Mic Preamp -- Is This a Silly Idea?

[chris319]

your DA converter which won't perform well at low levels. You could build an output PAD for the converter so that you can run it at high levels even for high gain circuits.

on paper these instrumentation amp designs will work at 80dB or whatever, but does it sound good?

The parts for a 70 dB pad are on the way from DigiKey and I will be re-testing my preamp to see how it performs at mic level padded down from line level, rather than with the gain reduced through the D/A.

The use of a 1:10 transformer is not lost on me for additional low-noise gain, but they are not cheap compared to an IC.

if you are amplifying speech for internet chat then perhaps I am off-base, but if you are trying to record a sitar with a coles 4038 and have it sound as good as possible, then that is what Im talking about.

Perhaps not quite up to the sitar/Coles model but close.

What do people think of this? It is the successor to the LM833. There is a mic pre on page 29 but I can't grok their gain calculation:

http://cache.national.com/ds/LM/LM4562.pdf

 

[JohnRoberts]

The 4562 is a very good opamp but at 2.7nV/rt Hz input noise it will be a few dB noisier than a good discrete bipolar front end, or transformer design.

Their gain equation looks correct as printed and is nominally 40 dB.

JR

 

[chris319]

Their gain equation looks correct as printed and is nominally 40 dB.

John, how did you get that? I worked the equation given:

V0 = (1 + (2R2/R1)) * (R4/R3)

and did not come up with V0 = 101(V2 - V1), or 100 (40 dB). I'm doing something wrong.

 

[JohnRoberts]

[quote author="chris319"]

Their gain equation looks correct as printed and is nominally 40 dB.

John, how did you get that? I worked the equation given:

V0 = (1 + (2R2/R1)) * (R4/R3)

and did not come up with V0 = 101(V2 - V1), or 100 (40 dB). I'm doing something wrong.[/quote]

This is from memory because I don't feel like downloading the file again and going to the schematic but from memory it was (1+20k/200) x1. I did the dB in my head, 100x is 40 dB so 101x is 40dB + a pinch...

JR

 

[PRR]

> I'm doing something wrong.

Back when I flunked EE, it was required to show your calculations. So don't say "did not come up with" and expect any help. How can we possibly know where you went wrong?

This one is worth knowing from Basic Inspection, because 97% of all inputs are wired this way.

First: the second stage is almost always wired Unity Gain. So ignore it.

Next: while (their) R2 and R5 kinda "must" be equal, it is useful to write both.

Recall that a happy op-amp has zero voltage between its inputs.

What makes this one happy?

Apply an input to the two external pins. The voltage on R1 200R "must" be the same. How can R2 have voltage? Current flowing to it from R2 R5. Why is there current? Because the output nodes of the two opamps will (we hope) assume a proper voltage.

Opamp inputs take zero current. (Not dead-zero, but usually utterly negligible for gain computation.)

Therefore R2, R2, and R5 all have the same current.

Assume we put 1V in. Assume the amp gets happy and 1V appears across R2 200R. This is 1V/200= 5mA. This must flow in R1=10K and R5=10K. 5mA in 10K is 50V. So the differential output must be 50V+1V+50V, or 101V. The circuit gain is 101V/1V= 101.

Another point to note. R2 200R has thermal noise similar to a 200 ohm dynamic mike. Even if the chip could be dead-quiet, noise figure is 3dB. If you want even higher gains, you do not {EDIT}increase R2 R5, you decrease R1. And if you want huge dynamic range, you decrease R2 R5 to 5K or maybe 2K (limited by how much strain the chip will stand). It seems reasonable, for low-noise design, to take R1 down to more like 5 ohms. However pots have large and unpredictiable wiper and hop-on resistances. Some of the glitching you hear near max gain is pot imperfection. Also, as shown, the DC gain is equal to audio gain. Using old rule of thumb, a 5mV DC input error is a whopping 505mV error at the output. A cap in series with R1 fixes this, but it should be bipolar, and for 20Hz at 5 ohms it must be like 2,000uFd. And now between pot and giant cap, the stray inductance in the "R1 arm" may bite you at the top of the audio band.

