diyAudio (
-   Solid State (
-   -   Input stage idea (

lumanauw 23rd July 2005 07:46 PM

Input stage idea
2 Attachment(s)
I never saw a diamond differential before. NP just submit it.
Like full complementary differential, but all the transistors are altered, N-P, P-N. Quite strange to me.

I interested in the part of "where you see the diff pairs operating push pull without limitation to the potential current they can deliver ", "This allows for the current to deliver enormous slew rates.", but advoiding this part "running an input device in Class AB is not a popular approach."

A differential (LTP) is really elegant. Only 2 transistors, but can control the whole amp and can control DC offset.

Has anyone tried this input stage? At a glance it resembles a differential, but it is not. It doesn't do "differential-ing". But it can control the whole amp and can control DC offset, like the main function of a differential.

In a differential, total current (I1+I2) will always be the same. This means if I1 is increasing, I2 will decreasing. In bipolar transistor, if I is rising or decreasing, the VBE value will also follow it.

But in this idea, if I1 rising, I2 will also rising, not decreasing. This means there is no current limiter for the input stage. Current limitation is determined elsewhere (in my drawing it is limited by CCS1 and CCS2). This also helps the input stage to be always biased.

The main fuction is done by transistor A (+C, forming 2 VBE's), in the form of changing voltage between it's base and emitor. If Y is fed by input, then X have to be fixed somewhere. X is fixed by transistor B-D, and with help of transistor C it reaches emitor of A. By VBE drops adding and substracting, automaticly the DC offset will be at ground point. I can expect this, because the current in the path I1 and I2 will always the same due to current mirror, that means all the VBE drops between left path and right path will be the same value(s)

R1 and R2 is needed there, to prevent oscilation. It may be a small valued resistor.

I'm using folded cascode for the next stage, so it perform inverting amp with gain of RF/Rin.
Funny thing is, it doesn't matter where to put the folded cascode. In both path I1 or path I2 it forms the same inverting amp.

A differential will cancel harmonic. Is this idea have the same properties (canceling harmonic), or do nothing to it, or adding harmonic? Anyone can guess what it sounds like?

darkfenriz 23rd July 2005 08:55 PM

what you've drawn can hardly be named 'diamond differential'.
I'd say it is not differential at all, the amp is inverting so you don't use differential, you don't substract, you add voltages by Rf and Rin.
Your input stage probalbly could be best described as common emitter, cause emitter of C is at constant voltage potential. E and F are probably redundant.

lumanauw 24th July 2005 01:52 AM

Hi, Darkfenriz,


I'd say it is not differential at all
Yes, it is not diamond differential, not a differential at all :D. I just trying to get the merits of diamond differential but not having it's drawback, so I end up with that drawing.

Your input stage probalbly could be best described as common emitter, cause emitter of C is at constant voltage potential
Yes, this one is for functioning one important aspect, to control the whole amp.

E and F are probably redundant.
This one is for another important function, that is maintaining DC offset regardless of thermal and AC condition.

With close thermal coupling, the DC offset can be maintained closely to 0, because transistor A+C will have the same current as B+D, due to current mirror done by E+F.

If we do not place E+F, we have to put a CCS on the emitor of D, and collector of C can go straight to (-rail). But this way, the DCoffset tracking is worse, because the VBE drop of A+C is not exactly the same as VBE drop of B+D. Putting E+F ensures I1=I2, so VBE drop of A+C = VBE drop of B+D

Tom2 24th July 2005 03:59 AM


I actually designed and built a mm phono preamp and line stage
preamp based on the input stage idea back in 1990!

I only built one channel and used a modified Jung-Didden regulator
design with it.

The sound I remember seemed very good and the regulator
improved the sound very much compared to a standard 317/337 regulator.
I built the circuit on Radio Shack experimenter plugin boards and
somehow I got the circuit to be stable with 100pf caps and
lots of ferrite beads on the supply lines. I did this with no scope
and just a Radio Shack analog meter. I tested for oscillations by
using a small fm radio nearby. I hope the local airport nearby didn't notice.

The input stage from my perspective is derived from
current conveyor toplogies. On your schematic c,d,e,f form a
first generation current conveyor(or CCI). I got the same idea
from the CCI and noticed like you the Vbe equalizing effect
due to the two current mirrors c,d and e,f. This is the idea
behind the CCI topology.

For the input stage I used jfets. At the time I use two paralled
J110's pair for the mm stage and an pn4393 pair for the main preamp.

The lower current mirrors I used a sort of Wilsonized Wilson. A four
transistor pair that cancels base current effects fairly well up to
about 10ma and with emitter resistor. The circuit was made popular by
Brett Wilson back around 1990.

Thus I had an input pair of j-fets and 8 transistors for the lower two
current mirrors.

My output was different than yours. I use an emitter follower with a common base stage
as sort of a level shifer with current and voltage gain. This feed into a current
source like yours connected to the output. I ran my stages as non-inverting with
integrators to provide dc stablity and get rid of coupling caps.

I have drawings on C sized vellum. If you wish I could make a similar schematic
using the Ltspice program I have and post it.

I considered the input circuit to be a perfect first-order circuit when using jfets.

The main problem I had was a latch up problem with turn on and circuit stability with
all those transistors in the current mirrors.

