Discrete DI-Box

[recnsci]

[quote author="clintrubber"]
BTW & semi-OT, apart from eventual lesser headroom, any disadvantages for this 2*JFET+BJT-threesome ?[/quote]

I dont know, I just see advantages

 

[Samuel Groner]

BTW & semi-OT, apart from eventual lesser headroom, any disadvantages for this 2*JFET + BJT-threesome?

The CFP often shows some peaking at high frequencies and pretty ugly asymmetric slewing. If this matters or not is as often another question.

Samuel

 

[clintrubber]

[quote author="Samuel Groner"]

BTW & semi-OT, apart from eventual lesser headroom, any disadvantages for this 2*JFET + BJT-threesome?

The CFP often shows some peaking at high frequencies and pretty ugly asymmetric slewing. If this matters or not is as often another question.

Samuel[/quote]

Just thinking out loud, not verified, but I expect the peaking could be tackled by a cap across the PNP (as previously mentioned by Brad in other CFP-related threads).

And the slewing by something like this:

(from a well-known site - http://www.dself.dsl.pipex.com/ampins/discrete/cfp.htm )

The 'phase-splitter' (...) arrangement of the circuit further above now
needs to be adapted to be both splitting & push-pull.
So (if at all possible without to much fuzz) then better consider
it as a SE-pushpull-stage driving an again added output-TX.


But then again, I don't think that for that 2*JFET+1*BJT circuit
there are severe consequences of any asym. slewing of that
three device 'CCFP'(TM), this because of the EF's after it.


Regards,

Peter

 

[clintrubber]

BTW, Samuel, I hope you don't consider the previous stuff too OT - I'm of course not after messing with your design (who am I to...), but interesting thread so nice to discuss alternative approaches.

 

[clintrubber]

Hi,

[quote author="clintrubber"]
http://twin-x.com/groupdiy/albums/userpics/condenser_mic_preamp.jpg[/quote]
FWIW, here's more info on the 2*JFET+1*BJT-based circuit from above:

http://www.pat2pdf.org/patents/pat7072478.pdf

Bye,

Peter

 

[Gus]

I don't understand what the patent is for

It seems to be a version of the standard schoeps OK the front end is a little different setup as a phase splitter And the output has maybe a better bias R8,R2 are added.
The Schoeps circuit you can find "seems" to be aware of the P50 drain to gate voltage as are the neumann schematics using P50s

Cascode?
http://www.josephson.com/pdf/C617set_ds.pdf


What is the patent for the active shielding drive from the source? Something like a active guard ring for opamp inputs?

Question what is the Schoeps circuit with the long extention tubes with the capsule at the end?

 

[recnsci]

[quote author="clintrubber"]
Just thinking out loud, not verified, but I expect the peaking could be tackled by a cap across the PNP (as previously mentioned by Brad in other CFP-related threads).

And the slewing by something like this:

[/quote]

Maybe I'm wrong, but slewing will be determined by total capacitance at
pnp base. Thus, it will be limited by 0.6V*1K2 on one side, and lower
device colector/drain current (which will give as much as you want) on
other. This will be the cause of asymetric slew.
CFP will have plenty of stability C because of Cgd of 2sk170. OTOH
Cdg will be highly nonlinear, which will have odd effect.

BTW that fet cascode + pnp thingie will keep Icascode constant, thus
Cgd of first device virtualy perfectly linear in audio band. Plus THD drops
by order of magnitude. I cant envision better capsule buffer device
than this cascode-CFP combo. If somebody has one please post schematic

cheerz
urosh

 

[clintrubber]

[quote author="Gus"]I don't understand what the patent is for[/quote]
It's not fully clear to me either, but it's at least a nice source of info this way.

I hadn't seen the combination of a cascode JFET for a CFP before, but it'll have been done before I could imagine. If it hasn't have a name yet let's call it a CCFP: Cascoded CFP (or C Cascoded FP if you like).

