Active Ribbon Preamp/Impedance Converter Idea - opinions?

[PRR]

Rossi, you have a private message which you apparently have not read?

 

[Rossi]

[quote author="PRR"]Rossi, you have a private message which you apparently have not read?[/quote]

I did get the message and sent you an email right away. Looks like it got lost somewhere across the atlantic. I just sent it again.

The circuit discussed here is a different solution for those adventurous enough to rewind the transformer. :grin:

 

[Rossi]

[quote author="Samuel Groner"]
Did you verify this with a suitable output load (e.g. 1k)? Not sure what current gain the phase splitter got but the load might get trough and increase this frequency.[/quote]

Just did a simulation with 1k (I usually use 2K): -3dB rises to from 11.4 to 13.5 Hz. 1u still gives 19.7 Hz.

[quote author="Samuel Groner"]

Would the 2SK170 still be superior to the 2N3819 as a source follower?

Noise wise for sure. Note however that you actually would like to have gain as early as possible. This would ask for a different output stage topology though. Personally I'd go back to the Dorsey circuit I think (miccircuit.gif). Or do you see a specific advantage of your circuit?[/quote]


I was a little curious about using a phase divider on the output. It seemed to be such a simple solution, but then again maybe there are reasons why it's rarely done.

So, maybe I should just go back to the Schoeps/Dorsey circuit. My initial mistake was using the hind part only, as I didn't need an excessively high input impedance. Around 100k would be enough. (Some transformer ringing wouldn't be so bad, as it would actually give some "free" treble boost, which you often need with ribbons anyway).

[quote author="Samuel Groner"]
And how did you measure the microphone noise? Looks like a relatively difficult thing to do.[/quote]

I just used my ears, made some recordings and compared to other mics. The difference was distinct. The mic's self noise was spec'd at 14 dB-A, which was about correct. Maybe my ears are more sensitive to noise than most peoples, I don't know, but I find that I can tell self noise with some precision by just listening and comparing to other microphones. Below 12 dB-A it becomes difficult, of course, and it's impossible to determine exact noise figures in the single digits using this method, due to room noise. Figures above 18 dB-A are really easy. Most passive ribbons are about 18 dB-A on an super low EIN preamp. Incidentally, I had a passive ribbon mic using pretty much the same ribbon motor to compare to.

[quote author="Samuel Groner"]

Yeah, maybe the 22 nF at the output are a little high, although, according to spice, the frequency response remains flat to several hundred kHz.

Frequency response is not the problem, but the capacitors present a load to the output stage, causing frequency-dependent distortion.[/quote]

Thanks, Samuel! I guess there's still a lot for me to learn. Not being educated in the field, I probably rely on simulation too much. I think the MXL version of the Schoeps/Dorsey circuit uses 22nF capacitors on the output. http://www.sdiy.org/oid/mics/MXL-603S.gif

 

[Rossi]

Another Question: Brad suggested to connect the FET gate directly to the transformer secondary. There shouldn't be much of an offset voltage, of course, but I'm a litte afraid, nonetheless, because of the high transformer ratio (backwards, now) and the fragility of the ribbon element. Could the ribbon be damaged when the mic is turned on? And could the miniscule DC current through the secondary be enough to magnetize the core in the long run?

 

[Samuel Groner]

With respect to noise I suggest to consider a few numbers here: if we ignore the DC resistance of the transformer and simply take the 4.5k Ohm figure you gave the thermal noise of the ribbon resistance translates to -116 dBu (20 kHz BW) at the transformer output. It should not be that difficult to design a buffer which has a low noise figure for that source impedance (the Dorsey circuit configured for at least 1 mA current per transistor and the BJTs replaced with low-noise types should do the trick, I think--perhaps the coupling caps should become larger to reduce 1/f noise). However, the DC resistance of the primary is very critical--even if it is one tenth of the ribbon resistance it will contribute about 1 dB. Neither can we ignore the DC resistance of the secondary--450 ohm is about the maximum which is acceptable. Unless you are very sure that your transformer is blameless I'd first check there.

