Microphonics in VF14 and U47

[terry setter]

How big is your grid resistor? The original 60M or more?

I've used 1 Gig and higher in various versions, but for my "exact clone"I felt that going with 60M was microscopically closer to my Neumann U47's sound. Also, I use two UF4007 diodes in series to ground rather than a cathode resistor. This gives me 1V of bias, irrespective of plate voltage. This is FAR quieter than using a resistor and, in my opinion, is closer to the U47s bias circuit design (of forcing the cathode to a positive voltage). The UF4007 is super-quiet, but even typical signal diodes beat resistors for noise. Schottky diodes are the only other ones I've tested that were equally quiet, but they yield different bias voltage. Feel free to mix and match to get what ever bias voltage you want. Try it, you'll like it (remember to orient the diodes correctly - cathode(s) to ground).

 

[Cqwet Dbdfte]

IR LEDs give about 1.2V at their rated current.

 

[rock soderstrom]

I've used 1 Gig and higher in various versions, but for my "exact clone"I felt that going with 60M was microscopically closer to my Neumann U47's sound.

Thanks!

Also, I use two UF4007 diodes in series to ground rather than a cathode resistor. This gives me 1V of bias, irrespective of plate voltage. This is FAR quieter than using a resistor and, in my opinion, is closer to the U47s bias circuit design (of forcing the cathode to a positive voltage). The UF4007 is super-quiet, but even typical signal diodes beat resistors for noise. Schottky diodes are the only other ones I've tested that were equally quiet, but they yield different bias voltage. Feel free to mix and match to get what ever bias voltage you want. Try it, you'll like it (remember to orient the diodes correctly).

That's interesting, as I'm currently experimenting with it. The UF4007 has been my standard diode for some time now and I always use it when a 1N400X is required.

My results with diodes as a replacement for the cathode resistor are promising, not only in microphones, but in a direct comparison in a headphone amplifier, I'm not sure whether I didn't like the resistor better in terms of sound purely subjectively.

Of course, my tests were not scientifically blinded and therefore not particularly meaningful. I'll have to do it again and get it right. Unfortunately, the sense of hearing is very easy to influence, so you have to be careful not to fool yourself.

At this point, I would also be interested in other people's experiences with diodes at this point, what do you think?

Overall the Terry Setter U47 recipe is now complete, I'm currently waiting for an M7 style capsule, when that arrives I'll check it out. I'm excited and have already built up a small stock of EF184s from Telefunken, Mullard and Valvo, so I'm ready to go.

 

[Cqwet Dbdfte]

An LED replacing a resistor for cathode bias would negate any negative feedback, a "big" cap on the diode may not make much difference.

 

[terry setter]

Rock Solderstom: Very interesting to know that you have already begun researching the diode bias method. As I said (and you apparently have found) UF4007s are REALLY quiet, as are Schottkeys. Other types can get quiet a bit noisier. IR LED's will give you a reasonable DC bias voltage, but they're very noisy. DON'T USE LED's for biasing. WAY too noisy, and, in the event that light should happen to hit them, your bias level will be effected. This usually isn't very probable, but I'm just letting you know what I found. I have a heavily modded Dynaco St-70 that uses 7 UF4007s in series to bias the driver tubes. Works really well: super quiet, fabulous sound. For those new to this idea, remember to orient your diodes with their cathode to ground (the end with the linen drawn on it).

Cqwet Dbdfte: You are correct. There is no need for a bypass cap on the cathode when using diodes to set your bias. You get all the gain, a very wide and flat FR, and have at least one less part in the circuit. You also get around the issue of low frequency phase delay that would occur if you were using a cathode cap, though if your mic's design is a good one, that would be happening at such low freqs as to not require much worry. That said, I know some pretty dang good builders and modders who do worry about it.

Final note for Rodk Solderstrom: If you can, use the Cinemag CM2461 NiCo (6.5:1) output transformer. You will be very pleased!!! I compared it with many contenders, including BV08 clones from from Andreas Grosser, Crimson, AMI, etc. and it won first prize, hands down.

 

[rock soderstrom]

If you can, use the Cinemag CM2461 NiCo (6.5:1) output transformer. You will be very pleased!!! I compared it with many condendors, including BV08 clones from from Andreas Grosser, Crimson, AMI, etc. and it won first prize, hands down.

I know two mic builders here who also swear by the Cinemag. I'll see if I can find one somewhere on my side of the pond. Shipping from the USA always makes things a bit more expensive, unfortunately.