Also look at a posts I made. It uses ideas also from current conveyors.


lumanauw 24th July 2005 08:43 AM

Hi, Tom2,

I didn't realize that I draw a Current Converyor-CCI (C,D,E,F) as the final result :D
At fist I only draw A,B,C,D, collector of C is straight to (-rail) and D is fed by CCS. D is only to make voltage reference, so this transistor is arranged as Diode.
But later I think about 0 DC offset stability is important, because I will build it as power amp with DC coupled output to the speaker, so I add E,F. The result, C,D,E,F is forming CCI, without my intention at all.

Many creative designs are limited use to preamp only. Like diamond buffer, very few makes it power amp (like SonnyA).

I'm hoping I can get good DC offset with this idea, like what can be done by usual differential input, because it is one important aspect for DC coupled output amplifier.


I have drawings on C sized vellum. If you wish I could make a similar schematic
Can you do that?:D I can't use simulator, but wanted to see what the harmonic pattern/ FFT looks like for my idea in #1.

Also, what is your "mm phono preamp and line stage preamp based on the input stage idea back in 1990!" looks like? those 2 links you provides seems not about it, the PassHawk still uses differential pair, the SuSyfun1a, I have to study it first, but at a glance it doesn't have 0 DC reference, and I8-I10 and I3-I4 is forming 1 CCS?

Tom2 24th July 2005 10:24 AM

2 Attachment(s)

Here is the schematic of the line stage.

The mm stage is similar with an riaa feedback network and using
J110's for lower 1/f noise.
I do not claim the circuit to be perfect.

Ltspice is free and easy to use.
I'll try to do an FFT related to your circuit.
Hopefully I can make it stable.


Tom2 24th July 2005 10:57 AM

2 Attachment(s)

Here is a buffer stage idea.

I1, I2, Q5, M5 can be high voltage devices.

Q1-4 and J1,2 can be low voltage devices.


lumanauw 25th July 2005 04:52 PM

Any commercial product use this kind of front end?

Tom2 25th July 2005 10:54 PM

2 Attachment(s)

Lumanauw said
Any commercial product use this kind of front end?
To my knowledge I have never seen it used. I have never
used it in a product or seen any US patents using it.

The circuit always seemed like a cool idea to me.

The attached zip file contains the front circuit like you posted
with no buffer stage or load. It has ideal current sources.
The input voltage is 1v at 1khz and the output is about 11 volts.
Also in the zip is the fft. Accuracy of the simulator????


lumanauw 26th July 2005 01:30 AM

Hi, Tom2,

Thanks for the FFT :D . There seems a thing that bothers me about this CCT, that is the high-order harmonics. The harmonics below 10khz is OK, but why it also generates harmonics in frequency over 10khz, and quite big too?

At first, I hope this cct to have one merit of the X1000 SuSy. NP said that X1000 and X600 uses the SuSy like in the patent itself, that is using 2ccs in the input stage, with a resistor bridging between emitors. This SuSy doesn't use 1 differential like other X amp, but rather using 2 ccs.

Some make explenation why this topology is maintained in X1000 and X600. Usually it is about easiness to adjust current balance and DC offset with 2 ccs.

But I think with X1000 approach, it also has one important properties that is not obvious. That is the less VBE cancelation between 2 transistors in differential. I hope this topology can boost the properties of not having "VBE cancelation" and can take advantage of the sound result.

I try to make explenation about "VBE cancelation", in very simple way. If you look at the drawing in post#1, in the left side I draw a differential. Imagine it is used as an inverting amp, where the base of transistor B is grounded. The equilibrum of the junction emitors is at -0.6V, input signal at 0V, because both TR shares the same Ibias, so both has the same VBE drop.

If there is positive input signal of 0.1V, the base of TR A will be a little more positive, lets say 0.1V above ground level. This makes the TR A have VBE of 0.7V, and more current is in I1 path, I2 has less.

VBE value is tightly related to the current passing IC.
Because I2 has less current passing through it, it makes the VBE of TR2 become less than 0.6V, lets say only 0.5V. Now, if VBE of TR2 is only 0.5V, and its base is tied to ground, that makes the emitor junctions is sitting at -0.5V. If the junction is at -0.5V and the input signal is at 0.1V, in the next cycle, that makes TR1 experience only 0.6V, from first 0.7V.

The differential is inherently inhibiting TR A from having excitation, making harmonics to be canceled.

This is a very simple way from me trying to explain what I mean with "VBE cancelation". It will not acceptable in theory discussion by EE's :D

You can see, a differential (like post #1) will always cancel harmonic, because the inherent properties of the differential itself. This harmonic cancelation will always happened in the differential, even in a non-feedback power amp. It happens because differential has only 1 ccs that has tobe shared between 2 paths.

In X1000 SuSy, this effect is less than it is in a differential. And with this input, this effect is nullified, even reversed, the topology will boost harmonics, not cancel harmonics (like in any differential).

I'm making a pcb for this topology right now, I'm using it for power amp. Hopefully, it will sound different than any amp that uses differential as input stage.

All times are GMT. The time now is 07:58 PM.

vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2015 DragonByte Technologies Ltd.
Copyright 1999-2015 diyAudio

Content Relevant URLs by vBSEO 3.3.2