Do you have the numbers for your transformer? It might be worth to determine the precise turns ratio as well as this will give us more information to optimise the active part.

Samuel

 

[Gus]

why not a BJT for Q? it is more predictable than a FET.

 

[Rossi]

Thanks for your input, Samuel. I know the transformer is critical (there was a related discussion in my previous ribbon tranny threads). I took great pains to reduce DCR on the primary. The DCR on the secondary is about 500 ohms. But there's only so much space on the bobbins, so you have to use really thin wire for the secondary. The transformer has two bobbins, each wound pri-sec-pri. The primaries use thick wire, all four primaries are wired in parallel, the two secondaries in series. I think it's about as good as it gets with the transformer core and bobbins of the original transformer. I'm thinking about simplifying the construction a bit to make it easier for readers to replicate, but I don't think I can improve DCR much, if at all.

Unfortunately there are factors I cannot change. The wires connecting the ribbon could be thicker wire. Unfortunately it's hard to measure resistences that low with any precision (at least with my multimeter) so I don't know how large a factor those resistances really are. If I were to build a ribbon from scratch, I'd be more scrupulous and use thick copper wire to connect the ribbon element and reduce line lengths. I actually tried to solder thicker wire to the top ribbon clamp, but it turned out to be too heat sensitive. There are some plastic parts that melt very easily, and sooner than expected the ribbon broke :sad: It was an experiment, and luckily I was able to safe the ribbon. But it's not something I can recommend.

So there are, indeed, limits set by the construction. Remember that this is a budget ribbon mic, not a Royer or AEA. Part of my article will actually be an investigation of the critical parameters of noise performance and frequency response. So I don't expect achieve spectacular noise figures. I just want to figure out how low I can get with this construcion. As I said, my schematic in the first post, even though flawed in some ways, gave slightly better noise performance than the same model with the stock transformer on a really low EIN preamp.

So, I don't have much in terms of numbers. Which has to do with the very nature of this experiment. The beauty of winding a transformer for an active ribbon is that you don't have to be very precise. You get a pretty good transformer by just reducing DCR on the primary and putting as much turns on the secondary as fit on the bobbin(s). Very doable with an improvised winding machine. You don't even need a counter. My tranny is pretty wound pretty sloppily except for the handwound primaries. Still, it sounds really good.

Of course, if this were not DIY but a commercial project, requirements would be different. But right now, I'm not trying to build the absolute best active ribbon, but the best that is attainable under primitive DIY conditions, at minimal cost, and without resorting to materials that may not be available to anyone. Basically just two spools of copper wire and a handful of electronic components and see how far we can get.

 

[Rossi]

[quote author="Gus"]why not a BJT for Q? it is more predictable than a FET.[/quote]

What kind of circuit were you thinking of?

Yesterday I tried a very simple solution with nothing but a BJT phase splitter. Same as in the revised schematic (without the FET part) but with higher values for R6 and R7. It works okay, but the FET circuit - even in my flawed verseion - is a little nicer and a little bit lower noise.

I think tonight I'll try the Schoeps/Dorsey circuit with just a lower value (say 470K) for the gate resistor. I'm still a bit scared to connect the gate directly.

 

[bcarso]

I doubt the FET could manage to clobber the ribbon---compare the motion of the ribbon when placed near a big drum versus what kind of energy you can back out from that circuit input.

Note that with the revised circuit you are still far from operating balanced Z. Using the polarity splitter directly as output is tricky, since an external disturbance on the "-" wire induces a signal into the collector and hence out on the "+" wire. If you model it as a couple of current generators on each line, the optimal value for rejecting induced noise is just under 2k in series with the emitter, higher than intuition might suggest.

Another approach which has the disadvantage of twice the noise power but other benefits: use two matched FETs (or a dual) and feed the gates from each side of the transformer. Refer the transformer to ground with good-sized Rs on each end. Couple to PNP emitter followers to drive the output lines. Send me a PM with your outside email addy and I'll show you an example, which you are welcome to reproduce.

