Quote from: fryingpan on October 15, 2024, 02:20:26 PMAntonis is your best friend. Or you are Antonis's best friend, I don't know.

Hmmm,let's see...

@mzy12: Could you elaborate the 2^nd term (40Ic x RE x Rb) in configuration B Zin calculation below. WITHOUT reading following edit note..??

https://www.diystompboxes.com/smfforum/index.php?topic=129250.msg1248337#msg1248337

Quote from: mzy12 on October 15, 2024, 02:18:16 PMCircling back to this (yet) again, would, having something like a 1M resistors in series with the base of Q2 not introduce a load of unwelcome Johnson noise?


EDIT: Also I don't think I'm even connecting it correctly in my schematic lol ;-;

A large cap across the resistor (as shown in my pic) prevent the noise.  The cap should be quite large since the "remaining" noise is then determined by the cap.

Something I forgot to say earlier.   When you decrease the current in the diff-pair it decreases base current and hence the input bias current.  The lower current will cause much less offset than the diff-pair operating at 3mA.    If you also use high hFE parts it will reduce that further.

Quote from: antonis on October 15, 2024, 07:12:29 AMHow did it come to op-amp/THD/noise circuits considerations..??

If you went through the trouble of building the buffer in this thread and it was noisy then it's not suitable for the task but with some targeted design decisions it's not hard to push it back on track.

That's why people use opamps.   The hard work is done for you.  When the NE5534 came out it certainly put an end to a lot of transistor designs because it takes some effort to beat its performance out of the box.

If the diff-pair is unbalanced then you might find the distortion of the diff-amp design isn't much better than the Sziklai type circuits (see fryingpan's post), which are much simpler.  Especially if you have a current source in the output on the Sziklai (similar to current design.)

Quote from: Rob Strand on October 15, 2024, 05:19:45 PMIf the diff-pair is unbalanced then you might find the distortion of the diff-amp design isn't much better than the Sziklai type circuits (see fryingpan's post), which are much simpler.  Especially if you have a current source in the output on the Sziklai (similar to current design.)

Of course, then you can go to Sziklai/CFB diffamp pairs... 

Although this is not strictly about buffets, I think you may find them interesting.

https://sound-au.com/amp_design.htm

This article is an interesting read. It focuses on power amps, but what is a power amp if not a (relatively rudimentary, high powered) opamp? And I've personally found that these days most discrete design pointers can actually be found in the power amp realm (because, after all, it's the only realm where you actually have to employ discrete transistors...).

In the input stages section it specifically compares a long tailed pair (granted, a simpler kind) to what is basically a Sziklai pair. The advantages of the simpler topology are basically the higher inherent stability and actually marginally better gain.

https://sound-au.com/project37.htm

This is a relatively simple preamp for hifi applications (low impedance, granted) which achieves very low noise and very good distortion.

https://sound-au.com/project13.htm

Again, a low input impedance microphone preamp which similarly achieves a lot out of very little.

Quote from: antonis on October 15, 2024, 04:27:41 PM@mzy12: Could you elaborate the 2^nd term (40Ic x RE x Rb) in configuration B Zin calculation below. WITHOUT reading following edit note..??

Hmm well Rb is the the bootstrap resistor and RE is the impedance that an AC current will see at that node, 4k7||180k(||The load impedance isn't included in your schematic but in reality it appears here too. Should be negligible. The 40 part in the 40Ic is giving me pause, is it related to the current flowing through the 100k resistor under the +9V?

And that's as far as I can get without looking at that edit note

Quote from: R.G. on October 15, 2024, 05:47:25 PMOf course, then you can go to Sziklai/CFB diffamp pairs... 

So many options hahha

Quote from: fryingpan on October 15, 2024, 06:15:55 PM...

...Again, a low input impedance microphone preamp which similarly achieves a lot out of very little.

