CJ

Matching Transistors for Valley People MicPre
« Reply #60 on: May 07, 2005, 04:08:25 AM »
That makes sense. Also, what ever I am driving will have either a transformer or a cap input, ehh?

Weird, I breathe on the current source and the output voltage goes from 22mv to 12 volts! I need to pot that current source in epoxy.

Thanks for the help! :thumb:
If I can't fix it, I can fix it so nobody else can!
Frank's Tube Page: www.mif.pg.gda.pl/homepages/frank/vs.html
Guitar Amps: http://bmamps.com/Tech_sch.html


bcarso

Matching Transistors for Valley People MicPre
« Reply #61 on: May 07, 2005, 02:11:34 PM »
"Weird, I breathe on the current source and the output voltage goes from 22mv to 12 volts! I need to pot that current source in epoxy."

It (the current source anyway) shouldn't be that sensitive to either heat or humidity.  You might want to use a little heater that is more precise (a warm 1/8W resistor maybe) and probe around.  Is this effect happening at the highest gain?

CJ

Matching Transistors for Valley People MicPre
« Reply #62 on: May 07, 2005, 04:55:36 PM »
I will check it again. It may be just the humidity on the high impedance input. So maybe a coating treatment might be in order after I get this thing tweaked.
If I can't fix it, I can fix it so nobody else can!
Frank's Tube Page: www.mif.pg.gda.pl/homepages/frank/vs.html
Guitar Amps: http://bmamps.com/Tech_sch.html

bcarso

Matching Transistors for Valley People MicPre
« Reply #63 on: May 07, 2005, 05:11:09 PM »
The reason I say that is that the i sources are "ballasted" with lots of resistance in the emitters.  So changing the temp of the transistors will change Vbe, but that Vbe change will affect the current by delta Vbe/R, which should have a small effect overall.

But it's hard to use a self-produced airstream to selectively heat/cool parts that close together.  And this comes from someone with a well-deserved reputation for hot air (fa dumph!).

CJ

Matching Transistors for Valley People MicPre
« Reply #64 on: May 07, 2005, 05:47:41 PM »
Yeah, I think it was the three pots of Hills Bro's.

Hey, I have a question for you:

By adding x-tra transistors to the array, will the mic still drive it?

I mean, the transistor is a current controlled device, so I need more base current now that I have 8 transistors instead of four.

Can I calculate base current needed from Hfe, like 4 ma per side, Hfe=100, so I need 1/100 th of 4 ma. which is 40ua to drive each side?

How much volts and amps does a typical SM57 put out into a 600 ohm resistor? Pretend it's Ella singing. No wait, she used to break wine glasses. Pretend it's Dino.
If I can't fix it, I can fix it so nobody else can!
Frank's Tube Page: www.mif.pg.gda.pl/homepages/frank/vs.html
Guitar Amps: http://bmamps.com/Tech_sch.html

bcarso

Matching Transistors for Valley People MicPre
« Reply #65 on: May 07, 2005, 06:14:45 PM »
"Can I calculate base current needed from Hfe, like 4 ma per side, Hfe=100, so I need 1/100 th of 4 ma. which is 40ua to drive each side? "

Yes.  And your offset issues btw may have more to do with base current shifts with temp appearing across the base bias R's than anything else---beta changes by about 0.5% per degree C.  Having the mic on there "shorting out" the two inputs to each other (if I recall the mic is d.c. coupled?) will make this effect much smaller, as it converts it to mostly a common-mode change for which you have rejection.

Whether adding more Q's will help noise or not, is the issue of noise current going up while voltage noise is going down---it depends on the source impedance through which the noise current flows as to which noise source is dominant.

PS: I assumed when you said you would add more transistors that you would be upping the total current as you did so.

PRR

Matching Transistors for Valley People MicPre
« Reply #66 on: May 08, 2005, 02:27:10 AM »
> Weird, I breathe on the current source and the output voltage goes from 22mv to 12 volts!

Something is VERY wrong.

> R6, the current source resistor, can be tweaked to help move voltages around.

I do think Q4 should run much higher current, R6 more like 10K and R2 dropped to get back to the design current.

I would short-out R4 R5 for DC testing. If you screw around enough, they will slam the collectors of Q3 Q5, which will throw your current WAY off. They may (or may not) be needed for lowest noise, but serve no DC purpose.