 

[Cqwet Dbdfte]

Rock Solderstom: Very interesting to know that you have already begun researching the diode bias method. As I said (and you apparently have found) UF4007s are REALLY quiet, as are Schottkeys. Other types can get quiet a bit noisier. IR LED's will give you a reasonable DC bias voltage, but they're very noisy. DON'T USE LED's for biasing. WAY too noisy, and, in the event that light should happen to hit them, your bias level will be effected. This usually isn't very probable, but I'm just letting you know what I found. I have a heavily modded Dynaco St-70 that uses 7 UF4007s in series to bias the driver tubes. Works really well: super quiet, fabulous sound. For those new to this idea, remember to orient your diodes with their cathode to ground (the end with the linen drawn on it).

Cqwet Dbdfte: You are correct. There is no need for a bypass cap on the cathode when using diodes to set your bias. You get all the gain, a very wide and flat FR, and have at least one less part in the circuit. You also get around the issue of low frequency phase delay that would occur if you were using a cathode cap, though if your mic's design is a good one, that would be happening at such low freqs as to not require much worry. That said, I know some pretty dang good builders and modders who do worry about it.

Final note for Rodk Solderstrom: If you can, use the Cinemag CM2461 NiCo (6.5:1) output transformer. You will be very pleased!!! I compared it with many contenders, including BV08 clones from from Andreas Grosser, Crimson, AMI, etc. and it won first prize, hands down.

Did you measure the difference in noise between the 4007 and an IR LED?
How much current thru your diode(s)?

 

[moamps]

Using different types of diodes (standard, LED, IR) or small button batteries for grid biasing has interesting advantages that have already been mentioned here. However, in the standard version of the U47, where the heating current actually generates the grid bias, these methods may be questionable. If an unregulated power supply is used, it is expected that its output voltage will change during operation, which may also have an impact on the anode current. This change is undesirable and will be less if for a grid bias cathode resistor is used. That's why I would still use a good cathode resistor in the case of the classic U47.

 

[terry setter]

Did you measure the difference in noise between the 4007 and an IR LED?
How much current thru your diode(s)?

Yes, I did. And I used more than one example of each. One interesting thing is that the color of the LED directly relates to the resulting voltage. The higher the frequency of the color, the higher resulting voltage.

The diodes can handle far more current than I even need to pass through them for microphone usage, and, as noted above, I used 7 UF4007s to bias the driver tubes in my tube power amp. No current-related issues ever surfaced, and I've had no failures to date.

 

[terry setter]

Using different types of diodes (standard, LED, IR) or small button batteries for grid biasing has interesting advantages that have already been mentioned here. However, in the standard version of the U47, where the heating current actually generates the grid bias, these methods may be questionable. If an unregulated power supply is used, it is expected that its output voltage will change during operation, which may also have an impact on the anode current. This change is undesirable and will be less if for a grid bias cathode resistor is used. That's why I would still use a good cathode resistor in the case of the classic U47.

Moamps: I appreciate that you know U47s through and through, so I take your opinion seriously. When I set out to "clone" my U47, I wanted to stay as close to the original circuit topology as possible, but I found that using tubes with relatively low heater voltages led to significant complications when trying to implement Neumann's bias scheme. My attempts to get their heater/bias arrangement to work gave me enough problems that I felt I had to find a good work-around. In the end, I found that I could achieve the sonic output- both subjectively and according to distortion measurements up to the 7th harmonic - that were closer to my 1957 Telefunken-branded U47 by using the diode biasing technique. I tried over 50 different types of diodes, as well as diodes + resistors, and diodes with and without resistors + a bypass cap, etc. I finally settled on using (two UF4007) diodes between cathode and ground. The Neumann circuit only uses a 29 Ohm resistor, so the resistor, itself, is quite innocent, noise-wise. But successfully applying a portion of the heater voltage directly to the cathode kept throwing me curve balls (even though I always use tightly regulated heater voltages in my mic designs). In fact, I think that Oliver Archut's "clone" that uses two tubes with 19V heaters wired in series (thus being able to accept 38V through a resistor from the B+ supply) was the work-around that he came up with so that he could keep the Neumann bias design. And I believe that David Bock ended up using a voltage divider on the full B+ to try and implement Neumann's approach. Of course a microphone is a synergistic system, so it wasn't until I found the right tube, transformer, caps, diodes, resistors, and topology combination that I was able to get my distortion measurements to within less than 2dB across those first seven harmonics.