The advantage is significantly reduced distortion. Also, the powerup behavior is symmetrical so any concerns about transients backing out of the input and tweaking the ribbon are pretty much laid to rest.

 

[Rossi]

So here are Brad's circuits. (Big thanks, Brad!). The first one is a high gain version which may be a little too hot for some applications; the second one is a lower gain version (requires more parts, though).

bcarso's high gain version

bcarso's lower gain version

I'm in a hurry to finish an assload of articles, so I may not be able to build one of those circuits till next week (I also need to order some 2N4391 for the second one). They look very tasty, though! :thumb: :thumb: :thumb:

 

[Samuel Groner]

Thanks for the circuits! Why not a BJT instead of the 2N4391s in the low gain version?

Samuel

 

[Rossi]

Brad sent me another schematic with output DC servo. According to Brad, the DC bias has a quite large temperature coefficient (which probably won't be a problem under studio conditions). The output DC servo compensates for that.

lower gain version with DC servo

 

[Rossi]

Oh, last night I experimented with the Schoeps circuit. I built a version similar to the stripped-down Dorsey version minus the parts not needed for a ribbon mic (capsule bias). I was a little scared to conect the gate directly so I used 100n in series and a 3.9 Meg resistor from gate to ground. Stuff that was a hand. Well, I made a silly mistake and ran almost full P48 through the ribbon tranny's secondary. Miraculously, both the ribbon and the transformer survived. The mic still sounds okay, seems like the transformer core doesn't get magnetized too easily. So at least this model appears to be more rugged than expected. I guess a little FET offset won't hurt too much :roll:

Initially I used the 2SK170BL. Noise was okay; I think my crappy-assed FET + BJT phase splitter was a little lower noise, though. I put in a bias adjust trim pot like in the Schoeps version (1 meg), but couldn't get any better results than without the trimmer. Then I used a 2N3819 instead and got slightly better results. The overall noise level may not have been lower, but there was less noise in the higher frequencies, where you hear it most. I've had excellent results with these 2N3819 before in condenser mics. They're Fairchilds, by the way.

Maybe I should try lower Rbb transistors in the output stage? I used BC560C. I have some 2SA1015, which I might try.

All in all, this stripped down Schoeps circuit looks like a good recommendation for beginners. Very easy to build and not hard to understand. Besides, I wrote about the Schoeps circuit in my previous article on condenser mic modifications.

But Brad's more elaborate circuits look very tasty. I simulated the high gain version, and the performance looks quite spectacular. :shock:

 

[bcarso]

The 4391s could be bipolars, certainly. I tend to grab FETs for cascode stages since often I'm using the built-in bias of the high pinchoff part and bootstrapping the lower part, so in some ways they are a remnant of that in these circuits. Also their gate current is negligible, so a relatively high-Z voltage divider can be used. But bipolars would be fine (and be easier to get, and ensure a tighter match of drain voltages for the FETs below). There would be a tiny increase in noise due to the shot and excess noise in base current.

The concern about damaging the ribbon is particularly moot in normal operation with the balanced topologies, since the two gates are intrinsically blocking any large differential currents at the input. As long as breakdown voltages aren't exceeded you can't get any but common-mode current to flow.

The principal purpose of the differential topology is to reduce distortion. As well, it is inherently balanced at the output, so you should get very good rejection of radiated and conducted noise with an average console's input. At the same time these benefits are gotten without overall feedback, which may have advantages in terms of lower high-order products. I haven't looked at IM in simulation yet and should do so.

There is of course a noise disadvantage of at least 3dB over a single FET, but the 2SK389/LSK389 parts are low enough voltage noise to have source thermal noise (let alone real studio background noise) almost certainly dominate.

