Valve mic preamp design incoherent rambling

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

GussyLoveridge

Well-known member
GDIY Supporter
Joined
Mar 15, 2012
Messages
116
Location
Cape Breton, Nova Scotia, Canada
Hi friends -

I'm going to be making a whole pile of notes here, kind of mostly for me to talk things out...out loud.

It's kind of long winded so I get it if you'd rather move on. Again. If you have some insight you'd like to share, I would seriously appreciate it. I'm in no rush - I figure I might live another 40 years or so I have time.

Designing valve based microphone amps

I am interested in designing some simple microphone amplifiers using tubes. I have a fairly decent knowledge of guitar and bass amplifier circuits and have been regularly studying, repairing and modifying guitar amps for about 8 years or so. I’m no genius but I can more or less understand how all the components work and generally how stages interact on a very basic level. I feel confident that I could copy a design and layout and wire someone else's design now, but I'm just looking to learn, progress and have fun.

I have been studying designing tube based guitar and bass amps by Merlin Blencowe for quite some time and am fairly familiar with his writings. But between that and Small Signal Audio Design by Douglas Self, I’m at a bit of a loss for where to turn. I have studied NYDave’s one bottle and Mila and checked out some of the Gyraf stuff including the dual micamp as well as looked at some classic circuits including the Pultec MB-1 and have been following some of the other threads here as well like Professor Dan’s. Of course I’m pulling knowledge from forum members PRR, abbey road d enfer, ruff records and Winston OBoogie who tend to dig in on the valve design stuff. Among others of course.

So some of the things I don’t understand (I have zero formal training) I think are pretty basic but I’m not entirely sure of even how to ask them.

For example, I more or less understand that we want reasonably high input impedances and also wish to have output impedances as low as we can really. At least I think those things to be true. There are somethings about creating oscillations that give some real limitations and the fact that we can’t have infinite input impedance and zero output impedance. I don’t think so anyway. I don’t really understand what kind of signal we are going to see at the output of the microphone and how much signal we should be trying to aim for at the output? Are we just looking for +4dBu (3.472 Vpp)? There is only some much head room in whatever is following the preamp be it an external EQ, compressor, desk, tape machine, converters, etc. so we don't need to amplify more than that but we do need enough gain for low output mics and quiet signals. Would all of these devices look for the same (nominal?) level at their inputs? Would most of them have approximately the same input impedance? I am using fairly high end gear at my studio limited now perhaps by my 003 interface which will likely be replaced in the coming 2 years (Probably long before I finish this project)

So I guess starting from the beginning, we are going to have an XLR input, positive voltage on pin 2 with respect to pin 3. That’s a good standard. I think I can figure out getting phantom power there, there are enough reasonable examples of how to do that, I can also figure out getting a pad there, no problem. I could easily figure out how to get a polarity switch here too, but I’m not sure I would do that here or at the output? I’m not sure why you would choose one over the other…? Anyway most of this stuff is covered by the Jensen papers. No big deal. So what’s next - we need to get to the first gain stage.

Here’s where I think I’m really running off the rails.

Okay so let’s say I want to have a transformer input…if I look at one version of NYDave’s One Bottle there is a 150:15K which I think would step up my voltage by a factor of 100 (Is that right? That seems like a massive step up? - Surely I'm missing something there, should I be taking the square root of this?) and reduce the current from by the same factor? On another schematic of that circuit I see a different transformer spec’d which is 200:50K upping my voltage by a factor of 250 (again - this seems like I'm wrong - same as above?) and reducing my current by the same amount? The Gyraf Gyratec IX shows an input transformer with a 1:6.45 ratio which only would step up the voltage by a factor of 6.45 right? These are good right? (Probably, I suspect that these things are well designed :) ) The microphone output would be relatively small and the reduction of current for gain in voltage is good. I want to send a fluctuating voltage signal to the grid. But why such a difference between the Gyratec and the one bottles? Because the one bottle wants to have a similar amount of gain as the Gyratec but with half the gain stages, so we get a boost up front by the transformer? It also gives us a little bonus with the signal to noise ration as well?

NOISE: So one reason we might add an input transformer is because the tube itself will have some noise - that noise we can’t change I don’t think. We can step up the voltage of the signal, but we are also stepping up the noise from the microphone too - but because we’re stepping both up by the same amount, the S/N ratio of the mic itself stays the same and the noise from the tube (when added) will account for a smaller percentage of that noise? Am I getting that right?

