Active ribbon-mic

[bcarso]

Ouch ---regarding the Oktava schematic! I wonder if maybe the first 100k is really tied from collector to base instead of supply to base. That would help a lot for openers.

I liked some of the specs on the THAT parts but they seem to have had trouble making parts do high beta. This is less important than it would be in some other applications. They do have low spreading resistances and hence good log conformity---good for some VCA apps. Maybe they have improved their process a bit since I last looked.

Linear Integrated Systems also makes a few bipolars but I haven't evaluated them yet.

Brad

 

[clintrubber]

from PRR:

> why better use big switching types i.s.o. low-noise types

If you have a free choice of source impedance, wind to about 1K-10K impedance and use a small high-gain transistor. ...

However the "low noise" transistors were developed 30 years ago when there were a lot of non-ideal excess-noise transistors around. They have high Beta and selected processing and are still good choices for impedances like 1K-10K (though most any high-Beta transistor will work about the same).

Thanks for the explanation - after reading your text it all sounds obvious. :thumb: Hadn't realized 'low-noise' should have read 'low-noise under this&that conditions'... interesting those switchers can come to the rescue.

Regards,

Peter

 

[Rossi]

Yes PRR, those THAT parts look nice. I wouldn't hesitate to get some for a nice preamp (perhaps your 48V single ended design). But for the budget ribbon upgrade project, I don't know. To me, the non-transformer design makes sense only as a low cost solution. If I have to spend that kind of money, I'd rather opt for the transformer solution, because that one can operate on phantom power. Plus, I'd have to order the THAT parts from yet another online store.

That said, profusion does offer some nice chips. I picked up their print catalogue at the Frankfurt music fair. The NJM2114 looks kind of interesting (generally speaking, not related to this project), it's a supercharged NE5532 at roughly the same price.

 

[clintrubber]

FWIW:
Triggered by the above explanations of base resistance of the various transistors, the thought struck me that there might be consequences for the electronics following Linkwitz-modified electret capsules.

I mean, in the standard non-modified form (common source) the output-impedance of the capsule is a few kOhm which will have been taken into consideration for the following electronics (the ESP#93 circuit comes to mind once more).

Once the internal FET of the capsule is modified to a sourcefollower-configuration both the source-impedance & output-level of the capsule drop though, likely causing a non-optimal situation w.r.t. noise.
In other words: one would think 'Linkwitz-modifying' electret capsules could require a modification of the subsequent electronics as well.

But I guess that it'll all not matter that much since the output-impedance of the internal FET will still be some kOhm - higher than a BJT at that 0.5mA.

And that's why I started this with FWIW, just a thought, back to ribbons again... :wink:
(but please shoot when I'm mistaken)

 

[PRR]

> Once the internal FET of the capsule is modified to a sourcefollower-configuration both the source-impedance & output-level of the capsule drop though, likely causing a non-optimal situation w.r.t. noise.

1) The change in impedance is small. Very possibly no change at all. The specsheet resistor is 2K and 1/Gm of that little FET at that low current is also about 2K.

2) The noise is dominated by FET channel noise and, for the $2 electrets, gate leakage noise. You would have to use a very bad BJT buffer to make things worse.

FWIW: I have a couple P-sonic omni capsules boosted by TIP120. Which if you look it up, is a POWER device, and also a Darlington. And connected as an inverting "op-amp". Between two B-E junctions, big-die leakage, external and built-in resistors, it isn't exactly low-noise. But the noise drops if I replace the capsule with a resistor. The capsule noise dominates. And in the room, the capsule noise is not an issue compared to room noise. It isn't a really quiet room, but not noisy.

 

[clintrubber]

The mechanism came to mind, but it's 'convenient' that the figures are still resulting in an alike situation. With this much Linkwitz-mentioning on the net I stll never had read anything about consequences of the mod, so a non-issue indeed. I was actually a bit surprised again about how 'lame' the output-impedance of a sourcefollower is at such currents :?

The capsule noise dominates.

Different situation indeed for those electrets vs ribbons !

Regards,

Peter

 

[PRR]

> surprised again about how 'lame' the output-impedance of a sourcefollower is at such currents

FETs similar to electret capsule FETs:

http://www.fairchildsemi.com/ds/FJ/FJX597JC.pdf
http://www.fairchildsemi.com/ds/PN%2FPN4119.pdf

Transconductance at a small fraction of a milliAmp is 300 to 1,200 micro-Mhos, which is 3,000 to 800 Ohms.

The output impedance of the Source Follower will be about the same. Drain conductance is in parallel but this is 0.5-1.5 microMho, about 1 MegaOhm, so no big difference. With the P-sonic 3-wire self-bias connection, Zout might be half, but Vout is also half due to the Source bias resistor. With the Linkwitz connection the Source resistor is much bigger than 1/Gm so makes only a small difference. The P-sonic 2-wire connection is the external Drain resistor 2K || the drain conductance 1 Meg, or 1K996.