Also I would really drop R9 and its counterpart (missing from your schematic???) from 100K to 10K. There is no reason to use such large values, and they are giving huge DC drop that will vary with temperature. This defeats all the Vbe matching you did. Now Beta match is important, and Beta is highly variable over temperature and time.

I would not try too hard to match the original voltages. Get the current right, and keep the voltages roughly right. The voltage on Q1 Q2 collector is not important, as long as you have a few volts across Q1 Q2.

> Since I dropped the collector resistors to about 2.2k, the ac gain of the opamp will go up, so I may need to lower the 3.3M feedback resistors.

No. Overall, while opamp AC gain goes up, transistor gain goes down, total gain is unchanged.

> I can not connect the plus and minus outputs directly to a buffer without coupling caps because of the offset voltage on the outputs of the instrument amps, right?

Major WRONG here. The output offset should NOT be 22mV. You can't swap opamps willy-nilly. This design DEMANDS "zero input current" opamps. TL072 is perfect. LT1078 is a terrible choice for several reasons, but it does have low input current. Drop in a 5532 and you will get a couple volts offset. Change the 3M3 resistors, and you lose bass until you up the 0.68uFd caps. To be halfway tolerable with 5532, you will be at 50K and 36uFd... at that point you would be better off doing it the conventional way, with DC coupling from BJT to opamp and a 2,200uFd NP in series with the gain pot.

The TL072 is the best opamp here. Selected TL072 will give less than 2mV output offsets, and if your next input can't take that then it deserves to die. The TL072's  main drawback is that you can't hang 600Ω on the outputs without buffering. 10K is fine. There may be stronger FET-input audio opamps that will drive 600Ω (300Ω per side) with ease.

> By adding x-tra transistors to the array, will the mic still drive it?

Of course. At the same total emitter(s) current, the base current is unchanged. Or very slightly changed because Beta will be different at the smaller current, but hardly-at-all for these parts at 2mA or 0.2mA.

You are increasing input capacitance, but even a gaggle of TO92 BJTs won't have as much pFd as a mike cable. And mikes are 200Ω just so they can drive a long run of cable.

And if there is any feedback happening, the dynamic input impedance is FAR higher than the naked transistors. The naked transistors are about 4,000Ω across the bases. With any proper audio opamp you will have 50:1 of feedback. The input impedance will approach 200K, in parallel with the 2*100K (or 2*10K) base resistors, in parallel with the 2*6K8 Phantom resistors. The base-base impedance is just not an issue.

> How much volts and amps does a typical SM57 put out into a 600 ohm resistor? Pretend it's Ella singing.

Relax. I had a date with Ella 20 years ago and she didn't show; odds are even lower now.

The '57 is 300Ω. If I put it in front of a lecturer I'll expect ~1mV open-circuit. If I load it in 600Ω I expect 1mV* 600/(300+600) or 0.67mV; for any live performance work, this is the same as 1mV. (Real lecturers and singers vary 10dB-20dB, so a little 4dB shift due to heavy loading vanishes in concert.) In loud performance in places people would want to listen, an SM57 can give 50mV-100mV (unloaded). If you put it places you would not put an ear, like against a Fender or close to drums, maybe 300mV. Close-micing a Cobra tailpipe you might get Volts; the '57 should be linear up there, until the hot exhaust melts the plastic diaphragm.

Kev

Matching Transistors for Valley People MicPre
« Reply #67 on: May 08, 2005, 03:03:31 AM »
Quote from: "PRR"
... the '57 should be linear up there, until the hot exhaust melts the plastic diaphragm.

lol
 :green:
I've mic a lot of things with a 57 but I've never done the exhaust pipe.
I'm sure someone has but perhaps there is a mic out there that does handle heat better.
Kev
DIY Factory

CJ

Matching Transistors for Valley People MicPre
« Reply #68 on: May 08, 2005, 04:29:37 AM »
I think Keef knows of a mic that will take an exaust manifold alright.

Thanks for all the help Paul! I need it. I am mixing two schematics, which is a bad idea for a proto attempt.

The original schemo I have Does use a 10K resistor under the trans-diode in the current source. Now that I have this thing running, I can play with values.

The breathing thing may have had to do with the amp being in oscilation.
Did not do it the second time I tried it.

