Bipolars for low noise input - Alternatives to 2SC2240

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I had some evenings listening to this MC stage. I started with 4 Volt rails, instead 12V, and gradully increased the voltage. At all voltages, I measured the voltage offset at the emitters of the two input trannies. Unfortunately, at rails > 8V, there was an increasing offset > 50mV, so did not dare to use this with my delicate Ortofon MC7500 cartridge.

Soundwise, this MC stage is way more transparent than the original Hiraga LePrepre. Though, when varying rail voltages, there were no significant effects.

This was with 2SC2546/2SA1084 trannies.

Next steps: either using a THAT's transistor array, or modifying the circuit to virtual zero impedance.

Hartmut
 
One really really :) does not want a current greater than a small proportion of a uA flowing through the cartridge ( Rs 3 to 5R typical). I have seen this statement from leading authorities and I have seen and played with common base MC preamps. And I have wondered where a current "preloads" the cartridge and what it does for distortion.

There are a limited number of options:-

1.) Always use an input coupling capacitor...several millifarad electrolytic
2.) Use an incredibly clever bias current cancelling scheme, I have tried several and they always have a fatal flaw. But some close to being cunning.
3.) use Mr D Self's servo integrator to cancel offset in a single ended amplifier.
4.) Use a differential amplifier and either a.) accept that you have just added 3dB to the noise component or b.) you have a cunning scheme to reduce noise.

I have finally arrived at No. 4 option b.) and it uses op-amps, :Pirate:


2SA1085 & 2SC2547 etc....with all due respect to others, I have never ever been aware of a "sound" to individual basic semiconductor devices. I understand the ancient phenomena of 1970's "transistor sound" and it's cause and why it might even be an issue with certain op-amps. One transistor amplifies just as well as another of similar parameters. Many parameters are never quoted by manufacturers, eg. if a certain transistor is manufactured with a high level of contamination by heavy metals it may have a high level of flicker noise, and I do not think any semi manufacturer has ever admitted that flicker noise exists, at least not in their own products :). (Actually YES they do when they publish 100Hz and 10 Hz noise figures I am unfair to them) Some transistor parameters are poorly understood by designers and devices are used innappropriately. Nevertheless it seems unfortunate to me that a perfectly good device might acquire a bad name without a clear engineering analysis of why so, and this can then become a reputation and urban myth, which is unfortunate.
 
I should also add to my previous post..

I do not include 100% symmetrical discrete amplifier schemes as a solution per se, because 1.) Input transistor currents necessarily should and must run at Ic of 1 to 3mA for minimal noise, implying several uA of input bias current. 2.) Matching NPN and PNP parameters eg. hfe over temperature etc within a fraction of a uA bias... not even in your dreams. Vbe is dependant upon Is which varies unpredictably between NPN and PNP.
 
One really really :) does not want a current greater than a small proportion of a uA flowing through the cartridge ( Rs 3 to 5R typical). I have seen this statement from leading authorities and I have seen and played with common base MC preamps. And I have wondered where a current "preloads" the cartridge and what it does for distortion.
...

Blackjack,

I intend to take some distortion figures for an MC cartridge using the circuit above and compare it to the same cartridge using other circuits free of DC input currents.

Hartmut
 
Could you explain the procedure?

Rbb came up in another thread but the link to measuring it was not available to read.

The voltage noise of a transistor, ( or an opamp that generally has 2 input transistors), comes principally from 3 sources. 1/f or flicker noise at very low frequency, say 100Hz down to below 10Hz and which we will ignore for now, so instead we will use a 1KHz frequency as useful. The other two are....

Johnson noise.. Vnoise=sqrt(4kTRB) and R in this case is rbb' which is the base spreading resistance. It varies from 4 or 5R in a 2SB737, 6R in each input transistor in a AD797, 30 or 40R in each of the pair in a LM394 (SSM2210). and much greater again in common or garden transistors. For obvious reasons it is lower in power transistors.

Shot noise....is fundamental just like Johnson noise. Inoise=sqrt(2qIB) where I in this case is the steady dc collector current of the transistor. But we also know Ebers-Moll that re the intrinsic emitter resistance...
re=25/Ic ( in milliamps ). At 1mA collector current the re is 25R and at 3mA it is 8R. The Shot noise current runs through this impedance and so adds as a voltage noise ......

....to the Johnson noise as an RMS sum (re is not a real resistance and so does not generate Johnson noise). This is en as sometimes quoted on data sheets. Now. As Ic increases so does the shot noise current, BUT only as a square root and the re reduces linearly, so the overall shot noise drops and en drops as a result as the current increases.

To all of this we must add in the current noise which is simply the Shot noise of the base current. This increases ( shot noise formula above) as Ic increases ( a high Beta/HFE is good for reducing base current).

The overall noise of an amplifier is therefore the Johnson noise of Rs the source impedance summated ( not simply added) with en and with in/Rs.

Noise figures and noise figure curves are an excellent guide to "how much worse" all of the noises of a less than perfect amplifier affect the amplification performance of a simple signal from known source impedance. Of course this signal source has it's own Johnson noise and hence noise figures are a ratio of perfect/real.

