Active ribbon-mic

[David Curtis]

Those nominal impedance figures look right. Voltage noise will be predominant in either case. Note that the voltage noise density is 5nV/rtHz for the NE5532 and 4.5 nV/rtHz for the MC33078. This seems low relative to most applications, but it might be excessive for an active ribbon amp. The noise density of the AD797 is 0.95 nV/rtHz, which is 14 and 13 dB lower than the other op amps (respectively). Also, the NE5532 looks a little tight on the current budget with phantom power.

I'm very interested in how your work comes out. Thanks for sharing, and please keep us updated.

--Da5id

 

[bcarso]

It's possible to do better in some ways than the AD797 with a suitable mix of discrete components, particularly if one has access to some good dual parts for different sections of the circuit, but it is not for the faint of heart. The really heroic effort would be to come up with a preamp that eliminates the transformer and has the thermal noise of 1 ohm dominating. You won't get one of those to work off of phantom power, that's for sure, at least with any amplifying devices known today.

But, given the transformer impedance transformation, you don't need terrifically low voltage noise and current noise to get the ribbon to dominate---for your 1:24 stepup you are looking at about 3.1nV/root Hz, neglecting any winding resistance noise and core noise in the tranny. If you use JFET's the 2SK170 (or the new LSK170) is already giving you the same e sub n of the 797 at a mere 1mA, somewhat better at higher currents. And at audio frequencies the current noise contribution is negligible. Of course, a single FET does not a phantom-powered preamp make. The 797 also has a clever low-distortion and high-bandwidth first compound stage design. The 797 current noise is rather high for the magnitude of the actual base current---I don't quite know why. Maybe there is some bias current cancellation circuitry not shown in the simplified schematic.

One of the taxing things about this sort of work is the low values of feedback impedances required to not spoil the active device noise, especially as improvement comes reciprocally as the square of the resistances. 62 ohms as the feedback divider R for ~1nV/root Hz isn't too bad, as long as the feedback R itself is much higher, that is, plenty of closed-loop gain. Just what kind of peak voltages does one see at the terminals of a ribbon driving a 1:24 stepup anyway, at the highest SPL's?

Brad

 

[PRR]

> High-performance audio transformers with high turns ratios are harder to design and build (and thus, more expensive or lower performance) than those with lower turns ratios.

I believe this is a mis-statement. Even though I have seen it from people who might know better.

The real problem isn't the ratio (with few exceptions), but the highest impedance.

The impedance is determined by inductive reactance at the lowest bass frequency, and whatever happens at the treble end of the passband.

To get high impedance at your bass-cutoff you need many-many turns. 99.9% of the flux couples the other winding, but 0.1% does not. That is leakage inductance and blocks the highs. It is roughly in proportion to low-frequency inductance: more turns raises both inductances about the same. With good core iron it is less than 0.1% so is not a real problem by itself. If you can avoid matching and use amplifier inputs much higher-Z than the winding, leakage inductance is very little problems. But any winding has capacitance, and this does not scale with number of turns, it is fairly constant. A few hundred pFd for ordinary small teansformers.

Hi-Fi audio is hard to get past 10K impedance. If you take winding and amplifier capacitance as 300pFd, then at 20KHz the capacitive reactance is around 25K-30K. Leakage inductance will also be in that range. You end up with a resonance near or just above the top of the audio band. It warps your flat response and makes ringy-waves on the o'scope. With tricky windings, 10K is often a perfectly acceptable impedance level for good audio. But consider a 1K winding: it has the same 300pFd capacitance. That is still 25K reactance at the top of the band, hardly bothering a 1K working impedance. Also the leakage inductance will tend to be around 10K-20K at the top of the band. The resulting "resonance" has a Q of like 0.1, not resonant. As you go to even lower impedances the inductive ratio stays about the same and the capacitance becomes totally moot.

Consider a 50K:50K transformer. It is 1:1, but with conventional winding it will be pretty marginal for good audio. Or 1Meg:1Meg... even though the ratio is "only 1:1", response sucks.