I will try messing with R4,5. In the app notes version, they sit on top of the emitters in the current source. Maybe this is throwing things off.

Everybody, including myself, misses R10 tucked inside the lower box formed by Q1-Q2 collectors. Badly drawn schenmatic. I think these 2N3906 transistors would saturate if I used 10K base resistors. Plus, I think one aim of having the 100K bias resistors is to keep the input impedance high, although there are still the phantom resistors...

I will try to match the current thru the transistors to the original values and see where that puts my voltages. Not a lot of headroom needed in a first stage of a mic pre.

I dropped the 3.3M feedbacks to 1.78M. This helped opamp oscillation. I don't need the full 70db of gain anyway. I have a V76 if I want to listen to the grass growing. I put in a couple of  390pf across the feedback resistors. Hopefully this won't chop much high end off. It seems to help improve phase shift.

I have to study the effects of paralleling transistors. I do not know of any books that cover this array concept.

The offset voltage of 22 mv was actually an error on my part. I was measuring offset with respect to the pwr supply ground. Since I am using kind of a instrumentation amp setup on the opamps, I should have measured from plus to minus output, as this is in balanced out mode, not single ended. When I wen back and measured from output pin to output pin, I read about 1 mv (still high by todays standards) on the B/B 2134. But the BB drags 4 ma, so maybe thats why. An analog devices SSM2275 gave me 50 uV offset. (now that's more like it).

The BB is a fet front end, so I think it will work as good as an 072, but I have a socket so I will try the 072 also, as it is the original for this circuit.



Ok, I have this thing racked, so time to try it out.
It's as light as a feather (ext. pwr sup).
I am going to make another one for field recording. I bet Lomax would have loved to have something like this back in his day.






see ya!

cj
If I can't fix it, I can fix it so nobody else can!
Frank's Tube Page: www.mif.pg.gda.pl/homepages/frank/vs.html
Guitar Amps: http://bmamps.com/Tech_sch.html

CJ

Matching Transistors for Valley People MicPre
« Reply #69 on: May 09, 2005, 04:52:18 PM »
Alright, I took the VP rack home and plugged it in.

I had a couple of major problems at first. It was late, and I was hoping
to have a listen before retiring, but it was not to be. All I got was a distorted signal,
 howls and squealing, and a big dissapointment.
All standard for a first time project.

But! I had all day Sunday to troubleshoot my blues away.  Now I am
a happy camper! The VP sounds like nothing I have ever heard. It's
the missing link. It sit's right in between the slightly darkened
tube/transformer sound, the saturated transformer Langevin
AM-16 sound, and the crisp but warm API discrete op amp sound.
In short, it gives back exactly what you put into it.

What was wrong with it the first night? Two things.

1) The high gain balanced output op amp stage does not take
 kindly to directly driving a capacitor input headphone amp, or a transformer input
Pultec eq. It will oscilate.
(It mention's this right on the schematic, Doh!)The op amp heating
up will be a clue, along with hideous distortion and low output.
 But, put a 600:600 UTC on the output and it will be very happy.

2) A serious design flaw, in my opinion at least, is those two
100uf input caps. 100uf is too big to be anything but an electrolytic. Remember, the
voltage rating on those caps needs to be above
 48 vdc. (phantom pwr)
OK, so what's so bad about having lytics on the front end?

 Look at the bias voltage on Q1 and Q2. On the stock schematic this voltage sist slightly
above ground. The stock schematic shows 1.5 volts on the
emitter, and 1 volt on the base. Mine sits a tad higher, about
3 on the emitters, and about 2.4 volts on the base. This means
that the lytics are in effect, installed with their polarity reversed! That is not good. In order
to be correctly biased when
phantom is applied, the plus side has to be up front.. So with the
 phantom supplied turned off, you
have plus 2.4 volts sitting on the minus side of the caps.

A second problem is that lytics leak current. How much depends
on the type cap and how old it is. But even a few microamps is
enough to cause trouble. This leakage current, if high enough, will
upset the bias on the bases of the transistors. I know, because I
had a leaky lytic in there, and when I turned on the phantom
supply, the headphones almost blew up from the scraming pre amp.

So I changed the lytics to 10uf/63 volt polyesters, and that problem
 was solved. I do not know how much bass I lost doing this, but it does
 not appear to be a problem, as there was plenty coming thru the other side of the pre.