(Close to the end now!)
Noise figure curves demonstrate the optimal operating point in terms of Ic for a given frequency ( 1Khz typically ) and resistive source impedance, where the contribution of en is minimal and so is the contribution of in. The curves cover a range of Rs, usually several decades up from 100R.

If you draw a line through the lowest optimal point of the curves, slanting back up left, you can make a rough estimate of in, and from that estimate en and calculate back to rbb'.

The quick and dirty answer is....if the noise curves optimise at 10mA or above for an Rs of 100R then is rbb' is a "few ohms". Optimal at 1 to 3mA then rbb' is a dozen or two ohms. Optimal at 1mA or less the rbb' likely to be a few tens of ohms.

And above is a clue that while power transistors have a lower rbb' they will have a higher in and quite probably worse flicker noise. The End:)
 
Burkhard Vogel
The Sound of Silence
Lowest-Noise RIAA Phono-Amps:
Designer’s Guide

Mr. Vogel's book contains a formula for calculating Rbb, but you have to enter lots of constants and a noise voltage from the graph of noise voltage vs frequency in the transistor specification then solve a quadratic equation. What I do is just look at the graph of noise voltage Vs frequency and calculate the value of resistor that produces the same Johnson (thermal) noise. A 1k resistor produces 4Nv-rt-Hz thermal noise, a 250 ohm resistor 2Nv-rt-Hz, a 62 ohm resistor 1 Nv-rt-Hz. Typically a low noise transistor will have a noise voltage on the order of 1 Nv-rt-Hz around 1ma collector current. The noise voltage is usually a minimum around 1 ma of Ic. Calculate the resistor value that generates the same noise voltage and this will give you an estimate of Rbb. I compared this method with the value calculated by Mr. Vogel's formula and it's typically within 10%, and it's a lot simpler to calculate.
 
How does the monolithic duals MAT02/03 fit in this picture.....?
Seen them used in some very low noise instrumentation preamps.
Quite expensive, but still made, AFAIK..........


"Sort of OK" is the answer. The history of the MAT 02 is involved.
Essentially SSM2210 ( and the pnp SSM2220 ) were an invention of Precision Monolithics who were bought by Analog Devices who replicated them as MAT 02 and MAT 03 ( there is a MAT 04 quad but each tranny has half the numbers! ). Nat Semi then replicated them as LM 194/394.
rbb' is variously quoted as 30R or 40R.

BUT they do not have a very high fT. Internal capacitance is highish and the PNP is a dreadful mismatch to the NPN, the PNP has a low Beta. So clever bias current cancelling circuits do not work and hf stability needs to be watched.

I have a very small bucketful of each and every variety rescued and unused.

Positively. The NPN variety have a very high Beta such that in is quite low and at around 1mA or so with about 100R or so source they make a decent low noise amplifier. The device matching is top class and they have very close to theoretical characteristics and so for logarathmic circuits such as multipliers they are all but irreplaceable or for posh current mirrors or for speciality opamp input replacement long tailed pairs etc.

For MC cartridge preamps or other very very low Rs transducer amplifiers, they no longer have any special characteristics. An AD 797 op amp with a similar en of 0.9nV sqrtHZ plus a well designed back end end with mega load drive capability and at less than £5 or $5 a pop makes them redundant.

I have the greatest respect for Analog Devices, I have been using their products since 1976. The reason that 2SB737 etc are not made any longer by Rohm or whoever is that an AD797 is the worlds very best low noise amp at low Rs. It meets strategic needs and anything else is a US military secret, and needs liquid helium to keep it at almost absolute zero.

HEY GUYS...If you could cool your preamps and the MC cartridge to about 2 or 3K then noise would never be a problem. The noise of the cooling systems out in the back garden might intrude!!!
 
"HEY GUYS...If you could cool your preamps and the MC cartridge to about 2 or 3K then noise would never be a problem. "

One little problem. Semiconductor amplifiers don't operate at that temperature. I have an article on the design of a low noise amp for a cryogenic photodetector (liquid helium cooling) used in astronomical work and they had to keep the amp portion at about 70 degrees kelvin or it stopped working. Those poor electrons and holes in the fets and bjts would just freeze to death at a lower temperature.
 
Blackjack8635;2465415 HEY GUYS...If you could cool your preamps and the MC cartridge to about 2 or 3K then noise would never be a problem. The noise of the cooling systems out in the back garden might intrude!!![/QUOTE said:
Hm... wouldn't 77K be enough......?
Many years ago, I operated an IR spectrometer through some winter seasons, topping up the detector coloumn with liquid nitrogen from a jug, every afternoon.......
Could be way cool to start an evening music session with some LIN fog in the living room.......:D
 
YES!. I said..."If you could cool your preamps and the MC cartridge to about 2 or 3K then noise" etc. and the preamp does indeed need cooling and the transducer does need cooling also, in which case preferably to less than absolute zero :headbash: But then if that were possible the transducer would also stop working.

