design with a single common-emitter stage for input rather than a diff. pair will not necessarily have 'faster' bandwidth......or wider foward path bandwidth than the later case.....

Details such as type of compensation used and the value of first stage transconductance used need to be known if such a comparison is to be meaningfull.

millwood said:



Another thing about those single transistor input stage is that they are so stable there is no miller cap on the VAS: you can still put one in but normally they are not required. for LTPs, if you don't put a miller cap in, they usually oscilate.


This is also not true...Indeed, the single CE input stage with a resistive load delivers TWICE the transconductance of a similarly loaded diff. stage.....all other salient details being constant....

So....if compensation (viz: foward-path unity gain bandwidth) remains unchanged in both instances, the diff. stage design should have greater stability margins than the CE type.....
 
I agree with Mikes :)

It is interesting. How can I use differential front end, but have big 2nd harmonic? In what part of differential should be modified?

There are at least two ways to make a LTP (long-tailed pair) distort. You see, it rather relies on the transconductance of each transistor being relatively constant. This usually means making the bias current of each transistor much larger than the signal current variation and making the bias current independent of common-mode signal.

To make the thing distort you do the opposite. Make the bias and signal currents of a similar size AND/OR put a low resistance in the tail to make common-mode signal vary the transconductance. This applies to both FETs and BJTs, although gm vs Id or Ic will have different characteristics. You'll have a significant common-mode signal if your amp is non-inverting and uses negative feedback.

[By the way, you can make an accurate analogue multiplier by applying one signal across the diff pair inputs and use another signal to modulate the tail current.]
 
What is the meaning of LTP? A differential with resistor current soruce?

Maybe Millwood wanted to say the single transistor input have wider bandwith, not faster. It is true for current feedback design, they usually have wider bandwith than voltage feedback.

How about DC offset in current feedback design (like in Borbely preamp). Is the DC offset as not as stable if we use ordinary voltage-feedback differential?

the single CE input stage with a resistive load delivers TWICE the transconductance of a similarly loaded diff. stage.....all other salient details being constant
Is this because in single transistor input all current are going into 1 transistor, while in differential, the current from CCS have to be divided to 2 transistors?

Make the bias and signal currents of a similar size AND/OR put a low resistance in the tail to make common-mode signal vary the transconductance
Aha, this is interesting. I have been wondering what this design like I attach aiming. At first I think it is to reduce the needed current entering the bases of differential (if the bias at BC560 is only 150uA, the base current needed certainly very-very low if we consider HFE).
But here Tradebam gave good hint about it. Is this design aiming at having big 2nd harmonic, because the bias at BC560 maybe as low as the audio signal entering the differential?

I liked this differential design. It can handle big bias current, but MAYBE it have great sensitivity, because signal is entering a very low bias (150uA) transistor.

I've got another amateur question. Which sounds better, differential with bipolars, with MOSFETs, or with JFET?
 

Attachments

  • differential amp mod.jpg
    differential amp mod.jpg
    16.3 KB · Views: 416
Disabled Account
Joined 2003
lumanauw said:
I liked this differential design. It can handle big bias current, but MAYBE it have great sensitivity, because signal is entering a very low bias (150uA) transistor.

as bias is a DC thing, you can use for example a mosfet to separate DC input resistance and AC input resistance: I am running a JLh with a mosfet input stage biased at about 15ma. beautiful sound.

lumanauw said:
I've got another amateur question. Which sounds better, differential with bipolars, with MOSFETs, or with JFET?

I have tried both: bjt and mosfet ltp. Quite frankly, i couldn't tell. But I try to run all stages of my amps with mosfets: mosfet input and mosfet vas. it allows me to run them very hot (10-15ma for input and 20-30ma for vas).
 
Hi, Millwood,
I have tried both: bjt and mosfet ltp. Quite frankly, i couldn't tell. But I try to run all stages of my amps with mosfets: mosfet input and mosfet vas. it allows me to run them very hot (10-15ma for input and 20-30ma for vas).
Yes, the interesting thing about using mosfet for differential is that we can run very high bias, without requiring DC current at the gate. But mosfets still have C input, that have tobe charged/discharged. It results in somekind of "AC current" needed to activate mosfet's gate.

What do you think about CFP differential like in post #17? We can have big bias differential with bipolars, but the input will need very low base current, due to very small bias current at the BC560.
 
lumanauw said:
In some writings, it is said that one of the advantage of having differential for input stage is that it can hold DC offset. It makes me think. Designs like JLH or NAD3020 power amp doesnt use differential at front end. They just use single transistor.
How do these amps hold their DC, without servo? (and if we dont use cap in the output). Will their DC wonders around when the amp gets hot?

Differential also cancels 2nd harmonic. What is the goal of the designer not to use differential. Is it to get overwheling 2nd order harmonic?


tHE only case I have heard on not using a diff. amp is in a tube amp where the linear operating region of the tube is huge. Frankly, less distortion, but $$$$$$$$$$$$$$$$$$$$$$$$$

I have made my own circuit designs from scratch and if you want to direct couple the speaker to the output stage, the best way to eliminate DC offset is to use the diff. amp to bias it because you can set a pot to adjust the bias on one of the transistors and not affect the AC. :D

Odd harmonics are caused be saturation of transistors, it is called a square wave. -----Fx= (1/N)sinN(wt)
------------> to infinity were N is an odd integer

As a transistor saturates and turns a sine wave into a square wave, high frequency odd harmonics are added to the signal. The differental amp must be able to create these high frequency harmonics at the opposite phase, so that they will mathmatically cancel out. Incidentally, a differential amp with high gain and high frequency responce(much higer than audio) like RF will be helpful in maximizing efficiency of your outputs if a global neg. feedback circuit is used. [feed the output back to the - input on diff. through a voltage divider]

Consequently, the driver circuit for the output devices needs to be over-rated so that they can feed these needed harmonics into the output transistors.

I have done some experiments with direct coupled Hi-Fi and this approach seems to work with getting more juice and less unwanted harmonic distortion and still get the wanted bandwith responce desired.
:cool: :cool: & not the unwanted and dreaded :hot: :hot: :smash:
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.