I havn`t see something similar to this before. Thanks to IvanGrig from vegalab phorum.
So, what do you think about this? It can be easily combined with darlingtons or mosfets output stage for power amplifier application.
Fully symmetric, low parts count, very good performer at simulators.
So, what do you think about this? It can be easily combined with darlingtons or mosfets output stage for power amplifier application.
An externally hosted image should be here but it was not working when we last tested it.
Fully symmetric, low parts count, very good performer at simulators.
Kuzmenko said:
Fully symmetric ... well, there is no such 'free lunch'
and PNP and NPN transistors are not made in the same way.
So they are always different in several parameters.
If this circuit performs better than our known and less complicated ones,
then good!
At least this one may have some ideas we could take, and make something better out of.
I have stated, I like to see peole who can THINK OUT OF THE BOX.
I never use cascading.
I always try for other solutions, like find transistors that operates better without cascade.
Why use 2 transistors, to do a 1 transistor job ??????
I can see this one circuit, in my opinion, is a bit sensitive to supply noise.
Which is almost the same as low PSRR (Power Supply Rejection Ratio ).
-------------------------------------------------------------
Questions:
Is this in use in any Amplifier?
If so, what was the result?
Regards
lineup
Hi
There is something I fail to understand: basically, the circuit is made of two differential amplifiers, but with complementary transistors in a series configuration, instead of the usual identical transistors in parallel. The inverting input is at the bases of Q7/11, and the non-inverting at the bases of Q10/9. But for the amplifier to operate correctly, the NI input has to be kept at the ground potential; here the only coupling to the ground is via the bias resistors R34/35, and is quite loose. Is this intentional? If not, the open loop gain of the circuit could be greatly improved by substituting two capacitors (or two strings of diodes) to the ground in place of C8.
LV
There is something I fail to understand: basically, the circuit is made of two differential amplifiers, but with complementary transistors in a series configuration, instead of the usual identical transistors in parallel. The inverting input is at the bases of Q7/11, and the non-inverting at the bases of Q10/9. But for the amplifier to operate correctly, the NI input has to be kept at the ground potential; here the only coupling to the ground is via the bias resistors R34/35, and is quite loose. Is this intentional? If not, the open loop gain of the circuit could be greatly improved by substituting two capacitors (or two strings of diodes) to the ground in place of C8.
LV
Elvee said:
yes, it might look like that.
But I am sure in this case the feedback is for an inverting operation.
This makes input impedance 4.7K and R32 / C12 is the feedback.
The input will create voltage/current changes in R26 and R27.
These resistors R26/R27 are connected to a defined operation voltage
by the divider R34, R36, R35 which is between V+ and V-.
So this would work.
But question is how good?
V+ and V- should be at very precise voltages.
This is why I wonder about the PSRR of this circuit.
I would say we need very good regulated supply for this amplifier.
But let's hear from some other people here ..
Regards
lineup
The potentials of the bases Q8+Q9 and Q10+Q12 are NOT fixed to the ground to avoid flowing currents thru signal ground. So, in this design there is no smallsignal ground at all 
There is only 1 powerground node on amplifier`s pcb. Because of this you can`t do mistake when you would route pcb`s traces
About "poor PSRR": are they? It`s not poorer than traditional diff+va design.
There is only 1 powerground node on amplifier`s pcb. Because of this you can`t do mistake when you would route pcb`s traces About "poor PSRR": are they? It`s not poorer than traditional diff+va design.
Re: Re: Re: unusual design of input stage
If you understand better what I means you can call it anything.
There are plenty of others around, that can't comment on topic subject,
but only post about all us non-english spelling mistakes.
you are nott the only one, I am afraidly you are nott.
So how much better would anyone reader get my message
if I is spellingly one letters vrongg?
You tell me, QSerraTico_Tico
This is what I said:
I never use cascading.
I always try for other solutions, like find transistors that operates better without cascade.
Why use 2 transistors, to do a 1 transistor job ??????