Now, some may want the predominant second harmonic of the unbalanced circuits for coloration (right now for the FET models I have in sim, when you push things the distortion is mostly third). Note that one could fine-tune this by intentionally unbalancing the differential stage. The coupling cap between the two FETs allows this.

In principle for a well-matched pair, the cap could be eliminated and the resistors consolidated.

Both the low gain and high gain circuits have been adjusted so that signal-induced dissipation shifts are not only small, but actually about identical for each device, following a technique used many years ago in oscilloscope vertical amplifiers. The precise values of the drain load resistors depend on the FET properties, but roughly they should be set so the voltage is about half of the available rail. Now, these effects are subtle and for the low-gain version cut into the available input signal range, so it is a tradeoff. As well, the increased voltage swing means higher Miller effect and also larger capacitance modulation, itself a mechanism for higher distortion if the source impedance is high.

Note that, from a noise perspective, the resistors in series with the lower FET drains on the low-gain version do not add significantly to noise, since they are effectively in series with the relatively high drain output resistance.

I don't claim these circuits are new, although I haven't seen them before, and apologize if I am treading on someone's prior art. They were a good way to procrastinate on some other work I really didn't want to do. :wink:

EDIT: Looking over the servoed version I notice I didn't take the 18V zener tempco into account, and this actually dominates the new d.c. tempco since it runs about +15mV/degree C. There is an optimization to be done here. As I mentioned to Andreas, I wouldn't worry too much about the uncompensated drift unless the circuit was being used for field recordings.

 

[clintrubber]

[quote author="Rossi"]@ Peter: Don't worry, the big T recently got a new shipment. The new ones have "normal" ribbon transformers (about 1:30). For a while they sold a version with an "overwound" transformer (about 1:80) that could be used for an active ribbon. No rewinding necessary. But I like to think that my rewound transformer is a little better. :wink:[/quote]
See, that's exactly what I mean; stuff from T. usable for DIY & modding - but by the time the mag is on the streets the suited stuff is already gone again :wink:

Kidding aside, wasn't aware of a temp batch with higher winding-ratio. Sounds a bit surprising actually, didn't people with 'normal pre-amps' run into some problems ?
Do you have any idea if these overwounds are still available ? The coward I am w.r.t. mic-mods, I fancy the idea of keeping mic-changes minimized to changing the XLR-3 to say a XLR-5 (to avoid mistakes & to about halve the plug-contact resistance while at it) and then going to an in-cable-pita that's fed from P48.

Bye,

Peter

 

[Rossi]

I'm sure the higher ratio transformers are still out there. Somewhere. I have a feeling that the 600 ohms version that Alctron produces (aka The Mother of Chinese Ribbons) is in fact a 3k version. The DC resistence of the overwound transformer's secondary is about 600 ohms. I'm making assumtions, but I have a feeling the people who build those mics may not know the difference between DC resistence and impedance.

Out of the box, the 3k version sounds kinda weird when plugged into a normal pre, which of course you only realize with a regular 200 ohms model (which is more like 300 ohms) for comparison. However, the weirdness is not just due to the transformer, but in part also due to some internal "improvements". With a few modifications the 3k version sounds waaaayyyyy better. Night and day difference. Out of the box, the standard version sounds is superior, so it's actually a good thing they returned to the original version. Except for modding.

But you can buy a 1:110 transformer from Lundahl, which I'm sure sounds excellent. I only have the normal 1:40 Lundahl; the 1:110 was unavailable at the time.

If you're scared to open the mic, you can build PRR's ribbon booster (in the "old" active ribbon thread). That circuit requires exotic super low Rbb transistors, though. And by super low, I mean super low. Anything else won't do. That circuit is designed for normal low impedance ribbons and can be built into an external box or into the mic itself.

 

[clintrubber]

[quote author="Rossi"]I'm sure the higher ratio transformers are still out there. Somewhere. I have a feeling that the 600 ohms version that Alctron produces (aka The Mother of Chinese Ribbons) is in fact a 3k version. The DC resistence of the overwound transformer's secondary is about 600 ohms.[/quote]
Thanks for the info Andreas, I'll check what mine might be. (with impedance-meter on non-auto ranging; I'm hesitant to throw a DC-check at it)
We're talking all RB-500 here, correct ?