We also see why the level does not change a whole lot when Linkwitz-ed. The Source resistance 1/Gm is about the same as the suggested external Drain resistor in the 2-wire connection, and they flow the same current. A 3-wire P-sonic capsule (now rare) converted to Linkwitz does show some increase of sensitivity due to higher source bias resistor. (And much higher overload level.)

http://www.fairchildsemi.com/ds/PN%2FPN4119.pdf also shows a typical noise curve. Take an eyeballed A-weighting, say 15 nV/√Hz, we have about 2 microVolts of noise. A BJT follower with 1mA current will be around 0.5 microVolts added noise, hardly any increase at all (much less than 1dB). You could follow with a '741 or a TL071 with only slight noise increase. And this assumes the mike capsule noise is mostly FET channel noise. In fact the acoustic damping adds noise, and these $1 mikes omit the GigaOhm resistor and let gate leakage "set" the gate bias, a very noisy trick. Many of these capsules are rated for noise above 30dB SPL. Compare with a $999 condenser: 14dB SPL. Most electrets are less than 30dB SPL self-noise... 2 to 5 microvolts is typical, but the factory want to cover their butts for wheezey samples with 10 or more microVolts noise.

 

[clintrubber]

Thanks PRR, interesting read. And nice to know some ballpark-figures of the FETs inside those capsules.

The PZM+ESP#93-output is fairly hot (I thought the dimensioning I used for the ESP does ~10dB gain) so some drop in level would have been no problem. I'm converting, let's see how it turns out. Noise is ~noticable but OK; in these loud times the problem is of course at the other end.

Bye,

Peter

 

[Rossi]

Okay, here's what I did last weekend. I built PRR's non-transformer battery preamp ("no-tran-ribbon-boost.gif"). So far I'm still without those 2N4401 transistors, so I used BC549Cs instead. As it turned out those were about the only NPNs in my poor collection.

As I didn't have an appropriate battery holder, I used a 9V battery instead of four AAs (6V). Current draw was about 15 mA, btw. Apart from that I used 2.2K instead of 2K for R8 and R9 and 220 µ instead of 100 µ for C3.

Despite my changes, the preamp wasn't so bad. Sound was good, noise wasn't quite as good, but not absolutely terrible, either.

Now, here comes the more intersting part. I remembered that I still had some AC187 germanium transitors, which I bought a while ago for use in a guitar stomp box. Quite surprisingly, the preamp was much lower noise with the AC187. That's kind of weird, because germanium transistors usually aren't known for low noise. Moreover, germanium transistors are usually PNPs with very few exceptions. The AC187 is one of those exceptions. It's the counterpart to the PNP AC188; the two were often sold as (matched?) complementary pairs. One of the main applications, AFAIK, was small output stages (about 1 W) in radios etc. So the AC187/188 are somewhere between small signal and power transistors. I suppose their size is where the low noise comes from. I'm curious to see if silicon switching transistors will perform as good or even better than the AC187.

BTW. I tried all ten BC549Cs I had; no difference between them. Of the AC187 I had three, which I tried in any combination. No difference between them, either.

As is, the AC187 equipped PRR-non-tran-ribbon-boost preamp is totally usable. I still wish it was phantom powered, but apart from that it really does the trick beautifully.

Here's a few mp3s (300-350 kB, sorry for the unnecessary stereo):

http://www.andreashau.info/audio/PRR_BR_Pre_transistors+Mindprint_Trio.mp3

This is the PRR pre in front of a Mindprint Trio, a brand new inexpensive all-in-one box (which I recently reviewed for German Keyboards Magazine).

http://www.andreashau.info/audio/Ribbon_passive_on_Siemens_and_Trident.mp3

This file is for comparison purposes. You hear the budget ribbon directly connected to a Trident 4T channelstrip (EQ and compressor are off) and to a Siemens V272 (with variable gain). The Trident is very low noise, the Siemens isn't quite as low noise, but otherwise a nice sounding fellow.

http://www.andreashau.info/audio/PRR_BR_pre+Siemens_and_Trident.mp3

This is the PRR pre in front of the Siemens and the Trident. The Siemens is a little lower noise here than without the PRR pre. The Trident becomes noisier with the PRR in front.

 

[bcarso]

Rossi, that's fascinating. Of course there is nothing per se about germanium that is bad--- a smaller bandgap, and more consequent leakage at a given temp than Si, and a tendency to misspell as geranium---but nothing to rule them out for some apps, except lack of availability. It is true that in the mostly pre-silicon days Ge was not so good, except for low saturation resistance and a gentler on threshold. I don't know the vintage of your parts but they could well be much better than I remember the old germanium to be.

