Darlington spinoff: bcarso's preamp circuits

bcarso was kind enough to design a couple of really neat looking circuits that can be used as discrete gain blocks for a mic preamp, and its my pleasure to share these with everyone!

He'll be happy to explain these circuits, and he'll do it better than I ever could. One is a FET cascode, the other is FET/Bipolar, and I think they're followed by a discrete servo stage.

My mouth is watering just waiting to prototype these bad boys...

Here's the link:
http://www.geocities.com/mattlaguardia/index.htm

Since I wasn't successful in deep linking these docs, FET1 is cascode, FET2 is FET/bipolar.

Thanks so much for your contributions, bcarso!
:guinness:

looking good..let us know how they turn out.

Bcarso is the man. :thumb:

Dale had a schematic of the TLM103 that KIND of looks like the fet 1 schematic.

Who knows.. We all know that there are only a finite number of possible working combinations of parts. It's very likely that he (RE)discovered portions of the circuit or just took working portions of circuits and put them together.. we'll let him tell us which.

Thanks for hosting those featherpillow. It's nice to see one's circuits available.

At the moment there is no d.c. servo, since with the current source ("sink") for the input device the d.c. gain is about unity. I show a bipolar 'lytic for the feedback divider resistor so there is no worry about bias, and also because Jensen and others say the BP 'lytics have the lowest distortion. One might want to bypass it with a smaller film C as well.

The first and simpler schematic has a higher pinchoff voltage FET in cascode with the input part. The reasons are that the 3819 is protected from excess drain-gate voltage and consequent high gate leakage, and also its dissipation change with signal swing is small, which is more of an issue at low gains.

I predicated the design on what featherpillow had in available parts and power supplies, hence the 3904/3906/3819 and +/- 15V rails. In the second schematic the function of the cascode FET in the first is served by a 3904, and it gets a floating base bias V from the current source above. The lower current source (sink) is adjusted to compensate for that higher net current.

The 1pF at the input looks a bit silly, and you may well have way more than that in strays and the trafo distributed C. It fixes a little peaking at around 10MHz which is partly due to the bootstrapped cascode. The 100 ohm is there to suppress parasitic oscillations due to lead and wiring inductances, and may have to be adjusted for a real layout. Also the trafo may need a Zobel at the secondary for best transient response---I didn't show that on the schematic.

The topology beyond that is pretty much a current feedback amp like tkhalmi's gainbloke. There is some gain, more at overall high closed-loop gains, from input to upper drain/collector at the base of Q5, but the major voltage gain is at the high Z node, the collector of Q5/Q6/base of Q7.

The use of a resistive pullup for the Q7 emitter follower is actually an enhancing detail for reducing some even-order distortion by introducing some of the opposite polarity. It will be interesting to see how close the model predictions are to the reality of specific transistors.

Any layout or breadboard should be tight because the 3819 is really a high frequency device.

With the trafo presenting a 15k source Z with a 150 ohm mic, the noise contribution of the preamp especially at high gains should be negligible.

The output swing before clipping is about 7V rms, if the input FET's current source is adjusted for a d.c. output voltage a little below ground. The best way to adjust would be to select R5.

The 1% values are not that critical and indeed with the FET variability misleading. But they sort of suggest that metal film R's should be used so I left them in as 1%. Near-value 5% parts can probably be used without a problem.

[quote author="bcarso"]At the moment there is no d.c. servo, since with the current source ("sink") for the input device the d.c. gain is about unity. [/quote]
Ooooh, that is clever indeed!

Alright, it's time for me to wade slowly and awkwardly through the rivers of technical data here:

So this is actually a current feedback topology? As I understand it (in my own limited way), the general benefits of current feedback over voltage feedback are as follows:
Gain bandwidth impedance
Faster Slew rate
Lower Distortion
The trade-off, in general then, is feedback restriction--for example, one could not short the output to neg input to form a buffer. This would cause oscillations.


Is this the formula to calculate CFB gain: Vo/Vi= (-Rf/Ri)/(1+(Rf/Zvs)) ???
Or does this formula not apply to this particular circuit?

Is the input of the FET/Bipolar gain stage also a form of cascode? It sort of looks like it to me...

With lots of loop gain like this has for even maximum closed-loop gain, the formula for non-inverting voltage gain is just like an op amp's: 1 + Rf/Rdivider. I set it up for the first potentiometer to be the ~equivalent of the mic pad switch in the original schematic with the darlington etc. So, the gain goes from about 1 + 2.74k/1.1k or x 3.5, +10.86dB, to 1 + 2.74k/100, or x 28.4, +29.1dB.

I don't know if it is unity-gain stable, with Rdivider open, but it is probably not without a compensation C, maybe optimally series R-C, to ground at the high-gain internal node. But, that's not what it is for. A unity-gain buffer I would do a bit differently.

You are right--both designs are cascode front ends. The bipolar one requires the extra stuff because it needs base current and to be biased up some volts above the input FET source so the FET drain has some voltage to work with. With the 5638 as the cascode part, this voltage is built-in. Borbely uses that arrangement quite a bit for example.

I guess I should make a second stage design as well, although this one will work---I might want to see a bit better line-driving capability there, maybe a diamond quad output buffer. I haven't looked at how much C load this first stage will tolerate, but if it is problematic a few hundred ohms in series with the cable should fix it.