I’m going to pull a few numbers here from space…bear with me.

Transformerless:
1 mV of signal from the mic
1 µV of noise from the mic
5 µV of noise from the tube

So if the tube has a gain of 100. We would have 100 mV of signal and 105µV of noise at the output?

So the tube noise is 4.67% of the total noise.

w/ transformer (1:6.45)

1 mV of signal from the mic via transformer gives us: 6.45 mV
1 µV of noise from the mic via transformer gives us: 6.45 µV
5 µV of noise from the tube

So again with a gain of 100 we'd get 645 mV of signal and 650 µV of noise (yes - I do know these numbers are whack - but is this at least correct in theory?)

And here the tube noise is 0.7 %?

I still don’t understand how what the signal from the mic looks like. I can look at the sensitivity of a microphone, for instance my SM7B which seems to be a somewhat popular microphone, relatively inexpensive and somewhat of a low output based strictly on a comparison to my other microphones, but it doesn’t really mean much to me yet. I don’t really understand how to read or make sense of it. So - at 1 kHz, open circuit voltage (whatever that means) I see -59 dBV/Pa (1.12 mV) - 1 Pascal = 94 db SPL.

So I guess if the microphone capsule sees 1 pascal of pressure (at a freq. of 1 kHz) the the mic will output 1.12 mV or -59 dBv? Am I mixing up RMS and PP?

The reading material for the mic says that typical speech three inches from the grill will need a microphone preamp with +60 dB of gain and that many modern microphone amps only provide 40-50 dB of gain.

You see, I kind of get how the circuit's DC function works. I understand the types of bias, how to read the DC load lines, more or less how to spec parts for each section kind of/not fully in terms of bandwidth (plate resistor, cathode resistor/bypass cap, interstage coupling capacitors, grid leak, grid stopper) I have a rudimentary understanding of negative feedback and how it works for the most part, but I feel like I’m missing some key knowledge when it comes to the small signal side of things. What do those signals look like - where are we wanting to take them? How much should we be amplifying those signals. I feel like one of D. Self’s Ten Commandments is thou shall not amplify and then attenuate but maybe another is also thou shall not attenuate then amplify…I would like to be able to design to fit those criteria. I know that he’s talking semiconductors but I would assume that it’s probably best practices for both? Ultimately the less attenuating we’re doing the less resistors we have in series with the signal and the less Johnson noise? Am I getting that right? Should I even be worrying about these noise issues when working with tubes or does the tube noise itself swamp any worry of other forms of noise? I don't need a pristine noiseless preamp. I have a nice GML 8403 and a few other nice mic amps for that.

I’m not looking to make a new product to take to market. I just want to have some fun and make some tools to use in my home studio. Thanks to this community and some other helpful folks, I feel like I’m getting closer but I am really having a hard time understanding what it is that I don’t know.

I will make another post here shortly with some more concrete planning that I've been circling around on.
 
Vacuum Tube Preamplifier Prototype

Requirements:

  • 65-80dB of gain (maybe?)
  • Transformer balanced input
  • Transformer balanced output
  • 48V Phantom Power
  • Polarity Reverse
  • XLR Input, XLR & 1/4” TRS Out
  • Pad?
  • Line in?
Ideally would build 4 channels with a single PSU


Potential topology 1:


1:10 Step Up
6072 in parallel
12AX7 cathode follower
Attenuator
6072(A) Gain Stage
6072(B) Gain Stage
12AX7 cathode follower
600:600 Output transformer

I'm not positive that this satisfies the above criteria yet.

Topology 1:

1:10 Step Up Transformer (20dB gain)
1:5 Step Up Transformer (14dB gain)
1st Stage (6072 in parallel)

Filament Requirements:

12AX7 & 6072
6.3 V @ 0.15 A
12.6V @ 0.3 A

So I'll need either around ~0.5 A at 6.3 V or ~1 A at 12.6 V per channel. I will address this further in the power supply design and decide whether or not to use DC for the heaters or just AC. It was interesting reading Ian Thompson-Bells notes on the EZ tubemixer (I think more specifically it was in the Lunchbox notes) that he saw no additional noise when moving from DC to AC heaters to reduce cost and free up space.