I have a matched pair of Blueline M7s from Thiersch and when I put one of them into my clone and one into my real U47, none of my engineer friends was able to distinguish one mic from the other. I take this to mean that I have the circuit pretty well dialed-in. Do my Thiersch M7s sound like an original Neumann M7? Well... they're the closest thing I've ever been able to find and, to my tastes, they are as close to original M7's (in reasonable working condition) as any two vintage M7 capsules are to one another. So I've been very pleased with the results of my particular combination of components and design. I can't hope for much more than that.

A closing thought about the future: All the original M7 capsules that Neumann made are now more than 60 years old. They're made with PVC diaphragms, and PVC just doesn't have that kind of service life. Many of them have "died" as a result of stiffening and/or loss of their gold sputtering. The rest are on their way to the grave, and I am told by one highly reputed repair person that there has been a sharp increase in M7 failures, due to age rather than abuse, over the past ten years or so. This means that it will not be very long before there will be NO examples of original M7 capsules that are working properly in ANY microphone. We'll be reduced to using recordings as our only basis for comparison, and young mic builders will never have heard dirrect, live output from a working mic with an original M7 capsule in it. Most of the younger engineers I know (below 35 years of age) have never held an original, working U47 or U49 in their hand, let alone used such a mic over a long period of time, so that they could really get to know what it was capable (and NOT capable) of doing. I fear that we are headed into a time when romantacized hype will be slathered over all new high end microphone designs as being the one that's able to "capture the magic" of those original mics. This will drive me nutz! But the saddest part will be that disproving (or supporting) such a claim will eventually become impossible to do. The advertising spin-doctors will have no restraints, and the people who knew those original mics will eventually be gone as well. Wow, I guess I'm feeling particularly dark today. My apologies in advance.

Moamps, it's guys like you that have done a great deal to keep us in a world where wild claims have to be proven and substantiated. Thanks again for your service to the industry (and for publishing the actual plate resistance of a VF14!).

 

[moamps]

Hi Terry,

thank you very much for such a thorough and detailed post. It should be put somewhere on the forum as a sticky because it actually contains so many good ideas. From the analysis of the original design, from the courage to try to reinterpret it with modern parts and available vacuum tubes using scientific methods and auditory observations at the same time, to facing the reality that if we haven't already, we will lose the benchmark or the beginning.

I'm glad if I was a very, very small grain of that very nice story about the U47, and yes, it's a shame that many engineers won't have that opportunity. And I make myself look a bit stupid when I excitedly talk to students about the history and facts of the U47, and they ask me later "which microphone with a USB connector is better".

Thank you very much for all the information and knowledge you share here on the forum (the idea of using the EF184 is very interesting), and I will continue to read your posts with interest in the future.

May the force be with you.

Milan

 

[Cqwet Dbdfte]

Yes, I did. And I used more than one example of each.

How did you measure the noise from the diodes?
I did, with a 'scope, 6-1/2 digit DMM, and a Fluke 8050A.
Do you have any numbers?

 

[terry setter]

How did you measure the noise from the diodes?
I did, with a 'scope, 6-1/2 digit DMM, and a Fluke 8050A.
Do you have any numbers?

I used my Fluke meter, a Tektronics scope, and Pro Tools recording software. Bearing in mind that my personal notes were not intended to publication, and against my better judgement, I'll attach some pages of my measurement notes that will help you see part of my process that led to me settling on the diode biasing scheme. I found the Pro Tools files to be very helpful. I could hear the character of the noise, as well as see it. I also came to use a playback emphasis scheme where I would add +20db at 100HZ and +20dB at 10KHZ to the recorded files (the output of the mic circuit being tested, fed from a 50pF cap rather than a working mic capsule). When it was hard to differentiate between the noise of various diodes or resistors on the cathode, this high and low emphasis on the recordings removed all doubt. With the emphasis added, the output signal from the circuit being tested would be well up into the levels where Pro Tools metering could reliably differentiate by 1/10 of a dB. Of course, there were some diodes that were very similar, but there were only a few that were close to identical among the quietest ones I tested. Again, the winners for me were the UF4007 and the Schottkys. I also liked the BA159 quite a bit. I use the UF4007s because they're small, inexpensive, and rated for one Amp at 1000V. Nothing is going to hurt them!