Thanks for more resources fryingpan. Will look into it

Quote from: mzy12 on October 15, 2024, 07:35:27 PMHmm well Rb is the the bootstrap resistor and RE is the impedance that an AC current will see at that node, 4k7||180k(||The load impedance isn't included in your schematic but in reality it appears here too. Should be negligible. The 40 part in the 40Ic is giving me pause, is it related to the current flowing through the 100k resistor under the +9V?

If (40Ic x RE x Rb) has units of [ohms] then 40Ic must have units of [1/ohms], or, 1/(40*IC) must have units of [ohms]

When you deal with BJT's and diodes you will come across the dynamic resistance.  For a transistor, the dynamic emitter resistance is,

      re = Vt / Ic

where Vt = 26mV call the thermal voltage, and Ic = collector current
(In reality Vt in increases with temperature.   I'll let you read up on it:
https://en.wikipedia.org/wiki/Boltzmann_constant#Thermal_voltage)

The 40 comes from 1/26mV = 38.   1/re = Ic/Vt = 38 * Ic.

From the units if Vt has units of voltage the formula for re has units consistent with ohms.

If you study transistors and solid state device these ideas come up all the time.

Trying to get the offset bias stuff to work. Not sure if I'm doing it right because even though I got the output offset voltage close to zero, the input offset voltage has remained unchanged.

Quote from: Rob Strand on October 15, 2024, 08:03:54 PMWhen you deal with BJT's and diodes you will come across the dynamic resistance.  For a transistor, the dynamic emitter resistance is,

      re = Vt / Ic

Aaah no wonder it caught me. I'm familiar with re (we love little re ), but in college we haven't got onto dynamic resistance in semiconductor's yet, or much to do with Boltzmann. Thanks rob.

EDIT: Looking back on last years notes we were just told that re = 25. We weren't given an explanation on that, which I guess makes sense because we were first years.

Quote from: mzy12 on October 15, 2024, 08:50:43 PMTrying to get the offset bias stuff to work. Not sure if I'm doing it right because even though I got the output offset voltage close to zero, the input offset voltage has remained unchanged.

The bias compensation only fixes the output offset.  The two inputs should sit at about the same voltage but they won't sit at 0V.    Don't forget your DMM will load down the voltages on the base.

The output voltage is at 0V and the grounded side of the 1M input input resistor is ground (0V).  If you think about the base current flowing through the 1M resistors then the voltage on the bases must be negative (due to the Ib*1M drop)

The only way to stop the negative voltage on the bases is to add resistors from the each of the bases to +V.  The idea is if these added resistor pass a current equal to the base current then no current needs to flow down the 1M resistors.  Some IC's use this trick internally (except they use current sources instead of resistors).  If the transistor hFE's aren't matched then you need different resistors on each transistor base.  Normally for audio you don't bother.  For AC signals the DC shift at the input doesn't matter.

---

I edited the post and added a few more details.

Quote from: mzy12 on October 15, 2024, 08:54:06 PMAaah no wonder it caught me. I'm familiar with re (we love little re ), but in college we haven't got onto dynamic resistance in semiconductor's yet, or much to do with Boltzmann. Thanks rob.

EDIT: Looking back on last years notes we were just told that re = 25. We weren't given an explanation on that, which I guess makes sense because we were first years.

If you are studying this stuff it gets burnt into your forehead.

The re=25 thing was probably an example to show that there's a bit more to the story.  re=25 only true at Ic=1mA (1mA is very typical for examples).

Quote from: mzy12 on October 15, 2024, 08:54:06 PM(we love little re )

But Emitter followers hate it..
(actually, their destation is mainly moved towards its existence and partially towards its value 'cause they put the blame on working current..)
Same stands for CE amps (for restricted maximum gain and non-linearity reasons) and low-noise designers (for Collector current shot noise through it, which BTW, equals to Johnson noise of a fictitius resistor of r_e/2 value..)

P.S.
I think we push moderators indulgence to their limits so I stop right now..