 

[chris319]

Shoot, I was reading the schematic wrong!

R1 is 200 ohms, not 10K like the other resistors.

That's what I get for trying to read schematics past 3 am :shock:

Just monkeying around:

R1 = 200

R2 = 50000

R3 = 2500

R4 = 50000

Gain = 80 dB

I'm no expert so there are likely problems with this. But if each stage could deliver 40 dB of gain ...

 

[chris319]

Here is my amateur effort at a two-stage preamp. I used the (free) TINA-TI simulator, hence the INA163 which should be substituted with a THAT 1510 running at 50 dB (built and working). The OPA627 I'm not so sure about, with that 300k resistor setting the gain to almost 30 dB. The OPA627 has a slew rate of 50 V/us to the NE5534's 13.

 

[Samuel Groner]

OK, not bad, not bad! Will probably work as shown but there are some things we could optimise:

• R3 serves no purpose--get rid of it
• to avoid amplifying the output offset of the frontend you want AC coupling (i.e. a 1 uF cap between U1 output and U2 +in as well as a 1 M resistor from ground to +in)
• the feedback network impedance of U2 is too high; scale it down to 10k/330 ohm
• as you seem to be impressed by large slew-rate numbers you could replace U2 with an OPA637
• there's no input RFI protection--check the THAT1510 datasheet for some suggestions
• you need PSU bypassing for the ICs--a 100 nF from each supply to ground for each IC
• to decouple capacitive loads from U2 output you want a small series resistor at the output--use 47 ohm

Samuel

 

[chris319]

OK, not bad, not bad!

Thanks Samuel, John and everyone who has chipped in.

there's no input RFI protection--check the THAT1510 datasheet for some suggestions

you need PSU bypassing for the ICs--a 100 nF from each supply to ground for each IC

to decouple capacitive loads from U2 output you want a small series resistor at the output--use 47 ohm

This schematic is a little misleading because it doesn't show the exact THAT 1510 side which already has the RFI and PSU bypass caps. There is also a 47-ohm series resistor on the output.

R3 serves no purpose--get rid of it

to avoid amplifying the output offset of the frontend you want AC coupling (i.e. a 1 uF cap between U1 output and U2 +in as well as a 1 M resistor from ground to +in)

OK.

the feedback network impedance of U2 is too high; scale it down to 10k/330 ohm

You mean 10k for R2 and 330 ohms for R1?

as you seem to be impressed by large slew-rate numbers you could replace U2 with an OPA637

Or make the whole thing with two OPA637s. :grin: I will have to study the data sheets. What do you think?

My original THAT 1512 preamp has worked great using two 9-volt batteries, but I'm thinking about using four (+/- 18V) for any new design.

 

[chris319]

OPA637 vs. THAT 1510

THD+N

OPA637: 0.00003%

THAT 1510: 0.0005% @ 0 dB, 0.005% @ 60 dB

SLEW RATE

OPA637: 135V/uS

THAT 1510: 19V/uS

 

[chris319]

If you want even higher gains, you do not increase R1 R5, you decrease R2.

I don't follow you. According to the gain equation, if R2 is decreased the gain will decrease.

Here is another interesting design consideration:

LM4562 apx. $5 US

OPA637 apx. $18 US

 

[JohnRoberts]

[quote author="chris319"]

If you want even higher gains, you do not increase R1 R5, you decrease R2.

I don't follow you. According to the gain equation, if R2 is decreased the gain will decrease.

Here is another interesting design consideration:

LM4562 apx. $5 US

OPA637 apx. $18 US[/quote]

He meant R1,, in low noise preamps the impedance of feedback resistors make noise too, so make R1 smaller instead of R2 larger when increasing gain to manage noise contribution from feedback resistors.

