How to calculate mic input xfmr ratio for tube circuit

Help Support GroupDIY:

ruffrecords

Well-known member
Joined
Nov 10, 2006
Messages
13,402
Location
Norfolk - UK
abbey road d enfer said:
Thermal agitation of electricity in conductors by JB Johnson Physical review July 1928.

Thanks. Mr Google found it for me.

Interesting that the 'working formula' is just plucked out of the air without any explanation of its derivation. Also interesting is that it makes tube noise directly proportional to idle current whereas gm tends to increase with idle current which means the 2.5/gm formula is inversely proportional to idle current

Cheers

Ian
 

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
12,355
Location
Marcelland
Well, mu increases also with current, so it's not directly linear.
Apparently, both the Johnson formula and the 2.5 are "working" formulas, which if I understand well, is olde saxon for empiric(al).
It seems that this subject is still current within the IEEE. I know there are quite recent publications on the subject; I'm interested but $32 for a download...I baulk.
 

ruffrecords

Well-known member
Joined
Nov 10, 2006
Messages
13,402
Location
Norfolk - UK
abbey road d enfer said:
Well, mu increases also with current, so it's not directly linear.
Apparently, both the Johnson formula and the 2.5 are "working" formulas, which if I understand well, is olde saxon for empiric(al).
It seems that this subject is still current within the IEEE. I know there are quite recent publications on the subject; I'm interested but $32 for a download...I baulk.

There is a famous old paper that derives that 2.5/gm formula but it only applies at radio frequencies. I am not sure how it became 'applied' to audio.

Are you an IEE member or is the $34 the non-member price? Do you have links to the articles? - I might punt 32 bucks for the latest research.

Cheers

Ian
 

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
12,355
Location
Marcelland
I'm not an IEEE member.
Pricing info is there:
http://www.ieee.org/publications_standards/publications/subscriptions/prod/mdl/ieeexplore_access.html

I have tried to locate the article that seemed of current interest but I can't...All I could find is the vintage stuff (Llewellyn, Pearson,...). Hope you're better at googling than me...
 

rackmonkey

Well-known member
Joined
Sep 20, 2016
Messages
895
Location
Texas
Summoning this thread from deep six.

What ever became of some of the discussed follow-ups here? Paul S mentioned having a class do some measurement work. There was talk about potentially purchasing an IEEE article on the subject. Did anyone follow this up with additional investigation? It’s an interesting topic with real world implications. Wondering what, if any, new conclusions have been reached since 2011.

BT
 

ruffrecords

Well-known member
Joined
Nov 10, 2006
Messages
13,402
Location
Norfolk - UK
Yes there have been many years of follow up. Bottom line is the 2.5/gm is only part of the story. The other major component is flicker noise which is ignored by most.. All I spent a lot of time to find more information on tube noise and I eventually found a book that goes into detail on it. The attached pic shows the noise formula for a triode. It has two components. The first is the shot noise component commonly known as the 2.5/gm component. The second is the flicker noise. As you can see it is directly proportional to an empirical constant K which varies with tube type and from tube to tube. There has always been an imperfect understanding of the mechanism(s) responsible for flicker noise in tubes which is why tubes used to be hand selected for low noise use. Bottom line is that for audio work, flicker noise is an important conponent of total tube noise. shot noise alone is not accurate.

After all this there is the question of the weighting network used in the measurement and the type of rectifier used both of which can significantly affect the values obtained. Some weighting networks give better noise figures and most measurements have been made with rms calibrated average reading meters which are OK for sine waves by typically under read noise. There is a European standard for audio noise measurement which uses a quasi peak meter and a C weighting network. This can give readings as much as 12dB higher than an rms calibrated average reading meter with and A weighting network.

Bottom line is tubes are noisier than the simple 2.5/gm would lead you to expect. To obtain a good noise factor you do need a 1:10 step up transformer. However, with careful attention to the implementation you can achieve and EIN in the region of -125dBu. To put this into context, even a completely noise free amplifier would only be 6dB quieter when feed from a 150 ohm source.

Cheers

Ian
 

Attachments

  • noiseformula.png
    noiseformula.png
    7.2 KB · Views: 13

rackmonkey

Well-known member
Joined
Sep 20, 2016
Messages
895
Location
Texas
Thanks, Ian. What's the name of the book you found, and did you find it online or was it a physical book? Looking at the formula, I don't know what some of the variables are (small "k" Tk, d). T is temperature, I assume, but temp of what (k)? Surely that's not the cathode temperature?  :eek:

It's surprising that you say 1/f noise has been overlooked in the past. My experience measuring resistor noise is that we have to focus on current noise because it's the type we have some control over, unlike Johnson noise.  Shot and 1/f noise above the background threshold is what you look for, then.
 

ruffrecords

Well-known member
Joined
Nov 10, 2006
Messages
13,402
Location
Norfolk - UK
Tk is indeed the cathode temperature. k is Boltzmann's constant.

d(Vn2) is the noise voltages squared per unit bandwidth. So to get the total noise you integrate this over the desired bandwidth.

The book is a physical one called "Amplifying Devices and Low-Pass Amplifier Design" by E.M. Cheery and D.E Harper published in 1968. It is not cheap:

https://www.gettextbooks.com/isbn/9780471153450/

Cheers

Ian
 

nille

Member
Joined
Mar 29, 2009
Messages
20
Location
Sweden
https://archive.org/details/CherryHooperAmplifyingDevicesAndLowPassAmplifierDesign1968RR
 

ruffrecords

Well-known member
Joined
Nov 10, 2006
Messages
13,402
Location
Norfolk - UK
rackmonkey said:
Nice find, Nille. Thanks!