I am leaning heavily on Blencowe’s “Designing High-Fidelity Valve Preamps (2016)” for all of this stuff. If you’re wondering where these equations and conclusions are coming from, they’re his. All the mistakes and misinterpreting of course is all me.

According to Blencowe, valves in parallel could possibly improve the SNR by up to 2dB. I thought this may be prudent for the first stage. I have as well considered using a smaller step up transformer for the input, which I believe would also reduce the noise entering the first stage (I’m taking this from a maybe correct memory of watching a John Hardy video of him discussing Jensen input transformers). 1:5 would possibly work, but I will wait until I have prepared the remaining gain stages until I can decided whether or not to sacrifice the additional ~6dB of gain.

HT = 245 V
Ra = 82k Ω
Vgk = 3.5 V
Rk = 3k3 Ω
Ck = 470 µF, 25 VDC

Gain is 23 or ~ 27dB - this is only the gain of that tube at that particular bias point. This is not the gain of this stage.

I have chosen a bypass capacitor to keep the zero as close to DC as possible for this first stage. I’m not entirely sure I’m going to bypass this first stage or use a degenerated stage. Perhaps the degeneration would be good to lower distortion in the first stage. Ideally I would prefer to be able to add some harmonic distortion should I want it, rather than to have it as the rule.

As I understand it, for low frequency degradation each stage will add cumulatively so best to keep this low in the first stage.

Ck = 2πfloRk
Ck = 2 * 3.14 * 0.1 * 3300
Ck = 482.5 x 10-6
Ck = ~ 470 µF

From here I think that I can start to understand the gain of this full stage.

For a fully bypassed stage:

ra = 22,800 Ω of JJ 6072
A(Bypassed) = -µ * Ra/Ra + ra
A(Bypassed) = -23 * 82,000/82,000 + 22,800
A(Bypassed) = ~ -18 or 25 dB

I’m dropping the negative here. I think that just means that we’re working with an inverting stage.

or for a degenerated stage:

A(Degenerated) = -µ * Ra/Ra + ra + Rk(µ + 1)
A(Degenerated) = -23 * 82,000/82,000 + 22,800 + 3300(24)
A(Degenerated) = -10.25 or ~20 dB

The bypassed gain is about 1.8 times the degenerated, which following in the Blencowe book, looks about right compare to his example. Same with the negative here.

So, what then is the output impedance of this stage.

For the bypassed stage:
Ro(Bypassed) = Ra x ra / Ra + ra
Ro(Bypassed) = 82,000 x 22,800 / 82,000 + 22,800
Ro(Bypassed) = 17,840 Ω

And for the degenerated stage:
Ro(Degenerated) = Ra(ra + Rk(µ+1))/ Ra + ra + Rk(µ+1)
Ro(Degenerated) = 82000(22,800 + 3300(10.25+1))/ 82000 + 22,800 + 3300(10.25+1)
Ro(Degenerated) = 34,623 Ω

Or roughly twice that of the bypassed. Which following in the Blencowe book, looks about right at least compared to his examples.

Miller Effect:

How do I calculate the miller capacitance on the input of a parallel stage?
I would think that I would add my inter-electrode capacitances as I would any capacitors in parallel.
For the 6072 I am working with the following inter electrode capacitances:

Cgk = 1.6 pF
Cga = 1.7 pF

So for my valve in parallel I can double these. As per Blencowe’s recommendation I will add some additional capacitance due to the proximity of the grid-anode connections and the grid-cathode connections. This is where I get a little confused. Am I doubling those values before I add the stray capacitances or after?

So should I be looking at 5.4 pF of capacitances or closer to 6.4 pF.

I will take the worst case ontario for now.

For the bypassed stage:
Cin = Cgk + Cga(A+1)
Cin = 6.4 + 6.4(19)
Cin = 128 pF

And if I go with the degenerated stage:
Cin = 6.4 + 6.4(11.25)
Cin = 78.4 pF


Am I way off here?
 
Last edited:
Tip #1. Tubes, even triodes, are relatively noisy so you need as much relatively noise free gain before the very first tube. A transformer is a handy source of nearly noise free gain. A 1:10 ratio will give you a handy 20dB of gain and a noise figure better than 2dB. Use a 1:5 and you throw away 6dB of gain and raise the noise level by the same amount.