So, those two fixes, the output transformer added, and the lytics
changed. Now I could have a listen with both the SM 57 and AT4033.
This preamp will not make your 57 sound like a Neumann, as it adds
very little color to the sound. Use your Tefeunken gear if you want to spruce up the
 Shure for vocals. For recording instruments, the Shure
 will sound fantastic. You need a good vocal mic to get a good sound
out of the VP. The AT was alright, but I suspect a U-67 would work wonders with this
 pre amp. Unfortunately, I do not own a
good Neumann.

Another mod I would recomend for this project is a stepped gain
switch instead of a pot,  for two reasons. One, the pot makes that terrible DC
scratching noise when it is adjusted. Bill Y is going to try
 and fix his with a cap on the wiper I believe.
Another reason is that I tried a linear 1K and an Audio 2.5K and they
both did not work very well as far as resolution is concerned. When
you get to the top of the gain, things become real touchy. I think if a
pot were used, it would have to be a reverse log.

Also, this amp, like the API, seems to like to run at a certain gain for maximum fidelity. The
 low 20 db setting does not really do it justice. I noticed a lack of dynamics
 down there, but that may have been just because of the lower output. I will
have to do some more tests.  

At the higher settings, close to the max gain, I found the transistor
hiss to be a bit much. Lowering the gain a tad quickly reduces this noise. This may be due
 to the transistors I choose, or the fact that I am using 8 instead of 4, or a combination of both.

But dial this bad boy in to about 2/3 to 3/4 max gain and it is like
 heaven on earth. You will find yourself listening to it for hours.

More tweaking to experiment with, going to try different opamps, change some
 resitors here and ther, etc.

I tried using two output caps instead of the transformer, but the amp
 still would not function properly. I do not want to run a transformer, so
I will try a buffer op amp next.

Thanks to BYacey fro hounding me to build this for the last 50 years! Now I know
 why he was so persistant.

cj
 :guinness:
If I can't fix it, I can fix it so nobody else can!
Frank's Tube Page: www.mif.pg.gda.pl/homepages/frank/vs.html
Guitar Amps: http://bmamps.com/Tech_sch.html


SonsOfThunder

Matching Transistors for Valley People MicPre
« Reply #70 on: May 09, 2005, 07:12:08 PM »
Quote from: "PRR"
Relax. I had a date with Ella 20 years ago and she didn't show; odds are even lower now.

LOL!!

Quote
The TL072 is the best opamp here.

OPA2134 should work about the same as far as o/p offset and input current, in my experience.

Quote from: "CJ"
The high gain balanced output op amp stage does not take
kindly to directly driving a capacitor input headphone amp, or a transformer input

With which OA?  or all of 'em?

Quote
seems to like to run at a certain gain for maximum fidelity

Makes me wonder if adjusting the diff pair current might change the sweet spot??

Quote
But dial this bad boy in to about 2/3 to 3/4 max gain and it is like
heaven on earth. You will find yourself listening to it for hours.

 :thumb:    :thumb:    :thumb:
"The sow would rather have her ear than a purse." - PRR

PRR

Matching Transistors for Valley People MicPre
« Reply #71 on: May 10, 2005, 12:16:40 AM »
> I think if a pot were used, it would have to be a reverse log.

This type of gain-set network ALWAYS favors a reverse-audio.

For foolin-around, wire a normal audio so that at mid-rotation, you get about 1/10th the nominal resistance. Now turn the knob "backward". This is the same as a reverse-audio. But the backward action is too confusing to real-time studio work.

> output op amp stage does not take kindly to directly driving a capacitor input headphone amp

It should. Assuming you take just one of the two outputs into an unbalanced input. (If you connect headphone amp ground to one of the mike-amp outputs, it will be unhappy.)

> put a 600:600 UTC on the output and it will be very happy.

TL072 will not be very-happy driving 600 ohm windings to high level. SoT suggests another opamp.

> the lytics are in effect, installed with their polarity reversed!

The original plan was not Phantom-ed, was it? You have to think when you tack-on Phantom. And you can't call this a VP design flaw, if they didn't Phantom it.

In this case: yes, the only simple answer is non-polar input caps.

> I changed the lytics to 10uf/63 volt polyesters, and that problem was solved. I do not know how much bass I lost doing this

None. The load on the caps is >100K if you use your values, or over 20K if you use my values. Either way, input cap roll-off is subsonic.