Definitely. Iwish I had known about that before I bought the book.

Tube noise starts on page 72 where both shot and flicker noise are introduced. Note the statement "Flicker noise in a given tube can be reduced only by reducing the anode current" Most "low noise" audio designs I have seen seek to minimise shot noise by running the tube at a high anode current without realising it increases flicker noise.

Note also the statement that "For typical tubes ....  flicker noise exceeds shot noise below a few kiloHhertz."

The detailed treatment begins in Chapter 8 and for tubes begins on page 372.

Cheers

Ian
 

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
12,355
Location
Marcelland
rackmonkey said:
My experience measuring resistor noise is that we have to focus on current noise because it's the type we have some control over, unlike Johnson noise. 
Actually we have a lot of control on Johnson noise, by either tuning the source Z to the OSI (using a transformer) or tuning the operating point to adapt it to the existing source Z, or both.
It has always been understood that vacuum tube's OSI is in the 100's of kohms range, so ideally a 1:50-100 xfmr would present the best SNR, however nobody knows how to make one suitable for mic inputs, so we have to make do with what's humanly possible. I think 1:18 is painstakingly achievable. Noise current in resistors should not be an issue today, with WW resistors being available in all the relevant values.
 

emrr

Well-known member
Joined
Apr 12, 2006
Messages
7,798
Location
NC, USA
There are several historical types that I measure at roughly 1:22 ratio which IMO also measure fine for practical response.  The practical bottleneck in modern practice is managing that 26-27 dB gain, which is already more than is needed with high output mics and loud sources. 

The most extreme 'full range' example I can think of is the Altec 458A/459A type, 1:31.62.
 

Bonnie1

Well-known member
Joined
Jan 24, 2016
Messages
47
Higher ratio's are possible with great success,
The V-series preamps 72, 76, 77 etc... designed by TAB and Siemens for German Broadcast designed input transformers with a 1:32 - 40 ratio.  Very labor intensive and extremely expensive to manufacture.  In the late 40's they used magnet wire wrapped in silk for isolation.
 

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
12,355
Location
Marcelland
emrr said:
There are several historical types that I measure at roughly 1:22 ratio which IMO also measure fine for practical response.  The practical bottleneck in modern practice is managing that 26-27 dB gain, which is already more than is needed with high output mics and loud sources. 
Agreed; optimizing for best NF is sometimes controversial with practical use. Same with SS xfmrless mic pres, their best NF is achieved at about 60dB gain.

The most extreme 'full range' example I can think of is the Altec 458A/459A type, 1:31.62.
Impossible to reissue today, according to
http://vintagewindings.com/tech%20swag/Vintage-Wideband-Audio-Transformers.pdf
 

ruffrecords

Well-known member
Joined
Nov 10, 2006
Messages
13,402
Location
Norfolk - UK
emrr said:
  The practical bottleneck in modern practice is managing that 26-27 dB gain, which is already more than is needed with high output mics and loud sources. 

A lot depends on the following electronics. Many tube mic pres have an output stage with a 4:1 step down transformer in order to drive a 600 ohm load and which drops the gain by 12dB. The total transformer gain is therefore often in the region of 14dB. The problem with historic tube mic pres was varying the overall gain over a wide enough range without compromising headroom or stability. Most designs have limited gain range (V72 for example) and rely on input pads to handle high levels. The exception is the V76 which can vary its gain over a wide range by altering both open loop and closed loop gain at the same time in order to maintain stability. This, in combination with a switched input pad allows it to handle a very wide range of input levels,

Cheers

Ian
 

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
12,355
Location
Marcelland
Bonnie1 said:
Higher ratio's are possible with great success,
The V-series preamps 72, 76, 77 etc... designed by TAB and Siemens for German Broadcast designed input transformers with a 1:32 - 40 ratio.  Very labor intensive and extremely expensive to manufacture.  In the late 40's they used magnet wire wrapped in silk for isolation.
And achievable in the context of 40-15kHz BW
 

emrr

Well-known member
Joined
Apr 12, 2006
Messages
7,798
Location
NC, USA
ruffrecords said:
A lot depends on the following electronics.

I don't see any preamp input stages that can take a +10 mic level output (common) followed by 20+ dB of transformer gain directly, nor any input transformers in use which operate in that territory anyway, leading to:

ruffrecords said:
a switched input pad allows it to handle a very wide range of input levels,

Somewhat defeating the point in some ways.  Not the ideal match of mic to preamp in that case. 

If the transformer can take the level, you could run directly into a CF stage feeding an output transformer to get low gain without an input pad. 
 

ruffrecords

Well-known member
Joined
Nov 10, 2006
Messages
13,402
Location
Norfolk - UK
emrr said:
(input pad) Somewhat defeating the point in some ways.  Not the ideal match of mic to preamp in that case. 
To achieve the signal levels you are suggesting (+10dBu)  the sound level of at the diaphragm of of even the most condenser sensitive mic would need to be approaching 130dB SPL. Hardly operating the mic at its lowest distortion point.  It has an electronic output anyway so matching to the preamp is not really an issue.

Cheers

Ian
 

emrr

Well-known member
Joined
Apr 12, 2006
Messages
7,798
Location
NC, USA
ruffrecords said:
To achieve the signal levels you are suggesting (+10dBu)  the sound level of at the diaphragm of of even the most condenser sensitive mic would need to be approaching 130dB SPL. Hardly operating the mic at its lowest distortion point.  It has an electronic output anyway so matching to the preamp is not really an issue.

People do have this habit of putting condensers close to drums......
 
Top