There is an exception. Bifilar windings offer an order of magnitude better inductive ratio; less leakage inductance relative to bass inductance. They have high capacitance, but even so they give good performance. Bifilar is great at 1:1, may be good for 1:2. But it only has advantage for whole very-small numbers. It really does not raise the useful working impedance, it lets medium-impedance transformers get VERY wide bandwidth. The 1:2 tranny used in Jensen's Twin-Servo is an example.

So when you start from 150Ω, a 1:7 or 1:10 ratio is as high as you want to go (7K5-10K secondary impedance). But when you start from 1Ω, ratios of 1:50 or a bit more are all about the same. (The special-case, a 1:1 or 1:2 bifilar, is not a lot of use at 1Ω impedances.)

A reason to use high impedances: amplifier noise voltage is a function of device current. With a 1Ω source, a BJT would need over 100mA of bias current to get a low noise figure, and we can barely get 10mA from Phantom. And the Phantom also has to power the line-driver, which could (worst-case) need over 5mA of current just to whack the line really hard (though 1mA or 2mA is often ample).
You can get pretty good noise figure in 100Ω-400Ω with 1mA or 2mA of bias current, which works OK under Phantom power limits.

> what kind of peak voltages does one see at the terminals of a ribbon driving a 1:24 stepup anyway, at the highest SPL's?

At 192dB SPL? High!

At my max reference level of 128dB SPL (orchestral peaks)? A 150Ω dynamic will give maybe 50mV-100mV. If the ribbon were as efficient as a dynamic, and was 1Ω through a 1:24 tranny, that's around 600Ω so you would get 100mV-200mV. Ribbons are not as efficient as dynamics so I'll ballpark 50-100mV at most, for 128dB SPL. You wanna stuff your ribbon up a trumpet, you hold the SPL meter and do the math.

The level at the cable need not be the same of course. With conventional low-noise 150Ω mike amps, you want to stay at/above the level of a dynamic. 100mV at 128dB SPL. Higher levels probably will not improve system S/N much.

 

[David Curtis]

Thanks for pointing out the ribbon noise. The 1 ohm figure may be high for actual resistance, though. I calculate the DC resistance of an aluminum ribbon 2.5 microns * 3mm wide * 5 cm long to be 0.18 ohms, which, with the 24x gain gives 1.3nV/rtHz. Please correct me if I'm missing something here.

On the other hand, the DC resistance of the secondary is significant--on the JT-347 it's 483 ohms, yielding 2.8 nV/rtHz noise density.

Just what kind of peak voltages does one see at the terminals of a ribbon driving a 1:24 stepup anyway, at the highest SPL's?

The output of my Oktava motor (no transformer) is about 30 uV at 94 dB SPL, 3 mV at 134 db SPL (more or less its rated max SPL).

--Da5id

 

[David Curtis]

I believe this is a mis-statement. Even though I have seen it from people who might know better.

Thanks for the explanation, PRR. I've heard this statement so many times from people I generally trust (some of the gurus on rec.audio.pro) that I assumed it was true.

 

[PRR]

> I assumed it was true.

It is "true" for the special case where the primary is a certain impedance, and you are comparing several secondary ratios. This is a popular case: source is a 150Ω-250Ω microphone and we want to optimize an amplifier input transformer.

It was also true for the once-common case of a small low/med-Mu triode driving the grid of a next stage, except there the limiting ratio was 1:2 or maybe 1:3 if you were desperate. Going 1:5 out of an old triode was counterproductive: frequency response at the next grid was so bad that you got "less sound" than a 1:2 or 1:3 transformer. Narrow midrange peak with no bass or treble.

Yes, for many ribbons the resistance is closer to 0.2Ω (according to Marik). Your primary resistance should be much smaller than that. If you are transforming to a line and a conventional mike amp, the secondary DCR should be well under 150Ω, maybe like 50Ω. For driving long lines you want the total mike output impedance no higher than around 300Ω, and not lower than maybe 100Ω because a ribbon is usually low efficiency and we need all the step-up we can get.

So a 0.2Ω ribbon probably needs pretty fat primary wire and around 1:30 ratio. This isn't really an MC phono design: MC phono is 1Ω-10Ω source, 10K-47K load, we want 0.2Ω source and 200Ω nominal secondary impedance.