Cheers

Ian
 
It's all good Gus. Erase it if you wish, but asking questions is absolutely fine, we all had those same questions at one time :)

What are your goals? Have you built or used a valve pre amp that you like? Or one you dislike?

I'd say, if you're new to building, best to build something with a proven outcome. Once you have that sorted, you can change stuff as needed.

Edit: On your noise questions - if you're recording harpsichords with ribbon mics, then the noise of a valve pre will probably be an issue for you, unless you go to extraordinary lengths with suped-up valves and topologies.

But for pop/rock type stuff, all of the 'usual suspect' decent valve pres will most likely be fine. A REDD.47 , any of Ian's designs, Jakob's G9, a proper UA 610, a V76, an RCA... Lots to chose from.

If you want super low noise but the voltage swing of valves, maybe you need to add some solid-state into the mix. - two or three parallel 2SK170BL J-Fets cascoded with, say, a 6922/E88CC would get your noise as low as you'd ever practically care about.
 
Last edited:
My goal is to design and build a two channel mic amp from scratch. I want something that I can use in my studio, modify it, experiment with it until I have what I want. I want to have things that can sound kind of nice or really ugly.

I have used some Summit, UA, Manley, Pultec, Thermionic Culture and Tube Tech gear before. Mostly of what I have in my home studio now is solid state, Focusrite, Showdow Hills, Vintech and GML.

I have built a number of guitar pedal from kits, a NYDave reamp, a Hairball Fet Rack, revived or moded countless tube guitar amplifiers, fixed loads of random studio equipment from powered monitors to microphones and worked on some Hifi gear. I am a beginner still for sure. This much I do know.

Mostly I just like working with the valve gear because I find it intriguing and fun. I live rurally and need to keep busy to keep my head on straight and stay out of trouble. I like to build things, do metal work, etc. I have a woodshop and a machine shop at my disposal. What I don't have is anyone I can talk to about this stuff that interests me and I don't have access to internet at my house.

I'm not overly concerned with super mega low noise. But I don't want to build something that is crazy noisy either. Mostly I would like to understand the limiting factors and do my best with what I have available. I get the broad strokes of how tube amplifiers work, but am struggling to understand the smaller details.
 
Feel free to ask as many questions as t=you like. This probably the one forum in the universe where you will receive civilized answers.

Tubes have an undeserved reputation for being noisy but it is not too hard to achieve and EIN of around -125dBu. You need to pay particular attention to the power supplies, use dc heaters and enclose the pre in a screened box but other than that it is nothing more than good engineering practice.

To put this into contest, the theoretical noise from a 150 ohm source is -131dBu. If you have 60dB of gain in your pre then the best the output noise can possibly be is 60dB above -131dBu which is -71dBu. With a decent tube mic pre it will be about -65dBu.

Cheers

Ian
 
"...

Miller Effect:

How do I calculate the miller capacitance on the input of a parallel stage?
I would think that I would add my inter-electrode capacitances as I would any capacitors in parallel.
...
Yep, and then to that parallel value you'd add the strays not inherent to the valves - wiring from the transformer etc. This value would be a guesstimate but you're probably about right in your values.


I will take the worst case ontario for now.

For the bypassed stage:
Cin = Cgk + Cga(A+1)
Cin = 6.4 + 6.4(19)
Cin = 128 pF

And if I go with the degenerated stage:
Cin = 6.4 + 6.4(11.25)
Cin = 78.4 pF


Am I way off here?"
No I don't think you are "way off". Without looking at the data sheets, it seems about right to me.

I'd say you seem to have a pretty good grasp of all the little bits that come into play here.

You know that doubling up on input devices gives about 2dB noise benefit out of the theoretical 3dB.
But that the trade off is increased Miller/input capacitance which, with a 1:10 input transformer reflected secondary impedance, will start to affect your high frequency response.

There's also the issue of the transformer's inherent response itself which, with a high ratio like 1:10, will have its resonant peak closer to the audio band.
So, do we let it ring? Or do we put a zobel on it which will most likely leave the transformer just about making it to 20kHz.

I've tried both the Cinemag and the Jensen 1:10's (not any of the Sowter though) and, my own preference is to go for a slightly lower ratio (1:7 or 1:8) and suffer the little bit less noise benefit I'd get from using the higher ratio and keep my audio mostly free of those particular and pesky transformer issues at the top end.. Horses for courses though really.