What I am very confused about is why Douglas Self removed the bootstrap configuration for BJT designs that was present in the previous editions of the same book. I just got a look at the third edition of Small Signal Audio Design and on page 124, he describes a similar bootstrap circuit that was recommended to me on this forum.

Quote from: mzy12 on October 15, 2024, 01:22:22 PMThey're still far more expensive than BJTs though. A single unit of the cheapest JFETs mouser has to offer could get me like ten MMBT3904s.

True, but you've also decided to build an overcomplicated unity gain buffer using seven transistors, rather than like two or three. So parts cost is clearly not a serious consideration here.

Quote from: merlinb on October 16, 2024, 09:36:47 AMTrue, but you've also decided to build an overcomplicated unity gain buffer using seven transistors, rather than like two or three. So parts cost is clearly not a serious consideration here.

Fair enough. I will try to find a JFET solution that uses less parts and gives satisfyingly similar THD results.

This is for guitar, right? You seem very concerned about triple-zero THD which makes me wonder if this is for something else?

Quote from: mzy12 on October 16, 2024, 10:03:35 AM

Quote from: merlinb on October 16, 2024, 09:36:47 AMTrue, but you've also decided to build an overcomplicated unity gain buffer using seven transistors, rather than like two or three. So parts cost is clearly not a serious consideration here.

Fair enough. I will try to find a JFET solution that uses less parts and gives satisfyingly similar THD results.

If you look at Doug Self's books they tend to show an evolution of designs.   Designs have certain flaws and there is a technique or circuit which counters that flaw.    Not all application need to remove all flaws.

The Boss Waza buffer follows a similar evolution.
https://www.diystompboxes.com/smfforum/index.php?topic=122929.msg1161023#msg1161023

The simple JFET buffer has a gain loss due to the finite output impedance driving the source resistance required for biasing.

Replacing the bias resistor with a current source reduces the gain loss and also THD.   (Not unlike the evolution of the BJT circuits in Self's book BJTs).

When a load is placed on source the load effect returns.  Basically the JFET buffer doesn't have a low impedance.  So an additional transistor buffer is added.  That does better than the JFET.   However the BJT is just an ordinary BJT buffer without any fixes, so you are back to square one on Self's evolution of BJT designs.

As drawn the Waza input buffer is only suitable for input signals.  The output buffer doesn't have any protection against oscillations with capacitive loads.  While the output impedance is low the ability to drive low impedance loads is limited.  A low impedance load will limit the available output swing.

Interestingly the Sziklai type buffer with a JFET at the input offers the potential for high input impedance, low distortion, low output impedance.   Because it's class-A the output drive is limited by the  bias current of the output transistor.

The problem with open ended designs is they tend to drift towards ideal complex designs.   When you have an application you can strip back things to the bare requirements and end up with something which is only complicated as it needs to be.

As a minor side note, for designing a pedal, I often assume that the parts on the PCB are free, unless there is some fancy high end IC, inductor/transformer, or other oddity. They're not really free, but generally the cost of Rs, Cs, jelly-bean ICs and so on totals up to within the rounding/uncertainty errors in estimating the enclosure, painting/finishing, jacks, switches, pots, wrapping paper, box, and all the rest.

I refine this as I go, of course, but it's been very good for a first estimate. A complex design with a dpzen transistors is not much different from a three-transistor circuit overall, unless the complexity tempts you to put in pots and trimmers to tune it in.

Here's an example of a Sziklai type buffer (ie. one with feedback) and a JFET input used in a commercial product:

https://adadepot.com/mods/ada_mp1a/ADAMP1pre.jpg

They chose this over the NE5532.  Probably a low noise JFET.

Quote from: Rob Strand on October 17, 2024, 04:56:02 AMHere's an example of a Sziklai type buffer (ie. one with feedback) and a JFET input used in a commercial product:
https://adadepot.com/mods/ada_mp1a/ADAMP1pre.jpg
They chose this over the NE5532.  Probably a low noise JFET.

Can't clearly see what it drives (other than op-amps) but about 16mA (rough estimation) seems to me a lot of current..