Out of the box, the 3k version sounds kinda weird when plugged into a normal pre, which of course you only realize with a regular 200 ohms model (which is more like 300 ohms) for comparison. However, the weirdness is not just due to the transformer, but in part also due to some internal "improvements". With a few modifications the 3k version sounds waaaayyyyy better. Night and day difference. Out of the box, the standard version sounds is superior, so it's actually a good thing they returned to the original version. Except for modding.

That's good info, thanks. Ah, sounds like non-electrical mods there.

If you're scared to open the mic, you can build PRR's ribbon booster (in the "old" active ribbon thread). That circuit requires exotic super low Rbb transistors, though. .

I recall that thread, I feel I'm fairly aquainted with the guy that started it :wink: Lots of interesting discussions there, I saw it's from two years ago already.

My main reason for avoiding modding the mic itself might be yet another open mic around :wink:
I understood your upcoming article will use the already present bobbin - that's nice. There's actually no real or further reason to be too scared from some mic-TX rewinding.

And by super low, I mean super low. Anything else won't do. That circuit is designed for normal low impedance ribbons and can be built into an external box or into the mic itself.

Hmm, I understand & recall that, but there's also a nagging voice that says: "but plugging in the stock mic into a normal pre seems OK enough..."
OK, let me first re-read that thread again.


Regards,

Peter

 

[Samuel Groner]

Thanks for the elaborations, Brad.

Then I used a 2N3819 instead and got slightly better results. The overall noise level may not have been lower, but there was less noise in the higher frequencies, where you hear it most. I've had excellent results with these 2N3819 before in condenser mics.

I would be interested to examine this further. Rossi, would you mind to send me some audio files? I'd need four of them (a pair for each FET configuration): one with a source at fixed level (e.g. pink noise from a loudspeaker) to calibrate the sensitivity of the microphone and one just with the self-noise of the microphone. 30 s each, 24 bit/96 kHz if possible.

Samuel

 

[Rossi]

Sorry Samuel, I have to write an assload of articles within the next few days, so there won't be time for that. I know how noise is measured, but I don't own a "Rauschbombe" (whatever that is in English) for highest isolation. But then again I wasn't interested in absolute values so much as in perceived performance. For comparison purposes I used an unmodified model plus PRR's ribbon booster circuit(using 2SC3329 transistors) plus a Siemens V272. That's so far the lowest noise configuration for real world gain levels (singer/speaker). The 2SK170 version had a little bit more HF noise than that, the 2N3819 version was just as good, maybe a little better than the above signal chain. I wasn't testing the 2SK170BL performance so much as I was figuring out, if I have to recommend it to my readers over the 2N3819 which is much easier to obtain. The current crop of 2N3819, at least the Fairchilds, are really good. I just ordered some more.

BTW. Michel Joly writes about his modified Oktavas that his version has less noise in the higher frequencies. According to his website he uses a FET that Neumann used, and I assume he is talking about the 2N3819. So far I that sounded weird to me as condensers have a falling noise curve anyway due to the capsule capacitance. But maybe there is something to his observations.

 

[clintrubber]

[quote author="Samuel Groner"]
[quote author="Rossi"]Yeah, maybe the 22 nF at the output are a little high, although, according to spice, the frequency response remains flat to several hundred kHz.[/quote]
Frequency response is not the problem, but the capacitors present a load to the output stage, causing frequency-dependent distortion.

Samuel[/quote]
FWIW this T.H.E.-circuit is quite hefty with the nanoFarads as well (C5, C7):

I guess a manufacturer has to go a bit more for 'sturdy under as many conditions as possible' - at the cost of a bit higher distortion.
In that schematic we also see R1,R4 to keep the EF's happy, as mentioned by Brad.

Regards,

Peter