What you got is an increase in low-frequency noise (random rumble). 10+10uFd is ~2Kohms at 20Hz, much higher than the nominal 200Ω source impedance. Base current noise will be high, at least on a dummy input load. My impression is that you may never find a room with so little rumble (due to HVAC, distant trucks, etc) that 10uFd caps add any electronic rumble to the acoustic rumble, but it is something to refine.

> lytics leak current.
> the pot makes that terrible DC scratching noise when it is adjusted.

The scratch is DC offset voltage. You obsessed over that, then used those unnecessarily high 100K base resistors. Input current (and cap leakage) is making several volts drop, so "small" Beta unbalance means large DC unbalance. And I don't think you can keep Beta balanced by matching lots of transistors: Beta is a pretty fishy parameter, not locked in Silicon like Vbe is. Dropping the input resistors to 10K will improve DC balance, reduce "scratch", and won't load-down Ella's mike or the input caps (if they are properly selected for low base/bass-noise).

> try and fix his with a cap on the wiper

Ugh! Fixing one problem with another: a big expensive un-clean electrolytic. I believe an original goal was to -eliminate- the awkward 2,200uFd NP cap on the gain-trim network (by hiding a DC-block between BJTs and opamps, where a small cap works fine). Find and fix the DC unbalance.

> close to the max gain, I found the transistor hiss to be a bit much.

With a mike, in a room? Open-circuit, or even terminated, sure the hiss will be gross at full-up. That's a heck of a lot of gain. With a mike in a room, I suspect room noise will dominate over amp noise. (If not, something is still wrong.)

PRR

Matching Transistors for Valley People MicPre
« Reply #72 on: May 10, 2005, 12:55:34 AM »
> Pretend it's Ella singing. No wait, she used to break wine glasses.

OK, it was just over 20 years ago, I was scheduled to do PA for Ella and at the last minute they told me she wasn't able to come, show cancelled.


The next year, 1986, she had big heart surgery, and also diagnosed with diabetes. Docs said she wouldn't sing again. Wrong. She did my town (without me), and Carnegie Hall in 1991 (her 26th concert there). Of course the diabetes did what it does: attacks the eyes and limbs. This forced removal of her legs, which reduced her performances. Died 1996.

Last album was 1991, so I doubt she will show in your studio. (If she does, get it on Memorex!!!)



Oh, it is widely known that the glass-breaking wasn't live, or Memorex, but a BB gun.

It seems there is no genuine case of an unaided human voice shattering a good glass, not even Caruso (according to his widow). Apparently the Memtek folks can shatter glasses by tapping it, getting a singer on the pitch, then cranking a speaker WAY up. Allegedly they can do this in sales demonstations. However someone who was there said the commercial was shot with a BB gun: more reliable.

bcarso

Matching Transistors for Valley People MicPre
« Reply #73 on: May 10, 2005, 01:38:35 AM »
One of my favorite old SNL skits when Ray Charles was the host: Garrett Morris as Ella, with enormous beanbag falsies, doing a "Mammorex" commercial with Charles.  The gag was that Ray's trademark dark glasses broke...

See also in this connection the sadly forgettable Dangerfield movie The Fourth Tenor for this gag taken to absurd extremes.

BYacey

    Where dogs wear thermal underwear, Alberta, Canada
  • Posts: 769
Matching Transistors for Valley People MicPre
« Reply #74 on: May 12, 2005, 03:41:01 PM »
With respect to the phantom caps, if you float the 6.8K resistors ie:( not tying them to ground when the phantom is off) there isn't any reference to power supply ground, or any other potential when you think about it. Therefore there shouldn't be any dc potential  across the caps. When phantom is switched on, they are correctly polarised. The only exception is if a mic has a ground reference internal, but most dynamics are floating with respect to pin 1.

All the voltages drawn on the schematic are my own measurements, and are just an approximation of what should be there. Far more important is the current through the transistors being equal and balanced.

I find that there isn't a lot of noise when varying the gain control, but it is noticeable. Tantalums (dare I say the T word) could be a good alternate to film  for the .68  coupling caps to the opamp.