Near the middle of the photo called "Discovery of LED Noise", you'll see the line:
"Output of UV LED is .2dB hotter than Rk = 685, but it's also 1.8dB NOISIER!?!"
I don't know how to rotate the photo so that you can see the entire page. But this was the moment I saw that although the signal output of my circuit was .2dB hotter when using a UV LED on the cathode to set the bias rather than a 685 Ohm resistor, the resulting, total signal-to-noise at the output decreased by 1.8 dB, for an overall loss of 1.6 dB when using the LED. Subsequent experiments confirmed that the LEDs were NEVER contenders for low noise. You can see some of those tests in progress further up that same page.

In the photo called "Relative Noise Levels", you will see circled numbers that represent the S/N of various biasing schemes, including several types of diodes and various cathode resistor values. (Overall one "Cree" Schottky diode yields .8V of bias voltage, one UF4007 yields .5V of bias voltage.) You can see that two 4007s gave me 1.04V of bias and a resulting S/N of better than 60dB, whereas a 681 Ohm cathode resistor gave me 1V of bias, but only 58.7 dB S/N. So, for the same bias voltage in an otherwise identical circuit, two UF4007s out performed the resistor by more than 1.3dB. [Remember, this is with +20dB emphasis at 100 and 10K HZ so when I say "60dB", that is not the measurement of the finished microphone's self-noise. The important consideration was the relative S/N number, not whether it was high or low as microphone circuits go.] Also, I tried adding bypass caps (big, small, different dielectric materials, etc.) and found they made no difference when using diodes, and fairly insignificant change when the cathode resistor value was quite low.

In the photo called "Some of the tested diodes", you'll see thirteen of the aproximately 50 diode types that I tested.

A long-winded answer, I realize. Sorry for that. I've got a large spiral bound notebook full of pages like these, all of which led me to the recommendations I mentioned in my previous comments on this thread, including what transformer, tube, types and values of of resistors and caps, B+ voltage, and heater voltage eventually brough my circuit closest to the performance of the circuit in my 1957 Telefunken (badged) long-body, chrome cap U47. Thankfully, I was on sabbatical and could give my full attention to this project. Otherwise, I could never have gotten so far into the weeds with all the testing and experimentation. But I must say, I learned more about how things actually work in tube mic circuits by doing this than in all the other mic design and repair I've done over the years. I hope you'll get a chance to do the same someday, if you haven't yet.

 

[Cqwet Dbdfte]

Thanks, you put in a lot of work!


I would disregard the noise data of the UV-LED, as the Vf, which like the blue, white, some of the green LEDs, gives a much higher than the bias voltage needed for the circuit. Your notes suggest that 681 Ohm Rk gives about 1V so should also any diode.

With a UV-LED the bias voltage get up to around 3V and and raises the plate voltage almost 100% and put the tube close to the B+, so a non-starter to use anything that puts the plate voltage away from design center, as you would lose dynamic range. OK maybe for measuring noise signals, but large signals would clip.

Diode bias with its low dynamic impedance would kill cathode degeneration and increase gain, versus an un-bypassed cathode resistor.

To evaluate an LED for the application it should have the desired Vf, in this case an IR LED could work, check the data sheets for suitable types. Digikey has good filters for selecting parts, including "Vf" for LEDS. Your circuits looks like it draws about 1mA thru the plate. The LTE-4208 looks like it could work, like 1.1V at 1mA, data a bit fuzzy at the bottom, may perhaps need a another mA to get it started in its linear range, where according to the slope of the curve would have an impedance of ~7Ohm providing very little negative feedback. This could be bypassed with a "big" electrolytic, but as the slope is very linear, whatever microvolts of feedback would have very little effect and can be disregarded. The original Neumann circuit ran the heater current (40mA) thru the cathode resistor thus reducing plate current (0.55mA) influence 38dB, making the need for a bypass capacitor small.

Using 1A, 1kV, fast switching diodes in this position is not necessary for a 1mA, 1V, DC only circuit, but wont hurt either.

If light inside the enclosure is an issue, an LED can be covered up with some black heathshrink.

I ran the noise test with two UF4007s in lieu of a IR-LED in a 45mA/V very low noise tube (gain about 80) used for moving coil amplification, which draws 20mA from a 185V source, (not necessarily a top choice inside a U47 body unless some neat heat sinking is used, 3W vs. 5.7W ).

I could not measure any advantage of the UF4007 pair with the 6.5 digit 300KHz DMM. My tube is fed by a constant current source which additionally suppresses any current variation in the tube. The current noise in the diode is converted to voltage over the diodes dynamic impedance in the tube circuit.

Signal to noise is important, is your noise measurement below the noise in a quiet room? Any signal would bury the noise and only be revealed during no signal passages.