 

[bcarso]

Hmmmm---even more challenging. So, are there impedance curves and/or equivalent circuits to show the series inductive component of such beasts? If I suppose that Z is the 1 ohm bandied about at, perhaps, 5kHz, I get an L of abut 25uH to be in series with the 0.2 ohm DCR---is this about right? (sadly I think my Eargle books are in storage...)

If it is of this order than I can see where even more care needs to be taken when it is stepped up by factor 30. Is it sometimes a situation like MM cartridges, where an anticipated mechanical repsonse is compensated by an electrical one?

Brad

 

[Rossi]

I didn't have much time, but last weekend I breadboarded a simple impedance converter with two PNP transistors. So far I can say that it didn't help with the stock transformer of the Thomann ribbon. It performed about the same as without the converter. So I suppose the mic's output impedance is already low enough. I wish I had a good high turns ratio transformer on hand to conduct further experiments. But then again, I would prefer an inexpensive solution to improving an inexpensive microphone. :roll:

Maybe simple discrete circuit with a little (but not too much) actual gain would be the way to go. Most of the better opamps would probably require more current than is available with phantom power.

Here's how I define my goals in improving/"activating" the Thomann/Nady ribbon:

1. mic should be 6-12 dB louder with little or no additional noise
(The stock mic's self noise is spec'd @ 18 dB-A)
2. mic output impedance somewhere in the 200 ohms area
3. circuit should work on regular 48V phantom power
4. circuit should not be too expensive

 

[bcarso]

What does the mic now like to see in terms of input impedance of the preamp?

Thus terminated, what is the output noise corresponding to the stated acoustical noise floor?

So, given that, you can readily come up with a discrete amp meeting your criteria of 6-12 dB more gain and the 200 ohm output Z. I am sure it will work adequately with phantom power unless you want a gigantic output voltage swing.

Brad

PS: Anybody out there willing to measure the inductance of a ribbon element for me?

 

[Rossi]

Here's what I conclude from my impedance converter experiment: The gain loss due to the converter producing a gain of a little less than 1 was a bit higher than whatever could gained by the mic now seeing an impedance somewhat higher than a usual mic preamp. From that I conclude that the mic's impedance (post transformer) probably is somewhere in the 200 ohms ballpark. Maybe a little higher, but not very significantly so.

So I suppose an input inpedance of 5-10 K would be fine. I'm not really an electroics engineer (in fact I'm writing a dissertation on American poetry :roll: ), so I don't design circuits just like that. But perhaps I can find and steal from a circuit so cool, not even I can corrupt it. :green:

 

[bcarso]

Do you solder well at least?

What do you think of the poem in Pale Fire?

Brad

 

[Rossi]

Well, I'm not a specialist on Nabokov, but at least I have a good friend who is. :wink: In fact he is writing a book length study on Nabokov's poetry and the interrelation between poetry and prose in Nabokov's oeuvre, including, of course, Pale Fire. He's a Canadian married to a (very beautiful) Russian wife and hence speaks both English and Russian. So he's got all it takes, including a brilliant mind. My own dissertation is somewhat simmilar, except that I write about Paul Auster (who is way more popular in Europe than in America). Auster was an obscure poet in the seventies, then made a clean cut and became a very successful novelist. Never wrote a poem again.

I digress... My soldering skills are okay. I write gear reviews to pay my bills. So I have a different perspective on electronics than an engineer. Just like a book reviewer is not necessarily a great novelist, I'm not really a designer but a gear pornographer and professional audio snob. My job is evaluating what is there and explaining it to normal mortals. I can say with some confidence that I'm pretty decent at that. It's actually not so different from interpreting a poem. But could I come up with a posting like PRR's? No way.

 

[clintrubber]

from Rossi:

Here's how I define my goals in improving/"activating" the Thomann/Nady ribbon:

1. mic should be 6-12 dB louder with little or no additional noise
(The stock mic's self noise is spec'd @ 18 dB-A)
2. mic output impedance somewhere in the 200 ohms area
3. circuit should work on regular 48V phantom power
4. circuit should not be too expensive

So I suppose an input inpedance of 5-10 K would be fine. I'm not really an electroics engineer (in fact I'm writing a dissertation on American poetry ), so I don't design circuits just like that. But perhaps I can find and steal from a circuit so cool, not even I can corrupt it.

