Transformerless Tube Input Discussion

[analag]

I present my idea of a Fet/Tube Input mic pre, this should be quieter than the one I'm presently using. I hope to slap it together on a breadboard soon and disclose what fixes were necessary to get that sound we like to hear.

Analag

 

[pstamler]

The 15k resistors at the input will be a problem. Calculating all the parallel resistances, the input/feedback network will add about 22dB of noise to the basic noise level of a 150 ohm microphone, plus whatever noise the FETs contribute. A good mic preamp should add only a few dB.

Peace,
Paul

 

[Winston OBoogie]

...

 

[analag]

Sorry bout that. The 15k should have been 1.5k that was an error. I will correct and repost, later tonight. Thanks for spotting that. As suggestions are made I hope to adjust the schematic to show the improvements.......... The schematic has been updated to reflect some of the ideas that were presented to me. Yes the B- voltage could be lower, in fact the one I have, uses only 50V, but then I would have to CCS the 12AU7 in a similar fashion, sonically less pleasing but cheaper I guess.

Analag

 

[pstamler]

[quote author="analag"]Sorry bout that. The 15k should have been 1.5k that was an error. I will correct and repost, later tonight. [/quote]

That'll still have a noise figure of about 13dB -- in other words, it will add 13dB to the inherent noise of the microphone, still a lot. And that's not including the noise from the FET. It'll work with a very sensitive microphone on a loud instrument, but for any lower input voltage, you'll get lots of noise.

Using this "inverting-amp"" type of design, with resistors in series with the input, will always be noisy; the resistors need to be high enough to not load the microphone excessively, and that makes them high enough that they generate too much noise.

Peace,
Paul

 

[PRR]

Return feedback to the input stage Sources. Since your output stage is beefy, the feedback resistors could be reduced to 22K. Then for the same gain, noise resistance would be 750Ω instead of 3K, 6dB lower noise. Also you could now use Gate resistors around 1Meg and go to 0.05uFd input caps instead of those awkward 22uFd.

The FETs add a lovely amount of gain. But as noted, you may be squandering their low-noise feature. With 1.5K input resistors, you could get the same noise with just tubes, no sand-state. 6DJ8 input with 12AU7 output would give open-loop gain around 200, ample to get your closed-loop gain of 55 while not sucking all the life out of the tubes.

> a noise figure of about 13dB -- in other words, it will add 13dB to the inherent noise of the microphone,

Noise voltage of AKG 414 (other big condenser mikes are similar) is 1uV, or about 14dB above the self-noise of a 150Ω resistor or dynamic mike. So in that now-common application, this design's NF is a small issue. As a dynamic mike amp: I have actually used something similar, and it hissed, but the gig went on no problem. Most mikes are quieter than most rooms.

 

[analag]

Let's see that schematic, Kubi. I am spicing up your suggestions PRR.
Hope to reflect the changes, soon.

Analag

 

[clintrubber]

Some Tek-122-stuff here: schematic - Front Pic

In its original form gain is 100 or 1000. I haven't adapted these boxes for mic-pre-use yet, but Marik has - and reported good results. Bit I thought he added an input-TX.

Regards,

Peter

 

[analag]

The above schematic has been revised per PRR's suggestions. Unfortunately I couldn't go below 10uF caps at the input and get the low frequency performance I am looking for. However dropping the feedback resistors value to 22k shows a marked improvement in noise, bandwidth, low impedance matching capability. Kicking out the cascode will improve PSRR too.

That Tek122 pre seems interesting, is it really a mic pre? I wonder why they used those super high resistors in the grid circuit.

Analag

 

[NewYorkDave]

It's not a mic pre. It's a general-purpose, high-gain laboratory amplifier. A typical application would be to amplify a very weak signal for input to an oscilloscope.

But with a suitable input transformer, it would probably make a fine mic pre if you needed tons of gain. Or you could use it as-is for a DI.

 

[clintrubber]

from NYD:

it would probably make a fine mic pre if you needed tons of gain.

Yes, the gain is a bit clumsy in the original version. I don't know what Marik did, I thought he used a pot somewhere.

I'm not sure about the advantages of this circuit over the usual non-balanced ones w.r.t. sound, being balanced etc, so some 2nd order cancelling - but I've been wondering that before, I should just try.

Or you could use it as-is for a DI.

That'd be a nice application. Have been thinking about using both those balanced inputs (I have 2 of such boxes) for Fender-Jazz-PU's, pre-blend - but that's a bit too much of non-standard wiring - where that bass goes those 122's won't always be.

I also remember Jakobs suggestion for this circuit: as an opamp for an active EQ.

Regards,

Peter

 

[analag]

It might be a lot of work finding use for those without modding it, and then it becomes something else, like the schematic above perhaps.

Analag

 

[FredForssell]

http://home.hetnet.nl/~chickennerdpig/ FILES/Tek122/Tek_Type122_circuit_ampsection_rt.jpg

Yeow!

Man, are they using low current operating points for the front-end stages.

 

[rafafredd]

So, the inevitable question...

How many triodes would one need to parallel to get a comparable or close enought noise figure (using low impedance mics) of an input stage with a 1:5 input transformer?

I was thinking about trying some three triodes compactrons. All paralell, two tubes, three triodes per side... But I would also like to know if anyone knows about some low-noise three-triodes in one envelope tubes. Maybe I will have to use two compactrons, six triodes per side, for a real all tube transformerless low-noise mic preamp. Come on, I think it´s worth it... After I´ve seen that ridiculous tube clock with some 100+ tubes, I thought "If they do it for a clock, why wouldn´t we do it for a micpreamp?"...