I suppose that the minimum effective temperature depends on a.) your desired bandwidth and b.) the mobility of your charge carriers :eek:
 
I have not read all this thread so I may be out of line or going over old material but the late John Linsley Hood used a pair of medium size power transistors as low noise input transistors in a number of his MC (and MM) pre-amps. Cica 1985/95. They were BD438//BD437. The circuits are easily accessible at Paul Kemble's collection of JLH schematics. Just Google "Linsley Hood Paul Kemble" and it should turn up.

EDIT: There are two separate schematics. One is about 2/3 of the way down the page and the other close to the end.
 
Last edited:
I have not read all this thread so I may be out of line or going over old material but the late John Linsley Hood used a pair of medium size power transistors as low noise input transistors in a number of his MC (and MM) pre-amps. Cica 1985/95. They were BD438//BD437. The circuits are easily accessible at Paul Kemble's collection of JLH schematics. Just Google "Linsley Hood Paul Kemble" and it should turn up.

EDIT: There are two separate schematics. One is about 2/3 of the way down the page and the other close to the end.

Yes, indeed, and he also designed one or two MC preamps looking somewhat similar but using the transistors in common base mode ( cascode ). Medium power transistors as you suggest do have some of the required characteristics, but also potential problems.

1.) 1/f ( flicker noise) is not usually specified for medium power trannies.
2.) they normally have a low Beta which implies a high in current noise. However a correspondent mentioned the ZTX 653/753 and others in the old Zetex range that are medium power but also have a high beta and so they are indeed worth consideration.
 
www.hifisonix.com
Joined 2003
Paid Member
" have the greatest respect for Analog Devices, I have been using their products since 1976. The reason that 2SB737 etc are not made any longer by Rohm or whoever is that an AD797 is the worlds very best low noise amp at low Rs."

Blackjack, I think the real anser to this is much more mundane! From the early eighties, integrated solutions began to emerge, and ultra low noise transistors found themselves in the company of IC's, even though the IC's were clearly inferior in noise terms.

I've had the great fortune of working in the consumer electronics industry here in Japan ( as a semcinductor supplier) for the past 4.5 years, and I can tell you the drive to get component costs down is absolutely tremendous. Ditto the computing industry out of Taiwan. As to whether IC solutions are a bit noiser . . . If the system cost came down by using an IC, but the noise went up 3-6dB the IC wins.

Having said that, the AD797 is in a class of its own, but I don't think it was the reason the Rohm part was discontinued . . . .
 
Here is a pic of the MC stage currently in use.
The thin black/red cables left are outputs. The thin cables to right are symmetrical inputs. Clip-Leads black=ground, red=pos.rail, white=neg.rail.

Hartmut

IMG_7906_MC_stage.jpg
 
" have the greatest respect for Analog Devices, I have been using their products since 1976. The reason that 2SB737 etc are not made any longer by Rohm or whoever is that an AD797 is the worlds very best low noise amp at low Rs."

Blackjack, I think the real anser to this is much more mundane! From the early eighties, integrated solutions began to emerge, and ultra low noise transistors found themselves in the company of IC's, even though the IC's were clearly inferior in noise terms.

I've had the great fortune of working in the consumer electronics industry here in Japan ( as a semcinductor supplier) for the past 4.5 years, and I can tell you the drive to get component costs down is absolutely tremendous. Ditto the computing industry out of Taiwan. As to whether IC solutions are a bit noiser . . . If the system cost came down by using an IC, but the noise went up 3-6dB the IC wins.

Having said that, the AD797 is in a class of its own, but I don't think it was the reason the Rohm part was discontinued . . . .

I am sure your assertion is spot on.
I have just scrapped a Pioneer CD player recorder that was an award winning design in its time. It has 4558 opamps in the output.
And replaced it with a Marantz SA-KI Pearl lite, purchased just yesterday for £700 ( I know this is slightly off topic :eek: ) and it sounds very nice. The point is I chose it over a £270 Marantz CD6003 and even my genuinely poor hearing could detect a world of difference.

The point is, what on earth is the difference in component quality?. A 4558 might cost a manufacturer 5 pence where a nice 5532 might cost 12p or 15p. To get a toroidal mains transformer with a flux band one has to pay £700 retail. There is a manufacturing conspiracy to cheapen products unecessarily and then wrap it all up in mumbo jumbo. oxygen free silver capacitors and audiophile grade wire, with special polywhatsit capacitors rolled carefully on a dusky maidens thigh.

Back on topic. :) I am just playing with a simple discrete MC input preamp cct. design and will use the 2SA1085's that I have to hand, I will then strap the input with resistors from 3R upwards, put some serious gain on the output and use the audio test set to measure noise at different bandwidths. Next step to put it through an RIAA filter, still with a null input, and do some simple listening, ie. can I hear a rustling sound?.

Should also be able to do a battery power versus conventional LM317 regulator supply.
 
diyAudio Chief Moderator
Joined 2002
Paid Member
ie. can I hear a rustling sound?.

The sensitivity of your loudspeakers and your system's total voltage gain and listening distance plus background quietness will play a major role in that test. Noise in applications is always relative. Best engineering goal should be 'adequate' noise level in any specific application than absolute minimum IMHO.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.