And when did I say anyone should do things my way?
What would I be able to do to stop them from Kaskading transistors
when there are good alternatives to keep stages simple.
There are benefits with KASckading, but there are also several drawerbacks.
This you would know, too.
So to avoid these drawbacks I will live with a more Cleene And Shimply cirrkut.
I understand that you are a kaskading man
This is why you feelt I was critisissing jou.
But I did not critisise nobody. Why should I?
If they want to kassekuding everything, so let them ...
I only told my opinion, how I do things
and I asked:
Why use 2 transistors to do a 1 transistar's jobb?
Splitting bases of Q7 and Q11 and apply feedback in another way?
Maybe it would work,
but this would make a different circuit.
And some of the point with this new idea would be lost.
Can ask the question in this way:
If it is better to separate Q7 and Q11
why did not this IvanGrig do this ....
If it is performing better.
Regards
lineup
QSerraTico_Tico said:
If you understand better what I means you can call it anything.
There are plenty of others around, that can't comment on topic subject,
but only post about all us non-english spelling mistakes.
you are nott the only one, I am afraidly you are nott.
So how much better would anyone reader get my message
if I is spellingly one letters vrongg?
You tell me, QSerraTico_Tico
This is what I said:
I never use cascading.
I always try for other solutions, like find transistors that operates better without cascade.
Why use 2 transistors, to do a 1 transistor job ??????
And when did I say anyone should do things my way?
What would I be able to do to stop them from Kaskading transistors
when there are good alternatives to keep stages simple.
There are benefits with KASckading, but there are also several drawerbacks.
This you would know, too.
So to avoid these drawbacks I will live with a more Cleene And Shimply cirrkut.
I understand that you are a kaskading man
This is why you feelt I was critisissing jou.
But I did not critisise nobody. Why should I?
If they want to kassekuding everything, so let them ...
I only told my opinion, how I do things
and I asked:
Why use 2 transistors to do a 1 transistar's jobb?
QSerraTico_Tico said:
Splitting bases of Q7 and Q11 and apply feedback in another way?
Maybe it would work,
but this would make a different circuit.
And some of the point with this new idea would be lost.
Can ask the question in this way:
If it is better to separate Q7 and Q11
why did not this IvanGrig do this ....
If it is performing better.
Regards
lineup
Off course there is a small signal ground: it may be implicit, but it's definetely there, it sits at the junction of the + and - PSU filter caps. Making an amplifier without reference node is like lifting yourself in the air using your own bootstraps: impossible, except in the case of unity gain buffer.Kuzmenko said:The potentials of the bases Q8+Q9 and Q10+Q12 are NOT fixed to the ground to avoid flowing currents thru signal ground. So, in this design there is no smallsignal ground at all
About "poor PSRR": are they? It`s not poorer than traditional diff+va design.
Having the ground node at such an inconvenient place is certainly not an advantage: it means for instance you've to take your input ground from there; it also means there will be unwanted feedback paths via the supply leads impedances: when high currents are drawn, parasitic voltages will appear on the NI input.
If you're not convinced, try to include realistic impedances for each of your wires in your simulation: 0.1ohm/0.5µH for example.
Also, have you tried to calculate your PSRR? It's easy: R34 and R35 having equal values, it is 6dB seen from the input, but this is then amplified by the non-inverting gain of the amplifier, 22.3 in this case; this means a resulting PSRR of 6-27=-21dB!! (yes, minus!).
LV
It will:thanh said:
a) Become totally unpredictable because one of the current sources will always have a slightly larger current than the other, pulling the NI node (and consequently Vout) all the way to V+ or V- (the loudspeakers will appreciate).
b) Lose practically all open loop gain if the current sources are of a good quality.
The NI node needs an impedance as low as possible; the Thevenin impedance of R34//R35 is already bad, but with current sources it will become disastrous.
Note that I have nothing against the complementary differential topology; I have used it quite often myself, but using an unusual topology doesn't allow one to dispense with sound design rules.
