Preamp Project

Hi all,

in a short I will build the following MIC preamp as on part of a larger project including A/D conversion along with ADAT interface and so on.

But lets begin with the input stage first.
It is nothing new but if I can make it better than it will be worth to do it.

The schematic can be found here http://groupdiy.twin-x.com/albums/userpics/10057/preamp_v09.pdf

To lower noise would it make sense to add a second transistor in parallel to Q1 and Q2?

I also saw some schematic on which the transistors where directly connected to the input of the OPAMP. Is there a benefit doing so. In my opinion this will disturb the a little bit the input stage, or?

Suggestions welcome

DW

designmaster

Is bold name for the questions you ask.

Look at the valley thread. The mackie schematics . Some B schematics the project at ESP? #66. PRRs discreate design thread the amek 2500 is maybe what you are looking for for multabe transistors.

The look close at the green, FF schematic etc.

There are more threads here with the answers you seek

Just off the top, are you sure the input transformer is oriented the right way? I don't see the point of it connected in this manner, or is this meant to be a choke for HF rejection ?

@BYacey
it is not a transformer it's a common mode choke. May the symbol I have use is a little bit misleading.

OK, fair enough. The design looks reminiscent of a typical Mackie front end.

a. If you use a low offset JFET opamp (OPA134, OPA604, TL074, etc.) you don't need the C2, C3, C29 and C30 capactirors.
b. If you keep the caps as shown this circuit may not work due to missing resistors on U1. Take another look at the ESP #66 circuit and note what is missing.

Why are there Resitors missing?
Yes on the ESP #66 there are but there is no need for in my point of view. Look at the soun.cr.ft Series mixer consoles they don't have it also on the ADA8000 of B. they are missing.

Thanks for the hint of your point a. But than I have to protect the inputs. On the datasheet it is mentiones input voltage => Vsupply - 0.7V

OK. Can you calculate the gain of your amplifier as drawn? Please let us know how.

How do capacitors protect the inputs? They let any AC pass through. As far as I understand the capacitors are used to DC decouple the circuits for low offset at the output of the preamp. Also the gain of U1 can be set very clearly independent of the resistors used in the discrete front end.

I would just drop those caps and build the thing.

Cheers,
Tamas

Well caps do not protect, that's clear.

Assuming Rgain is set to 1k (R21) the gain is 22.5 (simulation result)

Gain calculation is a little complex. But I will come back to you.

Hi tamas,
If I am not totally wrong:
Gain = 2*((R15+R20)/(1/(1/(R9+R10)+1/Rgain)))

Q1 Q2 are drawn upside down: Emitters must be on top pointing down.

Direct coupling to the op-amp need not disturb it, but does force many many changes, typically leading to a 2- or 3-opamp design.

For these parts, paralleling Q1 Q2 will not produce any big reduction in noise for gain of 1,000, and no effect for gains of 200 or less.

Thanks PRR :thumb: , I think I must have been blind.

Direct coupling to the op-amp need not disturb it, but does force many many changes, typically leading to a 2- or 3-opamp design.

Click to expand...

But what do you mean with 2 or 3 opamp design?
Like on the VLZ series of Mac. one opamp per Transistor pair? Or in the meaning to get back a balanced output?

The plan is currently to have one balanced output for external connections and one with an OPA1632 to feed directly the input of the AD converter PCM4202).

But what do you mean with 2 or 3 opamp design?

Click to expand...

I assume he is referring to the Cohen design (one opamp per transistor); check JFETMP1.PDF and 9098micpre-1.jpg for a first reference. The later is claimed to be designed by a very well known person.

I would make the input coupling caps bigger (say 220 uF), as this will reduce low-frequency noise. Noise figures below 5 dB are a rather academic thing, but this change is cheap, so if you sleep better as I would, a thing to consider. On the other hand, the bigger cap has higher leakage, so watch out.

Samuel

> Can you calculate the gain of your amplifier as drawn?

At a glance: 2*4K7/Rgain

Or really +1 higher than that (if you calculated 100, it is really 101).