Very few of Fluke's handheld electrician grade DMM's have enough bandwidth to cover the audible range. The old and very slow Fluke benchtop 8050A does cover it up to 100KHz. Most DMM's give up at 1KHz. The top end handheld 87V MAX DMM barely makes 20KHz at below 0.6V.

While a diode junction will generate broadband noise also in the forward biased direction, not much of it is relevant to the audio spectrum in this application, quoting Mr. JR.

https://groupdiy.com/threads/do-leds-generate-audible-noise.82153/
Walt Jung's Directory listing <-- GLED431 ~1nV/sqrt Hz

Using diode(s) looks like you got 2dB lower noise over (unbypassed?) 681 Ohm resistor.

 

[terry setter]

... I would disregard the noise data of the UV-LED, as the Vf, which like the blue, white, some of the green LEDs, gives a much higher than the bias voltage needed for the circuit. Your notes suggest that 681 Ohm Rk gives about 1V so should also any diode.

Yes, red LEDs were the only one that were very close to my target bias voltage. But again, they were noisy in comparison to the UF4007 and Schottky diodes when selected to achieve the same bias voltage in my particular application.

Diode bias with its low dynamic impedance would kill cathode degeneration and increase gain, versus an un-bypassed cathode resistor.

So true!

Using 1A, 1kV, fast switching diodes in this position is not necessary for a 1mA, 1V, DC only circuit, but wont hurt either.

Being as my favorite sounding part was also built like a tank, I got even greater pleasure from the fact that I bought them in bulk for $.07 each! I never worry about a UF4007 failing on the cathode of a tube in one of my microphones. Peace of mind for under a dime...!

If light inside the enclosure is an issue, an LED can be covered up with some black heathshrink.

I used black electrical tape for my experiments.

Signal to noise is important, is your noise measurement below the noise in a quiet room? Any signal would bury the noise and only be revealed during no signal passages.

"Noise" can be a bit of a slippery fish. Do my 24-bit recordings sound better than my 16-bit recordings, even though there are no moments of complete silence anywhere within the piece of music? All else being equal (which is a big topic in itself) I think they do, so I continue to push for very low noise when possible.

I make my noise measurements without a working capsule, and I feel that even though the S/N of my completed microphones puts them well below the noise level of even very quiet rooms, I usually go for the least possible total noise in my entire recording and playback system. Most of us are familiar with emmulation plug-ins that have a "noise button", so that the user can decide whether they like it when it is as noisy as the original piece of gear or as quiet as current technology can make it [I tend to not push the button]. And some people have argued that reducing the noise in a classic microphone works against the "classic" nature of its sonic character. But in my studio recording life, we ALWAYS selected for the quietest tubes and were always happy to find one that was exceptionally quiet, and we felt that we "got the best sound" that way. Also, I know Grammy-winning engineers who swapped out their stock plate resistors when newer, quieter types became available. I'm a big fan of Vishay "Bulk Metal Foil" resistors for use as plate resistors, but they are VERY expensive. In my clone, I used a high grade carbon resistor from a Japanese manufacturer. I thought it gave the mic just a tiny, tiny bit more of an "authentic" sound (I don't think those resistors are available anymore). That is to say, I do try to fit the type of part I use to the specific nature of its application. And, admittedly, there were times when I would measure and measure, and listen and listen to some fairly subtle varient in my designs and NOT be able to come to a definitive conclusion as to which version sounded best. In such a case I made my final design choise with respect to multiple factors: what was done in "the classics", which configuration is easier and/or cheaper to build, which takes up less space in the mic, which component choice might be more reliable in the long run? etc. Half of what I love about reading the DIY forum is seeing how other people answered those same kinds of questions. Our challenges are like jewels - so many facets.

 

[Khron]

I don't suppose you had any "plain" 1N4007's to test for noise, did you? Wonder how they compare to the UF's...

 

[terry setter]

I don't suppose you had any "plain" 1N4007's to test for noise, did you? Wonder how they compare to the UF's...

I'd love to be able to say exactly what I found. I know that I did test a 1N4007, but, relative to the 1N4007 I only have a note-to-self saying that in my tests, the 1N group did not do as well as the UF group. I have more specific notes on the 1N4148, the 1N4933G, the 1N4934, the 1N4935, and the 1N4937. All these were recommended as being good by various folks I was reading on the internet at the time (2016), but they did not prevail against the UF4007, BA159, or Schottkys.