Playing with discrete class-A opamps on the simulator

[PRR]

> My first discrete project is going to be a moving-coil phono cartridge - to - line - level preamp

Sorry I missed this remark.

Phono preamps are tough; IMHO even tougher than good mike amps. MC phono has extra challenge. I don't think it is a good starter design.

The schematic you posted will work. It seems over-complicated for the job.

Your Stage_1 has the doubled-outputs, but the real load is R13 1K which can't possibly need over 10mA-13mA peak (and much less over most of the audio power-band).

OK, you also have to drive your 150+1Ω feedback network. Why 1Ω? Because your MC needle output is low and you want a low noise voltage, of course. But then you have R27 R28 220Ω each. Your input noise voltage is like a 440Ω resistor, plus the few-Ω needle coil, plus the 1Ω NFB resistor, plus the emitter dynamic resistances of about 300Ω each. Around 1,000Ω total noise resistance. Only a few ohms of that is source (good) resistance, the rest is just thermal noise.

So there is no reason for the NFB net to be 1Ω. 100Ω would be fine, much lower than the total 1K noise resistance, making R7 say 15K, so Stage_1 output only needs 15mA peaks, 8mA idle current, one output pair should be plenty.

OTOH, your source resistance (MC phono) is few-Ω, so can generate an electrical signal with thermal noise voltage like a few Ω, much lower noise voltage than your 1KΩ amplifier input noise. Transformerless MC phono inputs usually use HIGH bias currents to get low transistor dynamic resistance. The transistor parasitic resistance limits how low it can get, but several mA of bias current in a big-junction transistor is about as good as it gets without exotic parts. So for lowest noise with MC phono input, you should consider 1mA-2mA in each input device, with preferably zero emitter resistance, or at least less than few-Ω. However the output stage does not need much more than that to drive the RIAA EQ network. You don't need a many-stage amplifier.

When working at these noise resistance levels, diff-pair inputs double your noise resistance. They are a necessary evil in balanced-mike work, but not in home phono work.

And also: you do not need a "zero ohm" output, you need 1,000Ω into your RIAA network. For general mix-n-match hook-up, zero-Ω outputs are nice. But when the job calls for a specific load or output impedance, building that into the design tends to be simpler than aiming for some "ideal" and then padding-out to the specified impedance.

Also, you don't need a zero-volt DC output. You need C2 to set your 20Hz RIAA pole, and RIAA is actually zero response at zero Hz. As long as you need a cap, there is no point in great DC accuracy.

Look at another fact. Up through 2KHz, your EQ network has a loss of no more than 8:1. Your output stage has gain of 150. So the level out of Stage_1 can not be more than 1/20th of the level out of Stage_2. Stage_1 can be a MUCH lower-level design. I'm forgetting the fact that the RIAA loss will be 80:1 at 20KHz, 800:1 at 200KHz, etc. Music power declines above 500Hz, but some material has ample 6KHz power. Still Stage_1 runs at no more than 1/6th the output of Stage_2. It could run on +/-3 rails! Or it could be designed very inefficiently and simply.

This should work for Stage_1 and be lower noise:

I've loosely based this on National.com's AN222 fig 4, which is a low-noise MC preamp.

The output impedance IS 1K, so the RIAA network series resistance is already there.

The gain is almost 1,000/6 or 150, close-enuff.

The noise resistance is about 15Ω on paper, about 50Ω with available devices (parallel LM394; selected 4401 may be similar), much lower than your 1K noise-resistance design. Over 10dB lower noise floor for few-Ω sources. I've omitted an input cap, a nasty source of subsonic noise. The 4,700uFd cap is in the same place as far as noise, but has to be so big that its noise is a non-issue. (It also has some DC bias to keep an electrolytic formed.)

You may need an extra ripple filter on the + rail, since the output is really referenced to the rail not to ground. 100Ω and 4,700uFd ought to clean it good. You can do the same on the - rail but it has over 40dB of rejection already.

Output swing is over 3V peak. Around 6KHz the RIAA is 1/24, the output 150/1, so the output will be (trying to) swing 18V peaks before this stage clips. At typical MC cartridge sensitivity and recorded levels, this stage's output will be 1V or less.