Some of us (not-too)oldsters may recall the ads from Germanium Power Corp. (iirc), featuring the indefatigable, portly, and hirsute president who appeared in the ads as Oliver Germanium, touting the benefits of the material and their optimized devices for low voltage high current apps. I think they even dabbled in photodiodes at one point, which have a nice near-IR response owing to the smaller bandgap. They have been trumped by fancy composite materials since.

Now as a composite with silicon, germanium is making a serious comeback for ultra-high-speed logic parts. I believe Barrie Gilbert (high priest of analog at Analog Devices) talked recently about some amazing analog parts using the hybrid material as well (betas of >100k, etc. etc,). We are not likely to see these as discretes anytime soon though.

Brad

 

[bcarso]

PS: Si Ge also has a big app for RF analog parts

 

[PRR]

> I used BC549Cs instead

As you found, these work but not big enough for lowest noise.

http://www.onsemi.com/pub/Collateral/BC549B-D.PDF page 3 shows Rb for the BC549. At 4mA about 147Ω, at your ~8mA about 135Ω (numbers are not really exact!). So from base to base you have about 270Ω of resistance that makes noise but does not make signal, plus ~200Ω of resistance that makes both noise and signal. The preamp makes more noise than the mike.

The AC187 is a fat device. Ge can be low-noise, and for a few years was preferred over early Si devices (which had NF around 20dB when Ge was pushing 5dB). The small problem with Ge is excess current noise, which would not be a problem at this low impedance.

> I tried all ten BC549Cs I had; no difference between them. Of the AC187 I had three, which I tried in any combination. No difference

A transistor's noise is either as low as it gets (for size and bias) or it has a problem and makes excess noise. I remember when half of a bag of transistors would have excess noise. Even in the 1980s Radio Shack would sell you a bag where 10% had excess noise. That rate is probably down below 1% now, even for dirt-cheap transistors.

> The Trident becomes noisier with the PRR in front.

If it is transformerless, it is probably nearly the same input topology (diff-pair at a few mA) except they got the big transistors. So yeah: at the same total gain, the noise figure is surely degraded using BC549, and may be degraded using AC187. The idea was to make it possible to use a lesser input that did not have enough gain, without a big sacrifice in noise. Interesting that the PRR is not much worse than the industrial-quality Siemens V272, which was good enough for a generation of German broadcasters.

I guess you wanna Phantom it next. I'll have to get on the other PC. Taxes and dead-printer issues may not let me get to it for a few days.

 

[clintrubber]

PS: Si Ge also has a big app for RF analog parts

After reading about SiGe on page 6 I was about to add that yes, but it was already on top of page 7 ! :grin:
Right, as I understood it they 'Ge the base' and get benefits from that.
Which doesn't mean the +Ge is always better suited - I remember for a certain RF-function at work after a Si & SiGe comparison the Ge-less process was used.

Hey, this is far away from Ge's for a FuzzFace ! :wink: Hi Rossi, don't forget to make that stompbox - I did a FuzzFace-clone/variation a while back and used OC44's - sounded real nice and better than the originally intended AC128's. You were thinking about a FuzzFace as well ?

But back to topic, interesting Ge-observations.
My T-Bone arrived today, the toying around starts and want to get a feeling how much gain my use will require.

Bye,

Peter

 

[Rossi]

Brad, the AC187 isn't too old, I guess. I don't know when it was introduced, but I bought mine brand new in an electronics shop about a year ago. I bought all the germanium transistors the guy had. Which was a total of six: three AC187 and three AC188. Two AC188 ended up in a treble booster/distortion pedal. Those transistors are still available at some online stores such as Reichelt. I think I'm going to get a few more. I think I might try them in a mic pre as complementary pairs. Maybe they'll add some mysterious unprecedented germanium flavor.

Yes, I actually heard about that si-ge compound thing in computer chips when did some germanium research for my treble booster project. I didn't know AD is interested in that technology, too. Interesting.

PRR, if this thing could be operated on phantom power, it would be a very cool project for anyone ordering these budget ribbons. As I said it works wonderfully with simple preamps such as the one in the Mindprint box. And that's ultimately what the ribbon booster concept is all about. I don't think it's much of a problem that it doesn't work well with the Trident or similar preamps. The Trident is low noise anyways and has plenty of gain.
The Trident 4T preamp is based on a INA217 chip. But I guess John Oram doesn't just use the schemo in the data sheet. Input impedance, according to the manual is "auto tracking" - whatever that means.