Regarding the residual noise you mentioned with gain full up, I didn't experience this. With a 200 ohm resistor terminating the input, it is really quite low. Open input is a different story.
Bill Yacey
"Adjust R116 for least smoke"

CJ

Matching Transistors for Valley People MicPre
« Reply #75 on: May 12, 2005, 09:44:27 PM »
Thanks Bill. I will fool around with this circuirt some more. I thought about Tants, but 47 volts is usually the highest rating for those types which might short out with phantom.

I have not tweaked the CMR resistors, so maybe that's why I am getting hiss at the upper gain levels. It's not bad signal to noise wise, that is, with something feeding a mic, but with no mic signal, it's pretty loud.

I guess if lytics were letting a little current thru with phantom applied, the leakage would get shunted to ground thru the transistor bases and not cause any problems. The film caps sound fine, so I think I will leave them.\

What is the purpose of raising the minus supply a bit with this 360 ohm and 2.2 uf cap? Is it to put the transistors slightly above ground?
Thanks:



If I can't fix it, I can fix it so nobody else can!
Frank's Tube Page: www.mif.pg.gda.pl/homepages/frank/vs.html
Guitar Amps: http://bmamps.com/Tech_sch.html

BYacey

    Where dogs wear thermal underwear, Alberta, Canada
  • Posts: 769
Matching Transistors for Valley People MicPre
« Reply #76 on: May 13, 2005, 01:46:20 AM »
The 360 ohm and 2.2 mfd are just to isolate the rail a little for decoupling, it just gives a little bit cleaner DC voltage for the circuit. The positive rail  feeds the constant current source, so it isn't required here.  Same as power rail bypass caps for an opamp. You could do away with the series diode as well. I think it's there just for a safety precaution against someone wiring the factory modules with the wrong polarity. A smoke limiter.

The Tants I was refering to are the .68 mfd caps to the input of the opamp (s). I don't see a problem with 'lytics for phantom isolation. If it was good enough for Valley, it should be good enough for me. As I said, there shouldn't be any reverse polarity on these input caps if the phantom resistor network is left floating when phantom isn't applied.
Bill Yacey
"Adjust R116 for least smoke"

Matching Transistors for Valley People MicPre
« Reply #77 on: May 13, 2005, 09:01:44 AM »
Hi all!

Sorry for hi-jacking this thread but I have a couple of simple stupid newbie questions that really doesn't deserve their own thread.

Is the input stage of this amplifier (and other similar designs) really a differential amplifier? I mean, the stage doesn't have a common tail resistor from the emitters to the supply voltage. To me it looks more like two CE-stages with their outputs connected to the differential amplifier of the opamp.

What is the formula for calculating the gain of the input stage without feedback resistors? It doesn't make sense to use the gain formula for a diffamp (because of the missing tail resistor) and I don't know where to put the gainresistor value in the gain formula for the simple CE-stage.

Thank you for your help and sorry for the newbie questions. Look at it this way: the faster I learn (with your help) the sooner I'll start asking interesting questions.

/Stu

PRR

Matching Transistors for Valley People MicPre
« Reply #78 on: May 14, 2005, 01:12:53 AM »
> Is the input stage of this amplifier (and other similar designs) really a differential amplifier?

Yes.

> I mean, the stage doesn't have a common tail resistor from the emitters to the supply voltage.

So? That's just one kind of differential amplifier, sometimes called "long tail". It is an especially handy form, so we see it more than all others put together, at least when we are not doing something fancy.

> To me it looks more like two CE-stages with their outputs connected to the differential amplifier of the opamp.

That's what it is. Except what is the "Common" in "CE"? The emitters are not referenced to ground, except through high-value resistors or current sources. They are referenced to each other, through a small resistor (here adjustable as low as 10Ω)

This form has two "tail" resistors/sources. That's just a detail: we could instead use one tail resistor/source connected to a center-tap of the between-Emitters resistor. You sometimes see it that way for fixed-gain use. Here we have an adjustable between-Emitters gain-set resistor, so there is no natural "center" to connect one tail-source to. We could use a dual-gang pot, but a second fixed tail resistor is cheaper, and a second current-source is today cheaper than a dual pot.

Generally the presence of (one or two) tail-paths does not affect the signal gain. The between-Emitters resistor is much smaller than the tail-resistors so it dominates the gain equation. (But if we trim down to low gain, the between-Emitters resistor becomes about the size of the tail-resistors; this is one reason to pay for current-sources.)