But Rgain is tricky. Obviously it is the pot R21 (whatever it is set for) plus the anti-infinity stopper R2. Oh, and all that in parallel with R9+R10. In the mid-range, when R21 is around 500Ω, these 5Ω and 5KΩ fudges can be neglected for a hand-turned knob. At max gain, we have 2*4K7/4R7 or 2,000. At min gain we have about 2*4K7/(4K7||5K) or about 4, plus 1 is 5.

max: 2,000
mid: 21
min: 5

OK, I have omitted a fudge at the highest gain. We need to add the dynamic emitter resistances of Q1 Q2. In some other designs, overall feedback absorbs this fudge; here it does not.

Q1 Q2 run at about 1.2mA, so their raw emitter impedance is about 24 ohms. This 24+24= 50 ohms is very significant compared to the 5Ω Rgain, so it looks like maximum gain is more like (2*4K7)/(50+5)= 170.

That's why, when no feedback from opamp output to input emitters, we usually add Q3 Q4. This compound will lever the Q1 Q2 emitters by the effective current gain of Q3 Q4. Total current in R9 R10 is 6mA, 1.2mA to Q1 Q2, leaves 4.8mA in Q3 or Q4. Their emitter impedance is around 4Ω, their Beta say 120, dynamic base impedance about 500Ω. So Q1's dynamic collector current splits between R16 and Q3 about equally. The base current multiplies by Beta and is injected back to Q1 emitter. The effect is roughly to reduce Q1's dynamic emitter resistance about 60 times, from 24Ω to about 0.5Ω. This is now "small" compared to R2, though not dead-small: the max-gain falls from 2,000 to around 1,600. Distortion is better with Q3 Q4 because Q1 Q2 work semi-constant-current.

That assumes a "good" value for R16 R17. With just Q1 Q2, the transfer curve is a classic exponential V/I curve (actually the difference of two opposed curves because of push-pull). With Q3 Q4, the transfer curve is sharply bent and S-shape. I know I had a plot of that.... seems the dog ate it. If you get mid-bend on the S, it is fairly linear and MUCH higher gain (lower error) than a simple transistor, but you can be very far off too.

Thanks Samuel, good idea.

Hi PRR,
my brain has to work and is currently steaming. I think I need a little bit to follow your comments.

Is there a major benefit in using one opamp per transistor, compared to my design.

An updated version of the schematic is available:
http://groupdiy.twin-x.com/albums/userpics/10057/preamp_v091.pdf

Hi PRR,

That assumes a "good" value for R16 R17. With just Q1 Q2, the transfer curve is a classic exponential V/I curve (actually the difference of two opposed curves because of push-pull). With Q3 Q4, the transfer curve is sharply bent and S-shape.

Click to expand...

just to make sure that I got you right. With V/I curve you mean Vce over Ic, am I correct?

Is there a major benefit in using one opamp per transistor, compared to my design.

Click to expand...

Get the G.J.Cohen paper Double Balanced Microphone Preamplifier, AES Preprint 2106. It explains the design very well and also looks into the problems with conventiolnal diffpair-to-opamp topologies.

www.aes.org

It will cost you a couple of dollars but it's a good read. PM me if you for some reason can't buy it from AES.

/Anders

By the way,

I'm building a preamp with a cohen-style input stage and a simple diffamp configured opamp on the output.

I have an oscillation problem in the input stage. Has anyone build something like this (SSL 9k preamp, AMEK 9098 for example) and found a great solution that they want to share?

/Anders

What opamps are you using? Can you share the schemo? Do you use small caps around the opamps for local feedback?

Samuel

No schemo on computer. I basically built it from the AES paper.

I use the NE5532 (also tried an LM833 with slightly better results) and yes, I have 100pF from op out to -in. It's not the kind of crazy oscillation that would occur without the caps (yep, tried that too). It's more like too much high frequency noise. The frequency is in the area of a couple of MHz. I suspect it can be a bad PCB layout. The path between emitters are quite long.

/Anders