Yeah, the V272 isn't very low noise with ribbons. It's originally a 34 dB fixed gain preamp; it simply isn't made for such high gains. But together with the PRR pre (using AC187s) the total noise was better than the V272 alone. Very nice. So thanks a lot for the PRR Budget Ribbon Booster (PRR-BRB)!

As I said, I'm gonna try some si-switching transistors as soon as I can. Hopefully next weekend.

 

[bcarso]

Again, if you can get your hands on them, the Toshiba 2SA1015's have 30 ohms rbb', and the NPN complement the 2SC1815 50 ohms. Try to get the high beta varieties if possible: the GR for the 1015 and the GR or BL for the 1815.

 

[Rossi]

[quote author="clintrubber"] Hi Rossi, don't forget to make that stompbox - I did a FuzzFace-clone/variation a while back and used OC44's - sounded real nice and better than the originally intended AC128's. You were thinking about a FuzzFace as well ?
[/quote]

Peter, my stomp box is already done. It's based on a Dallas Range Master treble booster. Actually it's more or less two of them in series, only the second one has a much larger input cap. So only the first stage works as a treble booster; the second stage is a simple gain stage which is overdriven by the first stage and, at higher output levels, overdrives the amp input as well. So you can dial in any amout of insanity. I originally wanted to get AC151R transistors for it which are said to be a good, low noise replacement for the OC parts that are very hard to get around here. But couldn't find any AC151Rs at local electronics shops. The only germanium ones I could get were the said AC187 and AC188 transistors.

That said, I'm not entirely sure I like the germanium treble booster better than the silicon one I built before I had those germaniums. Maybe I'll build another one with silicon in the first stage and germanium in the second. That would be my simpleton way of si-ge compounding. :wink:

 

[Rossi]

[quote author="bcarso"]Again, if you can get your hands on them, the Toshiba 2SA1015's have 30 ohms rbb', and the NPN complement the 2SC1815 50 ohms. Try to get the high beta varieties if possible: the GR for the 1015 and the GR or BL for the 1815.[/quote]

Alright, I can get those at www.reichelt.de , but they don't say which variety. They're cheap, so I'll just order a few. I don't think I can buy those at a local store.

 

[PRR]

OK, Phantom power is easier than I thought:

--attached below

Max input level 60mV for 1% THD (almost all 3rd). Around 15mV (more typical for weak ribbon) it is point-Oh-something.

Transistor must be rated 40V. If you have mystery transistors, they won't be killed if they happen to be lower breakdown, they just won't work.

Current demand is 5mA for nominal 48V Phantom. It will work on 24V 6K8 Phantom, but overload is less.

You still want some "fat" transistors.

> I don't think I can buy those at a local store.

It has been many years since we in the USA could expect to find a good selection of parts locally. I can still usually get TIP120 at a Radio Shack in the next town, but all the parts shops closed years ago. DigiKey and Mouser are our "parts shops". I looked at www.reichelt.de but that's a very annoying web site.

What do big companies and DIY-ers in Germany use for transformerless mike inputs? There has to be something like the 2N4401 that you find in "all" US-design good mike inputs.

Basic TO-92 high-current switch transistors cost about $0.10 each, but if you can't figure out which one is good, there is always the LM394. Reichelt lists it, but: "LM 394 TO NPN-Transistor-Pairs, TO-99 (H08C)  ?7.70 inkl. MWSt.". Ten Bucks each??? (More or less depending on today's exchange rate.) These are in the metal-can, but it better be 18 caret Gold for that price. DigiKey will sell me the plastic-DIP LM394 for half the price.

Come on, Germans. Help Rossi find some good cheap input transistors in/near Germany.

 

[bcarso]

PRR: "Come on, Germans. Help Rossi find some good cheap input transistors in/near Germany."

With all the diplomacy flying about you'd think a little of it could be directed toward the needs of decent analog electronics practitioners.

Of course it's a matter of economics, primarily, that moves suppliers like Toshiba to not support many of their parts in many areas of the world. I prefer that to hoping to influence a bunch of commissars meeting to determine resource allocation. Nonetheless, it would be nice if someone could see a way to making a reasonable profit providing parts. Probably the import/export laws are intimidating as well, but there must be a way.

As I have mentioned before, the very nice bipolars from Toshiba are less than a US penny and a half in China these days. The freedom this gives one to do some fairly remarkable designs for cheap is wonderful. It doesn't hurt that the stuffing labor is cheap either.

Regarding the LM394---the TO-99 case is the problem there, although that's still a gouge. Those damn glass-metal seals. And although it has no effect for most of our purposes, it's disheartening to know that the borosilicate glass is not even a very good dielectric.

 

[clintrubber]

Hi,

If anyone's interested: (attempt to a) thread about the use of those budget ribbons is here:
http://www.groupdiy.com/index.php?topic=6277