> It doesn't make sense to use the gain formula for a diffamp (because of the missing tail resistor)

In the 2-transistor long-tail diff-amp, the tail resistor has no effect on the differential gain. The differential gain starts with the resistance between the emitters, either actual resistances or 1/Gm of the transistors.

> What is the formula for calculating the gain of the input stage without feedback resistors?

Apply a signal between the two input Bases. For a moment, assume the transistors work at large Gm, and have a resistor between the emitters. Transistor action forces the input voltage across the emitter resistor. If the resistor between the emitters is 100Ω, and we apply 1mV. we get 10mV/100Ω= 0.1mA emitter current. Essentially the same current flows in the collectors. If the collectors are loaded in 2K, each shows 2K*0.1mA or 200mV. The gain is 200/10 or 20. That's to one collector: in this case we have two such loads so the gain between both bases and both collectors is double, or 40.

We get the same answer if we look at 2K load and 100Ω in the emitters: 2K/100= 20. Or (2*2K)/100= 40 for differential output.

In this amp the emitter resistor goes as low as (I think) 10Ω. At first glance we think the gain is then 2K/10= 200. In fact if the transistors are biased at 3mA, then 1/Gm is about 10Ω each. From Base to Emitter through Resistor to Emitter to other Base is 10+10+10Ω, and only 1/3rd of the input voltage actually appears on the emitters' resistor. Gain (one side) is more like 60 than the rough-estimate 200.

The opamps toss in another gain of more than 1,000 over most of the audio band.

We have two 10K feedback resistors from the opamps to that between-emitters resistor. If wired correctly, this feedback cancels some of the current in the between-emitters resistor. With infinite gain inside the loop, the differential in-out gain is restricted to (2*10K)/Rg, where Rg is that emitters resistor. If it is 10Ω, we expect a gain of 2,010. Check: the transistors give gain of around 60 with 10Ω Rg, opamps add 1,000, 60*1000= 60,000. This is 30 times greater than the feedback-set gain of 2,010. Actual gain will be lower by about 1/30 or 3%, so we really get gain around 1,950 when opamp gain is 1,000. Opamp gain may rise to 100,000 in the bass, 60*100,000= 6,000,000, about 3,000 times higher than the feedback-set gain of 2,010. Error is 1/3000, real gain will be low by about 0.03%, so it will be ~2,009. The droop from 2,009 in the bass to 1,950 at high audio frequencies is 1950/2009 or 0.97, a 0.26dB error in frequency response.

Actually, gain of TL072 hits 1,000 at around 6KHz, is falling past 300 at 20KHz. So at maximum gain we will have about 1dB droop at 20KHz, -3dB around 40KHz. That ain't bad, considering it will be less when we increase Rg. For very large Rg (low gain) it actually tends to the opamp GBW times the ratio of the 10K feedback resistors to the 2K collector resistors. These force the opamp to work at gain around 5, so we will be 3dB down at just past 1MHz. A designer may add some caps to reduce this a bit (we don't especially want MHz response, and it will vary with opamp parameters), though much of the charm of this general idea is the fact that gain can be varied over a wide range and stay feedback-stable without compensation caps.

> simple stupid newbie questions

There are NO stupid questions. (Lots of stupid answers available....)

> questions that really doesn't deserve their own thread.

I don't know what you have to do to "deserve a thread" around here. This isn't CompuServe 1989, when thread-heads ate precious "slots" and forum managers budgeted their 450 slots. I agree there are some frivolous threads here, but I think you've raised a good question.

Matching Transistors for Valley People MicPre
« Reply #79 on: May 14, 2005, 05:13:32 AM »
Quote
I think you've raised a good question


Thank you, PRR. I think you gave an excellent answer! I have to think about this a couple of minutes or hours. Then I will ask another question or two. This time in their own thread.

Thanks for your help! This is an amazing forum!

Stu


 

Related Topics

  Subject / Started by Replies Last post
31 Replies
13004 Views
Last post May 30, 2006, 05:24:34 PM
by CJ
4 Replies
2348 Views
Last post September 21, 2009, 10:43:28 AM
by stereokillah
11 Replies
8091 Views
Last post December 16, 2009, 10:13:27 PM
by CJ
0 Replies
158 Views
Last post April 26, 2018, 10:28:55 AM
by Jitter