Also maybe the second half should provide a balanced output too, although you will be well out of induced line noise at these levels unless you are doing something very wrong.

> So this is actually a current feedback topology? As I understand it (in my own limited way)

Here's the thing. In voltage-feedback, changing the feedback resistor values does not change the open-loop gain. You can wire a 741 with 200+2K or 100K+1Meg feedback resistors, and the amp's open loop gain does not change (because a 741's input impedance is much-much more than 1Meg).

In current feedback, changing the feedback resistor values does change the opamp's open-loop gain. The feedback input pin is very low impedance, so open-loop gain is roughly proportional to 1/R where R is the feedback resistance(s). This means you can tailor the open-loop gain for your particular application.

This particular amp is in-between. The source of Q1 is maybe 200Ω(?). At high gain, the feedback impedance is 100||2K7 or about 2 times lower than Q1's impedance. Open-loop gain changes little as you turn R8 from zero to 100 ohms. At low gain, the feedback impedance is 1K1||2K7 or about 4 times higher than Q1's impedance, open-loop gain has dropped roughly in proportion to closed-loop gain. So for gains of about 10 down to 3, you can keep the same compensation capacitor (seems to be Q5's C-E parasitic) and get the same closed-loop bandwidth. For gains from 28 down to about 10, open-loop gain is farily constant so bandwidth is falling.

The usual problem is that if you have enough bandwidth at high closed-loop gain, it gets unstable at low closed-loop gain. This technique reduces open-loop bandwidth as closed-loop gain falls below ~10, so it stays stable and nearly constant closed-loop bandwidth.

"...you can keep the same compensation capacitor (seems to be Q5's C-E parasitic) and get the same closed-loop bandwidth."

Yep---actually the C-B capacitance, probably what you meant. It's a big miller effect there as well, with a voltage gain huge from the base to the collector of the PNP. If you take that away by another common-base stage things get unstable without adding a bunch of C at the high Z node. In the end it looked like leaving it as the device C was the best way to go, and also didn't reduce output swing as the CB stage did.

This is great, it is nice to see a design that uses such common and easy to get parts, parts that most any one with a decent junk box will have. I will be giveing these a try in a few days most likely. Thanks for the circuits and the info on how they work.

adam

Thanks. I'll be keen to hear of your results.

It's true that there are some nicer parts to use in each location, but the difference in performance is not going to be a factor of two even, I suspect, in any particular parameter except voltage noise, which is already low relative to a transformed 150 ohm input Z with the 1:10 trafo.

I would throw one together on groundplane/pad-per-hole material but remarkably I have some urgent paying work ;-) so it may be delayed a bit.

Make sure the 2N3819 has around 5mA or more Idss, or adjust the NPN current source emitter R upwards accordingly so that the FET gate doesn't get forward-biased. Other FETs in that series that are fine would be the 2N3823, 2N4416, 2N5485 or 2N5486, etc., all the old National Semi process 50 or Siliconix NH.

If there are >100MHz oscillations, try making the gate resistor a bit larger, and/or use a ferrite bead. Those transistors are characterized for 400MHz preamp duty and will oscillate at the drop of a hat, especially if there is much lead inductance. But it should still be possible to get things to work with through-hole parts if the basic signal paths are short. SMD would work a little better but is a royal pain for breadboarding.

Star-grounding of the feedback divider R d.c. block cap and trafo cold end is advised. Probably most everybody else can use a ground plane.

I was just thinking of another advantage to the cascode bootstrapped input that I had not mentioned, namely that the input capacitance is small (even slightly negative at some frequencies) and fairly constant in the audio band. Thus issues with distortion arising from source loading by a variable input C, as described by Jung and others in connection with op amps for audio, are much less significant.

I will hope to incorporate this aspect into the other preamp, although in that case the impedances are low enough to where it is less of a consideration.

I'm hoping to have time to prototype these this weekend--I've got a shipment of 2N5638's coming from Mouser tomorrow. That way, I'll be able to build them side by side...

:thumb:
thanks a lot for the projects :green:

Ok, so where is threshold control on that compressor?!?!

Ok, so where is threshold control on that compressor?!?!

Click to expand...

Wrong thread...

Sorry, haven't had time to delve into that circuit all week. I'm getting married in August, so planning takes priority sometimes, especially when my fiance keeps changing her mind about where we want to go for the honeymoon! :grin:

I've got the weekend free, since she has to work, so I'll use the time wisely then. Also, I got a big parts order yesterday, so I can proto both of these gain stages!

especially when my fiance keeps changing her mind about where we want to go for the honeymoon!

Click to expand...

D*mn, so you need to start looking for surplus-shops all over again each time the destination changes :wink: :green:

Hey, I never thought of that...

although the choices thus far have been Quebec, Tulum (Mexico), and Puerto Rico...I don't know how many surplus shops I'll find in those places...

Been with my wife 19 years...I'll tell ya this. If I could live again...boy I would be the one who is single at being single and is single. If it's got a single in it I'm in. It's a journey that requires patience, forgiveness, servitude lots of that one, really, generousity, compassion, ability to make yourself believe your own lies so that she will believe you, a guiltless conscience(this for down the road when young love is worn away and reality steps in) the kids will make it fun though as long as you stay and not divorce. And that is why I stay. For the kids. Anything I left out you can fill it in down the road. Just kiddin man...go for it