About the best 1:10 I tried was Oliver Archut's CTV/72 1 and I'm now ruined! I just can't get the memory of that out of my head so, what to do?
Those particular transformers are very complicated to wind and the price to get one now is about 4 - 5 times the price of a Cinemag or Jensen.

An alternative I'm looking at is the Lundahl 7903 which is 1:8 and has a self resonance up at 80kHz which keeps my audio nice and flat until up there in the clouds.

Cascoding a couple of parallel 6072 sections?
Use another 6072 half as the top device and then the other half of that particular 6072 as a cathode follower so any following device has an easer time with its own Miller vs source impedance etc.

You'd get the 2dB noise advantage, you loose a good chunk of the Miller, and you get the gain of a pentode without any partition noise.

Choke load it and the load line becomes almost horizontal, linearity improves, your rms voltage swing goes way up...

Just some of my own incoherent ramblings in response to your really quite cogent post :)
 
Last edited:
Most 1:10 mic transformers have a response that depends more on the source impedance than the load. Typically they are dead flat out to 40KHz with a 150 ohm source. Change the source impedance to 50 ohms and you get a nice peak around 25KHz.

Cheers

Ian
 
I'm not really talking about load Ian, more that the Jensen and Cinemag 1:10's have resonance closer to audio bandwidth.
Mics that are built in the old school German or Austrian tradition typically have output impedance that's higher than 150 ohms - around 200 or a smidge higher. Might not seem much difference, but I've never got a dead flat to 40kHz once a suitable zobel to flatten resonance is applied.
I like to test at 200 ohms as there are a lot of U47-esque mics out there, and then also at 300 ohms which I find is nominal z for quite a few ribbon mics.

By no means am I saying the Cinemag CMMI-10C and the Jensen 115K are bad. There are, collectively, thousands of valve pre amps out there that have been sold, and that I had some sort of hand in. They mostly use Cinemags because that's what I spec'ed.

Just that, for almost all of what these valve mic amps get used for, I honestly never found that using a 1:7 or 1:8 gave me anything to worry about in terms of having a bit less 'noiseless' gain than if it were a 1:10.

Side observation: there are a valves with equivalent noise resistance around 100R so, in that case, even a 1:4 or 1:5 would get the mic signal's transformed source noise well above that.

Going much higher in transformed source noise is moot then surely, yes?
Genuine question there for Ian and the other the more mathematically adept than I readers :)
 
So if I am taking all of this into consideration. I am seriously grateful for the help here. I truly mean it.

In regards to the input transformer. I have been looking at the following potential transformers and I would like to figure some maths out for all of them.

Cinemag CMMI-8B3 1:10 (It was nice to see your name on this one Ian.) - I will look at the 10C as well.
Lundahl 7903 1:8
Sowter Type 9610 1:10
Jensen JT-115K-E 1:10
Edcore MXL8cs 1:8 (Though I still have some questions on this one.)

I am not set on a topology yet and will consider cascode, mu-follower, SRPP, White follower and regular old common cathode stages and cathode follower stages.

I'm not entirely sure what choke loading means? Is it giving you a CCS?

Finally - I'm not entirely sure how to calculate the -3dB point using the Miller capacitances I had figured before. I assume that we're talking about some kind of RC filter here but not entirely sure where to start or stop. It's AC too right - so there is a path to ground essentially toward the B+ but also the DC ground?
 
It seems nobody took care of the difference between quoting transformer ratio by the impedance or the voltage ratio.
1:6.45 is a voltage ratio. 220 ohms:50kohms is the impedance ratio, which, as the OP surmised is the square of the voltage ratio.
So the 200:50k xfmr has a ratio of sqrt 250, or about 16, and the 1:6.45 xfmr could be quoted as as a 200:8,3k ohms.
 
Abbey, I don't know that I've seen impedance being used when quoting specified ratios of transformers. If I see 1:7 then I assume 200 : 9K8 and thought this was standard practice no?

Though there are differences in how you sometimes see impedance being specified in the US vs say in Europe.
In the US, a 1:2 mic input transformer would be specified and identified by the nominal source and reflected secondary impedance as a 150:600 for instance.
Whereas, looking at the Carnhill 10468 1:2 Neve mic input, the reflected primary load is specified and it's a 1K2:4K8.