JR

 

[PRR]

This plan is clearly too complicated for you and me. You read R1=10K. I read R1 and typed R2.

If we understood it, as John does, such typos would be clear. Three resistors the same there gives gain of three, a "clue" that something is not as the mind thinks it to be. If I say gain increases and it clearly decreases, that is a "clue" that I'm befuddled.

I'll bow-out.

 

[chris319]

At any rate, here is what I have in mind for the LM4562. R5, 6 and 7 follow R2, 3 and 4 respectively:

R1 = 200 (or lower)

R2 = 10k

R3 = 200 (or 100)

R4 = 20k (or 10k)

The idea is to get 40 dB out of the second stage rather than unity.

LM4562 apx. $5 US

OPA637 apx. $18 US

It would take four OPA637 chips for a stereo, two-stage preamp. 4 x $18 = $72. A stereo preamp built with LM4562 chips would require 3 chips x $5 = $15.

 

[Samuel Groner]

It would take four OPA637 chips for a stereo, two-stage preamp. 4 x $18 = $72. A stereo preamp built with LM4562 chips would require 3 chips x $5 = $15.

I'm completely confused about this statement. Why do you need four OPA637 but just three LM4562? In any case you need the THAT1510 (or whatever) as frontend, neither of the opamps you mentioned are suitable as first stage (I think you're still missing the difference between opamp and instrumentation amplifier). And do not use too much gain for the second stage. 20 dB is perfect, more rather hurts (for several reasons).

It doesn't show the exact THAT 1510 side which already has the RFI and PSU bypass caps.

You mean it has RFI protection and bypass caps included in its little IC package? My dear, this is not so--you'll need to add them yourself.

Samuel

 

[chris319]

Why do you need four OPA637 but just three LM4562?

The LM4562 has two op amps per package and the OPA637 has one. A two-stage preamp would require two OPA637s per channel; a stereo pair would require four. Alternatively, a two-stage preamp would require 1 1/2 LM4562 chips or three for stereo.

You mean it has RFI protection and bypass caps included in its little IC package? My dear, this is not so--you'll need to add them yourself.

My already-built preamp has these caps in place externally, and they are in the app note schematic.

you need the THAT1510 (or whatever) as frontend, neither of the opamps you mentioned are suitable as first stage (I think you're still missing the difference between opamp and instrumentation amplifier). And do not use too much gain for the second stage. 20 dB is perfect, more rather hurts (for several reasons)

OK. The 1510 runs at (up to) 66 dB and the OPA637/LM4562 at 20 dB, for a total of 86 dB.

 

[chris319]

According to the TINA-TI simulator

this circuit turns 300 uV AC into 41.54 uV AC with the addition of C2 and R3. Without C2 and R3

Fixed that, thanks. Now 300uV AC becomes 6.61 V AC (+86.9 dB). Note the use of the much less expensive OPA134.

 

[jdbakker]

R3 is 1 milli-ohm. Most SPICEs (including TINA, I believe) use 'm' for milli and 'meg' for mega.

JDB
[and you wouldn't be the first to get bitten by this]

 

[chris319]

R3 is 1 milli-ohm. Most SPICEs (including TINA, I believe) use 'm' for milli and 'meg' for mega.

Who knew resistors would be so hard to deal with? :? Thanks, J.D. In this case all that was required was to use an upper-case "M".

As mentioned previously, the already-built THAT 1510 preamp will be used for U1, which already has the necessary capacitors, etc.

 

[jdbakker]

[quote author="chris319"]

R3 is 1 milli-ohm. Most SPICEs (including TINA, I believe) use 'm' for milli and 'meg' for mega.

Who knew resistors would be so hard to deal with? :? Thanks, J.D. In this case all that was required was to use an upper-case "M".[/quote]
Does TINA accept an upper-case "M" for megohm? Most SPICEs are case insensitive; have you checked the circuit sims as expected (ie ~80dB gain over the audio band)?

JDB.
[EDIT: Yes, it appears you have and it does]