So how about the ESP #93 circuit mentioned earlier up this thread ? As far as I can see it seems to fulfill all items from your list.

Bye,

Peter

 

[bcarso]

Hmmm---I had no idea Auster wrote poems.

BTW your English is ausgezeichnet

Reviewing anything is treacherous---a surrogate for hard-laboring honest-day's-work kinda stuff, and inevitably leading to envy and resentment. Or does it? There must be some reviewers out there of something who really just want to do that, and not the activity reviewed.

And reading reviews is also treacherous: how many books do I think of as having been read when in fact I just read the Times Literary Supplement or some such? Auster being a case in point. Although I'm good at buying books reviewed, so at least supporting the biz.

But sorry to be way off the subject. I'll look at the ESP #93 as suggested by Peter/clintrubber and duly pontificate (cough).

Brad

 

[bcarso]

Hmmm. Seems like the ESP #93 is much better for quite a bit higher source Z's than around 200 ohms. I would run more current through the input device, say at least half a milliamp---right now it is mighty low at about 30uA according to my estimates. That puts voltage noise somewhere around 3nV per root Hz I suspect, while current noise is low, probably around 0.3 pA, thus the optimal source R is around 10k. Also, there is a lot of noise contributed by the highish feedback network values. But then at some point you will run out of gas using phantom power, so scaling everything down equally may become impractical.

This design was undoubtedly fine however for the original application, which was almost surely limited by the noise from the electret capsule.

I also don't much like the zener diode D1, as the regulator for when phantom power is the source. However, the zener noise may not be terribly important: it modulates the Q3 collector current a lot less than it might due to the action of Q2. Everything else is filtered and/or referenced to circuit ground, so one is probably ok. I may do something a bit more quantitative and get back on it.

Brad

 

[bcarso]

I did a bit more work looking at how to reduce the noise of that basic topology, which is overwhelmingly dominated by resistor thermal noise in the schematic as published. I wound up with something pulling 6.7mA and with the feedback R and divider R down to 500 and 150 ohms respectively, with more current flowing in most everything. This would still be within the range of regulation, sort of, using phantom power so it's not too impractical. Still, after all that work I predict about a 3dB noise figure for a 200 ohm source. To do substantially better I would look at adding a class AB output so that it is easier to drive an even more stiffish feedback network. Sounds like a shift to surface mount based on at least the original desired board size and increase in component count.

Brad

 

[clintrubber]

Interesting comments, a nice read.

Yes, the ESP-circuit targets a different Zsource, I neglected that. It'd be tempting to give a ribbon-mike some non-standard arrangement, using more mA's than standard-phantom can provide, but that'd be too easy a way out, not ?

I'm curious to where this all ends up.

I've never done anything with such a topology, but for some reason 'common-base' pops up somewhere in the back of the head. Misplaced ? :roll: (There's probably a one-sentence remark that explains already that this is not the way to go :wink: )

Sounds like a shift to surface mount based on at least the original desired board size and increase in component count.

I was eager to use an already existing circuit but you're probably thinking about putting a circuit inside the mic itself, correct ?

Bye,

Peter

 

[bcarso]

Yeah, I was supposing that one wanted to put this inside a housing. If not then sky's the limit on parts, power supplies, etc. I like the idea of the convenience of phantom power but it is a severe limitation in terms of current (and even of power, supposing that one decided to do a specifically tailored switching regulator, with all the potential havoc that could entail!).

Common-base stages are marvelous for many apps, especially if you want very low input Z's. They have the same noise performance as common-emitter, although that can be concealed by the noise arising from the bias networks needed.