LV
i agree with Elvee, the inverting input is implicit, and sees any difference between + and - rail. At the very least, C8 should be split into two caps in series with the idpoint connected to ground. Even better, a proper stiff voltage reference should be used.
Also, Q8 and Q12 are not really necessary except if the voltage rails are outside the Vcemax or Pdmax of Q11 and Q7.
@Lineup: what is your aversion towards cascodes? (You say you NEVER use them). They don't really do a 1 transistor job, they do a 2 transistor job, since they are 2 transistors - they have their uses. but of course, i do agree with you that people sometimes tend to cascode everything, even when it does not justify the achieved improvement. i also find they are amongst the most trouble-free 2-transistor constructs
Also, Q8 and Q12 are not really necessary except if the voltage rails are outside the Vcemax or Pdmax of Q11 and Q7.
@Lineup: what is your aversion towards cascodes? (You say you NEVER use them). They don't really do a 1 transistor job, they do a 2 transistor job, since they are 2 transistors
- they have their uses. but of course, i do agree with you that people sometimes tend to cascode everything, even when it does not justify the achieved improvement. i also find they are amongst the most trouble-free 2-transistor constructs Ok, this is not a problem, here you are!
R36 R37 can be substituted with lighting diodes or zeners. But resistors are cheaper
An externally hosted image should be here but it was not working when we last tested it.
R36 R37 can be substituted with lighting diodes or zeners. But resistors are cheaper
Looks much better now.
Resistors may be cheaper than diodes, but the difference is, in my opinion, negligible and anyway, resistor+47µ cap is probably more expensive than a diode.
There is a good reason to stiffly set the voltage at these nodes: any inbalance in the supply rails voltages will cause a DC offset at the output.
LV
Resistors may be cheaper than diodes, but the difference is, in my opinion, negligible and anyway, resistor+47µ cap is probably more expensive than a diode.
There is a good reason to stiffly set the voltage at these nodes: any inbalance in the supply rails voltages will cause a DC offset at the output.
LV
V
Vek
I threw it in a LTspice file to have some fun, just in case somebody wants to play around with it ...
Some values differ from the last posted schem and I added the zeners. This could be a good base to illustrate changes to this circuit.
L1 in the feedback network is for easy switching between closed/open loop behaviour without loosing dc-reference. Just delete the short and you are open loop.
Have fun!
Some values differ from the last posted schem and I added the zeners. This could be a good base to illustrate changes to this circuit.
L1 in the feedback network is for easy switching between closed/open loop behaviour without loosing dc-reference. Just delete the short and you are open loop.
Have fun!
ilimzn said:
I have no aversion against cascading, in special.
It is just that I like simplicity so much
It would take very much to make me use 2 transistors
when I can do Well with 1.
Same thing, if adding one cap somewhere
and distortion drops only less than 1 dB.
I find it is not worth it. Because probably this cap can cause some other little trouble.
Cascading means not only add 1 transistor, as you know.
You also add a rail from power supply. And resistors and maybe also filter cap.
One reason, in my opinion, why simple amplifiers performs so well in compare
is they have fewer components.
And fewer rails for injections of troubles.
When you have nothing. This nothing can not cause any disturbance.
As soon as you add something, you have a media for disturbance
as well as for the good you have added.
This is my logic explanation for why simple, which theoretically should be inferior
in practical reality so often turns out so well.
While most complicated amplifiers can be almost impossible to tune
and get right.
It may be too many things to cure.
This is also in line with, why Single ended amplifiers, using only 2 power supply rails
V+ and 0V=Ground
can have such good sound.
While dual supply, 3 rails, have not only several paths for supply currents
but also more paths for unwanted 'feedback' and disturbances.
There are a lot of small currents, garbage flying around in a an amplifier
in motion.
Not only at audio frequencies.
thanks for asking, ilimzn
hope you get my thinking
lineup 'Keeps It Simple' man
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.