And you might use this scheme for Stage_2 too. In the top of the audio band, the source impedance into Stage_2 is falling from 120Ω. Noise is mostly determined by Stage_1 noise, but in the top octave a 1K noise resistance amp as Stage_2 may not be masked by Stage_1 noise. If you do that, raise the 270Ω resistor to maybe 1K. The peak swing of 3V is more than enough to toot a line-level input. CD player ab-max output is 2.8V peak. Depending on your cartridge sensitivity, it might try to make more than that. Make R9+R11 a 5K pot (with 50K in parallel to preserve the computed EQ), play your hottest LP, and set level similar to a hot CD or a bit less.

 

[magicchord]

Thanks, PRR. I can't see your schematic right now, but I can't see your avatar either so I guess your web host must be down right now. Really, this opamp is a project that can have many applications. The MC phono preamp is just something I can use right now

I've made more changes to the op amp since we last spoke. I'll post a new schematic along with some questions soon. (EDIT: here's the circuit.)

The breadboarded version seems to be really quiet even at a noise gain of 470 (about 53dB). Very clean, too. Thermal drift of the output transistors seems to be controlled well enough that it won't run away with itself.

How much input bias current do you think I could pass through a MC cartridge safely?

 

[magicchord]

Ah, I see it now, and yes you could make a 2-stage MC preamp out of it. Boy, them emitter bypass capacitors are sure big, but they don't have a lot of DC on them so I guess physically they'd be reasonable size.

Going back to my opamp... :grin:

The change I just made in this version to the output current source does work, but now when I turn the power supply off, instead of slowly "dying", the opamp shuts off with a +4v "thump". I guess I'll go back to the way I had it before.

 

[magicchord]

This is a cool circuit, PRR. I see you greatly simplified it based on the one that was in AN222.

Unable to leave anything alone, I realized that one could multiply this circuit by 2 and add current sources, and we'd have a nice, simple balanced instrumentation amplifier with good CMRR. I breadboarded this last night and it works well:

Just as an exercise, I'm going to hook an SM57 up to it and see how it does. :grin:

(EDIT: Sounds good.)

 

[PRR]

> multiply this circuit by 2 and add current sources, and we'd have a nice, simple balanced instrumentation amplifier with good CMRR.

Yeah; in fact I posted something similar but different feedback. Everybody wanted more iron, or FETs, or something.

Drawbacks:

Give it an unbalanced signal and it makes an unbalanced output. Many condenser mikes don't drive both pins: they are balanced impedance but unbalanced signal. At high gain this is unimportant. If you go to low gain (increase R3) the unbalance is significant. At the extreme: if R3=∞, your mike drives one pin, your unbalanced load takes the other pin, nothing comes out!

The CE-CE pair has a LOT of gain and can tend to be unstable up around 10MHz. 220pFd collector-base on the second stage helps, though it gives you classic slew rate effects. You may be quite stable at high gain, trouble tends to be at low gains.

Your current-sources improve CMRR but affect the noise performance. Since you probably don't need -15V input swings, raise the bias voltage and the value of R4 R8 to reduce the gain R3/R4 that reflects Q7 noise into Q1 Q2 "-" input (and same other side). It may be negligible for gain of 1:60, but is unimpressive for gains lower than 1:10 (a gain we sometimes want for hot condensers).

Bias current in Q1 Q3 might want to be a bit higher for mike impedance; but depends a lot on the transistors. In fact 680Ω does seem to be popular for mike-amps using 4401 and a Vbe load on the first stage. so 1K is fine.

Output swing is unspectacular for a 30V supply. I worked it all under one 48V rail for a little more level and one supply (cheaper to beef-up a Phantom supply than to add a dual supply).

 

[adrianh]

magicchord
Sorry to hijack this thread.
But Prr LM394 circuit got me going.
someone bleed all over this schematic.
http://www.mastertraxstudio.com/tech_talk/LM394/prr_m_pre_.gif
Mic pre with 1:1 input iron and maybe 1:2 output iron.
Powered from +48 volt high current supply.
heatsinks on transistors no problem.
yes I know I may have too much gain.
Prr what is the simple gain formula for one stage?