The feedback may be 'new' for micpre-JFET-frontend (R16, R15, R14)
but looked from another way it is the same principle as for instance here: the 150k below Q1,Q2 (the first drawing that comes to mind of a common circuit):
http://www.twin-x.com/groupdiy/albums/userpics/normal_Altec-1578A-circuit_unchecked~0.jpg
The details differ, but the essence is that there's negative feedback around the input-device for working point stabilisation.


Last but not least:

I'm puzzled by the 2SK170 & 2SK117 though: the patent/schematic specifies much lower Idss-ratings (0.8 .. 1.0 mA) than the datasheets of each of these devices show: all classifications (Y,GR,BL,V) are higher.

The -Y for the 2SK117 comes close at 1.2mA-3mA, but the 2SK170 doesn't seem to have than and starts at 2.6mA (GR).

Any idea here ?


Bye,

Peter


PS
The CCFP is a bit in contrast with the two simple EF's. So for some overkill, let's make those at least CFP's as well....

 

[bcarso]

I don't know about the discrepancy between the patent callout and the schematic posted, but I would hazard that running J1 at that low a drain-source voltage is not a great idea. It does conserve voltage in the stackup of parts but makes the capacitances awfully large, although at least not signal-swing-variable by much.

Rather than the prescription given, if using an SK170 I'd get hold of the highest pinchoff (V for violet) variety---although the max value in the Toshiba datasheet is 1.5V, which is probably never seen with any actual device and just makes the QA people feel more comfortable with their capability measures. The curves show a 10mA Idss device having only about minus 400mV Vgs at the approximate 0.5 mA operating point set by Vbe of Q1 and R13. That's a mighty low Vds for J1, which looks like it's just getting out of the "triode" region.

At least the gate leakage will be low.

 

[Gus]

some schematics of microphones have cascode fets. Some time ago I starting to trace my 103 it looked like it has a cascode fet. Dale posted a nice trace of the circuit so I stopped tracing the 103.

Look at the 414s, P48,ULS,TL at
http://www.geocities.com/siliconvalley/peaks/5212/akg1.htm

I believe Decca did some cascode circuits in microphones

For vocal microphones you might not want or need a cascode. IMO there is no best condenser circuit tube or solid state. Why is some input cap a bad thing?

 

[bcarso]

[quote author="Gus"] Why is some input cap a bad thing?[/quote]

If it were constant with voltage swing it would be of little concern. Often people say nonlinear capacitances cause distortion, but more correctly all you need is capacitance change with voltage---the C versus V relation can even be linear, although it probably never is in solid-state devices.

There could be situations, perhaps even in condenser mics, where complementary C variations could cancel somewhat---so it could even reduce overall distortion.

If signals are small enough there's not usually much to worry about.

 

[Gus]

bcarso

I sometime wonder if simple circuits sound good because of the "flaws" as you noted in your post.. I sometimes think I hear a difference if I reverse the backplate and front of a condenser microphone also flipping the outputs.

As time goes on I seem to like transformer output microphones better than transfomerless for vocals. Now the circuit like in this thread might be very nice for SD microphones for other applications.

The u87 Km84 type circuits are "simple" one fet gain stage and a transformer but they can sound nice IMO.

 

[bcarso]

Well I would certainly trust my ears in the final analysis. Having said that, I think the apostles of Simplicity as Holy Grail (Nelson "Zen" Pass as one prominent example among the highend audio folks) often go astray.

I suspect guys around my age, like Brad Plunkett (older) and PRR (probably younger), who remember when transistors were expensive and often not that reliable, and who have seen plenty of equipment failures, often value a certain measure of simplicity on this account. I had the luxury, when I did bleeding-edge research equipment for academia, of putting complex machines together with mostly high-rel parts and being personally responsible for each solder joint. I know how reliable such equipment can be provided the design is sound. But as a design approach for even low-volume production, it doesn't cut it. Basically, nobody cares enough to be that careful.

As far as what we really can hear, it's a fascinating area. Sometimes I begin to think there are extra-auditory senses being brought into play :roll: . Then I usually have another glass of Merlot or whatever is open and nod off.