This has caused confusion with a few folks (including a very well known manufacturer of various bits of expensive kit) and they look at the lower Z strapping of that Carnhill 10468 which is specified as 300:4K8 and think that this 300 ohm figure applies to the source impedance and assume this is correct for your common-or-garden 150-200 ohm impedance microphone.
It may well be an artistic choice as an effect or tonal change, but it's not the intended interpretation of the figure.
 
Last edited:
I'm not really talking about load Ian, more that the Jensen and Cinemag 1:10's have resonance closer to audio bandwidth.
You said " There's also the issue of the transformer's inherent response itself which, with a high ratio like 1:10, will have its resonant peak closer to the audio band.
So, do we let it ring? Or do we put a zobel on it which will most likely leave the transformer just about making it to 20kHz. "

Hence my confusion. However, I have tested the standard 1:10 transformers by Cinemag, Sowter and Jensen and with a source of 150 ohms there are all pretty flat out to 40KHz with no peaking. It is only with sources of less than 150 ohms that the just out of audio band peaking occurs. I have not tried sources above 150 ohms but I see no reason to expect the peaking to recur.
Side observation: there are a valves with equivalent noise resistance around 100R so, in that case, even a 1:4 or 1:5 would get the mic signal's transformed source noise well above that.

Going much higher in transformed source noise is moot then surely, yes?
Genuine question there for Ian and the other the more mathematically adept than I readers :)
Unfortunately the equivalent noise resistances quoted for tubes are all radio frequency figures, usually measured at a few MHz. They only include shot noise because at those frequencies 1/f noise is not a significant factor. However, in the audio band, shot noise is by no means the whole story. In the audio band, 1/f noise often exceeds shot noise from a few KHz downwards. It is an unfortunate fact of life that the high quiescent current required for minimum shot noise is not the one required for minimum 1/f noise so this does not usually result in the lowest possible noise. Merlin Blencowe has done some excellent work on the relationship between the two.

Tubes with quoted rf noise resistances of 100 ohms tend to be very high gm devices with fine closely wound grids. High gm can lead to parasitic oscillation, the usual cure for which is a grid stopper - but adding a 1K resistor in series with the grid rather negates the reason for choosing a tube with a 100 ohm noise resistance. In addition, fine, closely would grids tend to be very microphonic. In short I think mic transformer ratio is not yet moot.

Having said all that, I am not unwilling to be persuaded a quieter tube mic pre can be made. I have designed a version of the EZTubeMIxer mic pre gain blockusing a D3a tube to replace the usual 12AX7. I have also purchased some D3a tubes and plan to build one to see just how quiet it is.

Cheers

Ian
 
I have tested the standard 1:10 transformers by Cinemag, Sowter and Jensen and with a source of 150 ohms there are all pretty flat out to 40KHz with no peaking. It is only with sources of less than 150 ohms that the just out of audio band peaking occurs. I have not tried sources above 150 ohms but I see no reason to expect the peaking to recur.
In this instance, I wasn't referring to an issue with peaking per se, but rather an issue with more droop at the high end with 200 ohms sources on a 1:10 vs a 1:7 or 1:8 once a noble zobel is used.
The peaking I mentioned was the inherent resonance of the transformers, a 1:10 typically exhibiting this inherent/intrinsic peaking lower down, closer to our 20kHz top limit than a good 1:8

Just my own personal preference, nothing more, to forgo the couple extra dB noise advantage of a 1:10, to give a fair crack at the whip to a slightly lower ratio, for a top end response that behaves very well as high up as is practical.

That 2dB noise advantage loss then being overcome by:


Unfortunately the equivalent noise resistances quoted for tubes are all radio frequency figures, usually measured at a few MHz. They only include shot noise because at those frequencies 1/f noise is not a significant factor. However, in the audio band, shot noise is by no means the whole story. In the audio band, 1/f noise often exceeds shot noise from a few KHz downwards. It is an unfortunate fact of life that the high quiescent current required for minimum shot noise is not the one required for minimum 1/f noise so this does not usually result in the lowest possible noise. Merlin Blencowe has done some excellent work on the relationship between the two.