For the outputs, two triode-pentodes in one envelope (like ecl82,84,86 and many others) in a mu-stage config for low output impedance. So, it would be really all-diferencial-balanced-tube-only-low-noise-as-possible-transformerless-mic-preamp. :green:

Have a look here for the output stage:

http://www7.taosnet.com/f10/mustage.html

But in the end remains the question. Would it really sound that diferent from a transformer coupled mic preamp using high quality transparent transformers (read LUNDAHL)?

Have to build to find out...

 

[bcarso]

As I mentioned somewhere the 6C45 seems to show about 3nV/rt Hz midband. With a 1:5 trafo that's an input-referred noise that's way low. OTOH to get balanced without trafo it's a job, as you're doing both sides. With two tubes you are up to 4.2nV and thus need about 4 per side to be comparable to the thermal noise in a 150 ohm R.

Also as mentioned the C45 samples I looked at had a nasty microphonic ringing with a mechanical Q of about 3000 at around 9kHz. Peter Qvortrup of Audio Note remarked to a friend that they sounded "metallic" so maybe he was hearing the ringing.

 

[FredForssell]

Also as mentioned the C45 samples I looked at had a nasty microphonic ringing with a mechanical Q of about 3000 at around 9kHz. Peter Qvortrup of Audio Note remarked to a friend that they sounded "metallic" so maybe he was hearing the ringing.

I had exactly the same problem using 12AT7 and 6072A tubes in my early transformerless mike preamp designs. The center freq shift around a bit, and I'm not sure the Q was 3000. MU follower configuration SEEMED to have a greater tendency toward this behaviour. All in all, very frustrating.

Cheers

 

[bcarso]

I calculated that Q from the decay of the oscillation envelope after tapping the tube. It was brutal ;-). Reminds me of the gag in Spinal Tap when he talks about the sustain on the one guitar. If only it were at least about an octave and a half higher---but then you would get IM, since presumably there's some gain modulation too, so still not good.

 

[clintrubber]

from FredForssell:

Quote:
http://home.hetnet.nl/~chickennerdpig/ FILES/Tek122/Tek_Type122_circuit_ampsection_rt.jpg

Yeow!

Man, are they using low current operating points for the front-end stages.

You're right, it doesn't seem to exceed a few tenths of mA's per side. Hadn't realized that before - I mean, OK, some resistor to a negative voltage for a simple something i.s.o. a 'real' currentsource, but had ignored the actual figures so far. Hmm, what could they be aiming for with biasing like this ?

I guess it's something apart from (unrelated to) the 10M Rg, correct ? Grid leak bias, of which RDH4 doesn't mention too many usable benefits on page 489 for intended use as a micpre, but I assume they originally did that to get the high Rin (ignoring the cap input capacitance).

Regards,

Peter

 

[PRR]

> I was thinking about trying some three triodes compactrons.

One good TV Tuner tube will beat a dozen of the utility triodes they put in Compactrons.

If you need three triodes to do three things, use Compactron (though a couple dual-triode bottles is often a better plan).

If you need one FAT triode to do one thing, get a fat triode. And the TV Tuner tubes were made for that work.

6BQ7 is the ancestral twin-triode tuner tube. One in differential mode can get you to 1K2 noise impedance.

6DJ8 is a TV Tuner tube and will do around 600 ohms noise resistance.

TV Tuners could use transformers so a very-low noise resistance was not essential. Also TV sets were very price-constrained. Aside from tube cost, low noise resistance implies high operating current and clean power supplies cost money which eats profit.

The next step up is microwave and radar tubes. AT&T, USA, USSR didn't whine so much about cost when 1dB better noise meant seeing the bombers 10% sooner, getting defense up a minute faster. (At AT&T, 1dB better noise meant putting repeaters 10% further apart, which adds up over a long haul.)

6c45pi is very high Gm but a single. You might consider 6H30, which is a twin-triode, each half roughly like half a 6C45pi. But the heater power for the transconductance is better with 6c45pi, and 6C45pi has higher Mu.

Taking 6C45: Gm is about 45mmhos at 40mA. Re is about 1/45mmho= 22 ohms, noise of a hot cathode is about 3 times higher or 60Ω, two tubes is 120Ω. Noise figure re:150Ω about 2.5dB. For good voltage gain you are looking at 24 watts of B+, 18 watts of heater. Quite a beast. For 1dB noise figure, you want about 8 pairs: 200 watts B+, 144 watts heater. A monster. You gotta ask yerself: do I need EVERY last dB of noise-figure? Often you don't, or not at that price.

High-Gm tube info

 

[FredForssell]

I calculated that Q from the decay of the oscillation envelope after tapping the tube. It was brutal ;-).

Ah, well you had more success at it than I did. I just watched in on Stanford Systems Real Time Spectrum Analyzer. I found that tapping with a really list striker ( I used a piece of shrink tubing) help get it going.

I guess it's something apart from (unrelated to) the 10M Rg, correct ?

Yeah, I think they may have had trouble with space charge with those high values resistors. I'm kind of surprised they didn't use a bootstrapped cathode-follower to feed the diff-amp to take care of the miller-cap and provide really high input Z. But then again they could have used (and did later I think) a cascode diff-amp to take care of the miller-cap.

Don't get me wrong... Tek was an amazingly good company back in its day. Even though I still use Tek scopes, they kind of lost me about the time the 465 scopes came out. Their best, imo, was the 647a and the 453 type scope. I digress... sorry.