 

[Gus]

I saw this at the Liberty Instruments site

http://www.libinst.com/cepst.htm

 

[bcarso]

Skimming it just now, that seems like an interesting article. I seem to recall the Bogert et al. paper too---I may even have that book. Tukey is probably responsible for the puns etc.---he's quite the humorist.

 

[clintrubber]

[quote author="Gus"]Look at the 414s, P48,ULS,TL at
http://www.geocities.com/siliconvalley/peaks/5212/akg1.htm[/quote]
Thanks for that link.

Ah, they did the 2*JFET + 1*PNP like in the T.H.E.-circuit already, be it that AKG didn't use it as a 'phase splitter' (T101,102,103):
http://www.geocities.com/siliconvalley/peaks/5212/buls_1.pdf

Saw that some of those other AKG 414-circuits use a cap C4 from EF-out to JFET-drain, might work as well for Cgd:
http://www.geocities.com/siliconvalley/peaks/5212/C414E1.pdf

Regards,

Peter

 

[clintrubber]

[quote author="bcarso"]I don't know about the discrepancy between the patent callout and the schematic posted, but I would hazard that running J1 at that low a drain-source voltage is not a great idea. It does conserve voltage in the stackup of parts but makes the capacitances awfully large, although at least not signal-swing-variable by much.

Rather than the prescription given, if using an SK170 I'd get hold of the highest pinchoff (V for violet) variety---although the max value in the Toshiba datasheet is 1.5V, which is probably never seen with any actual device[/quote]
I see what you mean.... Funny thing though is that I understood that several AKG 414 versions use an alike topology (2*JFET + 1*PNP) with the 2SK223 (Sanyo) for both JFETS and that type has a VT range from -0.2 to -1.2, so the same idea as the T.H.E.-circuit :roll: ....

 

[clintrubber]

[quote author="bcarso"]Well I would certainly trust my ears in the final analysis. Having said that, I think the apostles of Simplicity as Holy Grail (Nelson "Zen" Pass as one prominent example among the highend audio folks) often go astray.[/quote]
Certainly, absolutely agreed on keep using ears.
I admit the interest in these circuits may have gone perhaps a bit too far in the direction of 'let's add yet another cheap component to remove yet another potential distortion adding influence' (that's at least where I'm heading myself).

But it's fun, learning from it and for a clean-intended DI it'll indeed be a good thing - an eventual nice sounding attribute will go out of the window w.r.t. simpler 'sound-shaping' DI-circuits but that's just the consequence of the direction that was taken.

 

[Samuel Groner]

A quick update on this one. I designed a cute PCB for the circuit (check the first post for the layout file) and finished soldering the first board today. The emitter resistors for the BJT output (R8 and R9) will need different values as the quiescent current got a bit low but otherwise everything seems fine. 100 kHz square wave response is about as clean as it gets before the electrons hit the iron and headroom is--as easily predicted--very high. My HP signal generator doesn't have enough volts to make it clip.

Samuel

 

[Samuel Groner]

It turns out that a 120 ohm resistor in series with D3/D4 is the better fix for the output stage quiescent current. As shown it was only 400 uA--never seen Spice to be that much off with DC analysis. The resistor value is ideally selected for every board for a 30 mV voltage drop across R8 (or R9) as the quiescent current depends on several relatively ill-defined parameters.

In case I'd feel like doing another board revision I think I'd replace D3/D4 (and the additional resistor) with a trimmable Vbe multiplier, generously bypassed with a capacitor.

Input headroom at 1 kHz is about 30 Vpp (+22.7 dBu). At significantly lower frequencies the transformer starts saturating earlier and limits the maximum output voltage to lower values.

Charging C6 takes a few minutes. At the beginning D7 conducts and injects serious common mode noise into the output. No problem with that as after start-up this will stop completely. It's pretty funny to observe the noise waveform though as it changes drastically with zener current. Once it gets towards zero the waveform becomes very jagged with hefty one-sided voltage spickes.

Samuel