Tubes with quoted rf noise resistances of 100 ohms tend to be very high gm devices with fine closely wound grids. High gm can lead to parasitic oscillation, the usual cure for which is a grid stopper - but adding a 1K resistor in series with the grid rather negates the reason for choosing a tube with a 100 ohm noise resistance. In addition, fine, closely would grids tend to be very microphonic. In short I think mic transformer ratio is not yet moot.

Having said all that, I am not unwilling to be persuaded a quieter tube mic pre can be made. I have designed a version of the EZTubeMIxer mic pre gain blockusing a D3a tube to replace the usual 12AX7. I have also purchased some D3a tubes and plan to build one to see just how quiet it is.

Cheers

Ian


I am indeed talking in part about some valves which were intended for non audio use. Not necessarily RF in all cases though. Some were just robust 10,000 hr life, audio frequency, but deep-sea conditions ready buggers.

I've measured very low noise (comparatively speaking) at audio frequencies with some of the usual suspects, and oscillation hasn't been an issue so far. for me Microphonics on a few have been troublesome however.
If I were to experience oscillation, I would look to some other method than a grid stopper of a value that would negate/swamp out any noise advantages.

Before they became more expensive, back when they were plentiful and about £6 squids a pop, I played around a lot with the D3a's myself. One or two were microphonic. But never did I need a 1K grid stopper. Layout, topology come into it of course and others' mileage may vary in their own particular rendition.

There are a couple of Russian pentodes that seem to be about on par with the D3a noise wise. Slightly less gain in triode mode at mu of 40, but again very linear in triode mode, and currently about £2 each until the Hi-Fi crowd gets wind.

I'm also not averse to using forgotten tubes that are still inexpensive and good enough to parallel 2 or 3 on the bottom of a cascode.
The Brimar 6BR7 is about equivalent of a good EF86 but I'm still seeing them for £6 each.
I just used two in triode on the bottom, with another in triode on top of a cascade configuration.

Basically, I wasted some heater power and needed 70V more HT supply in order to use 3 pentodes in triode mode to emulate their behavior as a single pentode. But without partition noise, and a 2.5dB noise advantage from the paralleling of the lower 2 valves. One could even add a third 6BR7 on the bottom for going even more overkill :D

Anyway, this is DIY, and no one in their right mind would do that sort of stuff for a commercial product.

But cost for the 3 valves was still less than the ever climbing dosh currently needed for a good EF86 or EF806S though.

If you can be bothered with a grid on a top cap (double ended valve) then the Brimar 6BS7 is even cheaper and supposedly consistently lower noise still than the 6BR7.

All in, I'm not convinced the so called 'low noise' audio tubes we've all been using for decades are actually even all that quiet compared to some. I believe there are better out there that us civvies didn't get our hands on or just plain couldn't afford when the military were buying them at £4, 3s, 6d each in old, 1969, money.

Cheers Ian. I'll be curious what you find out re.the D3a :)
 
Last edited:
"... looking at the following potential transformers and I would like to figure some maths out for all of them.

Cinemag CMMI-8B3 1:10 (It was nice to see your name on this one Ian.) - I will look at the 10C as well.
Lundahl 7903 1:8
Sowter Type 9610 1:10
Jensen JT-115K-E 1:10
Edcore MXL8cs 1:8 (Though I still have some questions on this one.)

Most all look like good choices and options to me, although I have no experience with the Edcor so can't comment on that one.

I like pretty much all of the Cinemag CMMI- C stuff, have used boatloads of 1:7's, 1:10's.
I'm currently on a path with the Lundahl LL7903 though, and I really like what I've seen so far but, we'll see I suppose...




I'm not entirely sure what choke loading means? Is it giving you a CCS?

Actually yes, for a good part, it is. But it's commonly referred to as parallel feed whereby the anode is loaded with an inductor which is then cap coupled to the following device - whether that be transformer, another valve, or frontal-lobe electrodes.



Finally - I'm not entirely sure how to calculate the -3dB point using the Miller capacitances I had figured before. I assume that we're talking about some kind of RC filter here but not entirely sure where to start or stop. It's AC too right - so there is a path to ground essentially toward the B+ but also the DC ground?

It would be defined by your source impedance against your Miller C
thus:

1
----------
2 x Pi x R x C.

You can just ballpark estimate it for now and, after you build it, measure it if needs dictate and must.

I honestly don't see you having issues with understanding any of this stuff, you pretty much already do in my very humble opinion.

Keep going :)
 
Last edited:
Back
Top