And the circuit will overload easy.
Just needed a simple 60 dB gain block.
And this looked like a nice place to chime in.

 

[adrianh]

PRR;
Why did you in your circuit take the output off the emitter?
In the AN222 the output is off the 2N4250's collector?
http://www.national.com/an/AN/AN-222.pdf

 

[PRR]

> Why did you in your circuit take the output off the emitter?

I can't remember if I took my pill an hour ago..... you ask about a tossed-off sketch from a couple weeks ago?

There are many ways to skin cats.

> bleed all over this schematic.

No Base-bias(es). As drawn, nothing will happen. The battery won't even go flat.

Output will not drive 150Ω as you seem to think.

 

[magicchord]

[quote author="adrianh"]PRR;
Why did you in your circuit take the output off the emitter?
In the AN222 the output is off the 2N4250's collector?
http://www.national.com/an/AN/AN-222.pdf[/quote]

As I recall, we wanted to specify an output impedance of 1k ohms to drive a RIAA network. PRR's circuit was a nice easy way to do that.

 

[magicchord]

After reading Douglas Self, I thought I'd try doing over the opamp to put some of his ideas into use, and also change the output to a class B circuit:

I haven't breadboarded this yet, but it looks like it'll be unity-gain stable with 470p as CDOM, and still have about 16v/us slew rate. The voltage divider on the output of the emitter follower is an attempt to correct for inherent offset errors. There a better way to do this?

 

[bcarso]

I tend to prefer Selfless designs myself but...

Your d.c. offset errors at that kind of gigantic d.c. gain are going to be as much or more affected by the mismatch of the 4403's in the input stage---adjustment via the voltage divider in the emitter of Q12 is a simulation fool's paradise I am afraid. BTW at those closed-loop gains I suspect you don't need emitter R's at all, and they will add to errors especially if they are mismatched 5% parts as the values suggest. They also will add significantly to your noise, which ought to be set by the thermal noise in Q7 and Q8's rbb', pretty low for 4403's but not nearly as low as some parts.

A very large C in the ground leg of the 4.7 ohm is one way out. But if you really needed the response down to d.c. then a better-balanced and matched input stage is called for. I would at least match some 03's, or try to find some good PNP duals (sadly the 2SA1349 is discontinued---you could parallel some of them [not exactly directly though] with a little additional work and get the noise down). Linear Integrated Systems or THAT Corp. might have some parts that would work.

To further clean up the input stage: use a modified Wilson current mirror with three Q's and some ballasting R's in two of the emitters to reduce noise. Now your unbalancing force will come from the loading by the Q12 stage; you don't know what its base current is exactly, but you can compensate for it in a few different places. Or, use a small MOSFET like a 2N7000---you have room to accomodate its higher gate-source voltage---and help it work better by making R9 smaller. And you might want to put a protection resistor in the drain to the +rail (or for that matter in the collector of the 4401 as it stands) to prevent big base currents in Q3 under overload conditions, although the '01 will probably survive with the 100 ohms there.

Provide a d.c. trim adjust somewhere though, and if you want to avoid future agonies replace any trimpot with fixed r's after value determination.

It is unfortunate that for the app (MC cartridge pre) the noise voltage is root 2 more than one 4403 due to the differential configuration. It is still pretty low at those collector currents (the good Moto, now On Semi parts if they make them the same way get to less than 1nV/root Hz midband). But your MC cartridge may be in the region of 10 ohms, with a thermal noise density of 0.4nV/root Hz. Also, at those collector currents the shot noise in the base current of order 25uA is just beginning to creep up into noticeable territory for highish MC cartridge Z's. And when you look at the 1/f component in the base current noise it gets more worrisome, especially when you apply the RIAA compensation.


Brad

 

[magicchord]

Simulation fool's paradise, eh? Probably right.
I've been matching 4403s up to now; not difficult though time consuming.
I don't think I'd use this opamp for a MC preamp. This is more of an exercise to make a simple general purpose opamp for lower gains than I show on the schemo.
I think from now on I'll wait to post these things till after I've breadboarded them