In this app one wants to lightly load the Z seen after the tranny, and that condition is easier to achieve with the input going to a base, particuarly with the series feedback similar to the circuit under discussion. But to get noise down for looking at a 200 ohm source the current in the transistor has to be healthy, its rbb' low compared to 200 (easy for most parts) and the noise in the gain-determining series feedback network has to be low. This dictates low values for those R's and the ability to drive them comfortably at maximum output levels. With "normal" audio and if necessary a big fat reservoir cap (getting squeezed out of that existing housing again now), and with a class AB buffer stage to drive the feedback network and the load, you could get down to well below thermal noise in 200 ohms. Actually it not quite as bad as all that: with the R's set up for similar gain to the existing circuit and PRR's estimates on the output from the ribbon and 1:100 transformer you could just about get there: with a 100 ohm feedback divider R, 330 ohm feedback R, you'd need about 6-7 mA peak if I've done the maths correctly.

Notice though how feedback is constraining us here though. If we contrived an open loop design with acceptable distortion performance (no mean feat) we could avoid the stiff feedback divider drive requirement.

Brad

 

[Rossi]

There's two things that I don't particularly like about Project 93, i.e. the unbalanced input and the use of electrolytics in the signal path.

I'm not quite sure whether to put the circuit in a little box or into the mic itself. The mic's quite big, so it probably would fit in there. But I would actually prefer using the active circuit as an option, so an external box would be preferable. In the case of an external box, the input should be balanced and I wouldn't want to use a transformer for that.

Electrolytics on the output could probably eliminated. Some other electret mic circuits don't need them. The Schoeps and Dorsey circuits also don't use electrolytics on the output. It may be more difficult to get rid of the input caps. I would at least like to be able to use polyester caps. Rising the impedance could be a way to go, I guess. At 16 k input impedance a 1µ cap would do (10 Hz lowpass frequency). Given the bassy response of the mic itself, a bit of low cut would be acceptabe.

The more I think about about this project, the more I realize that it is perhaps more akin to a microphone preamp than an electric mic circuit. Source impedance is around 200 ohms and output impedance should be around 200 ohms as well, perhaps lower. Input and output should be balanced. Special requirements are that it should operate on phantom power and produce optimal performance for 6-12 dB of gain.

That actually made me think of PRR's 48V single ended preamp.
Here's PRR's thread. Maybe that design could be scaled down for as little gain as we need and as little current as we can get from phantom power.

@ Brad: I don't think reviewing necessarily leads to envy and resentment. That may be the case when it comes to books, because a lot of reviewers are frustrated novelists or poets. But that frustration, I think, came before they became reviewers. I became a gear reviewer by chance. If I was frustrated by anything it was the sometimes sloppy work of other gear reviewers. I'm definetly not a frustrated designer. In fact, I have nothing but the greatest respect for good engineering. A good schematic is a bit like a good poem. Everything must fit and work hand in hand. You see a brilliant mind at work, intelligent decisions being made, and sometimes there is an element of staggering creativity. At times there's something that you don't understand, but, hey, it's that mysterious little something that makes it sound so great. Both circuits and poems are artifacts of the human mind. So, if I sometimes make excursions into designing, it's only to understand and thus appreciate things better.

re Auster: get The New York Trilogy or Moon Palace. Those two books are as good as it gets, I think. BTW: his wife is a successful novelist, too. Her name is Siri Hustvedt.

 

[bcarso]

Rossi: thanks for the Auster recommendations. Of course I was somewhat tongue-in-cheek about reviewing (maybe I should have indicated that with the odd smiley or something) but points well taken on your part.

For one cap to go away, just connect the tranny output between the base of the input Q and its bias network---a microamp or so of base current shouldn't hurt the tranny too much. The C in the feedback divide leg is tougher however. For this topology and lowest noise, it needs to be large if you are going to preserve low freq response.

The design is balanced as regards impedances as is, which is the significant consideration from an induced-noise rejection perspective. However, the highish output voltage on one leg and none on the other could be an issue at some downstream inputs I suppose, unless the first thing you hit is another transformer. Of course you could throw an inverter in and drive the other line, but you burn up precious power that way. Maybe abandoning phantom power is the best way to go---it's such a farking constraint.

I have to hit the road and will be out of internet access for a while, also probably simulator access. All the better to use pen and paper and calculator---maybe I will work on a balanced-with-complementary output voltage design. Or maybe I'll work on taxes (yecchh). Or relax (hah).

Brad