It is, that's why I (and Elektor back then) use BC550, whose noise figure is 0.5-2dB compared to 2-10dB of the BC857.
Lowest current noise, probably. Do you have a circuit handy? I'd like to build it and compare it to the supra.
If one end of the coil is connected to ground (which has to be connected to chassis ground somehow for safety reasons), it is no longer floating. That end will be signal return and shielding at the same time, which is detrimental to CMRR.
I'm gonna use a Dual Salas shunt regulator, which should be pretty hard to beat by anything else.
A small signal transistor is the wrong device for low noise. The only thing paralleling transistors does is reducing the parasitic base resistance. Now the typical base resistance is about 200-500 ohm of a small signal transistor...
The "low noise" in the BC557 really means only low flicker noise.
First if you want to build a MM preamp it does not make sense to parallel transistors as the typical MM coil has about 500 - 1500 Ohm, not counting the inductive part..
And if you use 68 kOhm as emitter resistance any attempt to parallel transitors to reduce noise is completely futile.
Balanced inputs are modern because of marketing. There is no benefit but a 3 dB noise disadvantage and 2x number of transistors in comparison to single ended.
A lot of lazy designers use instrumentation amps as they are really cheap today...
The only problem with single ended configuration is if current can flow in the RCA shield. It is easy to prevent this.
You can look up service manuals of older amps (Philips/Grundig) for simple working circuits. The Holdman preamp or Naim 323 is good too.
The "low noise" in the BC557 really means only low flicker noise.
First if you want to build a MM preamp it does not make sense to parallel transistors as the typical MM coil has about 500 - 1500 Ohm, not counting the inductive part..
And if you use 68 kOhm as emitter resistance any attempt to parallel transitors to reduce noise is completely futile.
Balanced inputs are modern because of marketing. There is no benefit but a 3 dB noise disadvantage and 2x number of transistors in comparison to single ended.
A lot of lazy designers use instrumentation amps as they are really cheap today...
The only problem with single ended configuration is if current can flow in the RCA shield. It is easy to prevent this.
You can look up service manuals of older amps (Philips/Grundig) for simple working circuits. The Holdman preamp or Naim 323 is good too.
Last edited:
The Holman has a FET input for the phono section (I don't like FET's except for small current sources) and the 323 uses five bjts in parallel (322 has a single one).
If a single transistor run at a low current is supposedly superior to several in parallel, why aren't there more low noise opamps around? The venerable 5534 is easily surpassed by the supra circuit, as you may have seen a couple of posts back.
If a single transistor run at a low current is supposedly superior to several in parallel, why aren't there more low noise opamps around? The venerable 5534 is easily surpassed by the supra circuit, as you may have seen a couple of posts back.
It all depends on the source impedance. MM has rather high source impedances where
the current noise is as important as the voltage noise (e.g. 500mH+500Ohm = 6800 Ohm @2kHz).
Paralleling transistors does not decreases voltage noise but only decreases the base resistance,
assuming that the total collector current is constant.
If base resistance is already well below the source impedance nothing
is gained by paralleling. And paralleling degrades other important parameters.
Anyway paralleling small signal transistors is unaware engineering as a single medium power transistor
has a much smaller base resistor right from the start.
With very low MC impedances < 100 Ohm current noise is negligible. Here you must minimize the base resistance of the transistor as it is in series to the source impedance (the emitter resistance too is in series).
The only advantage of Jfets is the neglibible current noise. They shine with source impedances > 5-10 kOhm.
A good low noise MC preamp should have an eqivalent input noise of < 0.4 nV/sqrt(Hz) to get 65 dB SNR with a 100 uV cartridge.
For a MM preamp 4 nV/sqrt(Hz) is enough to get 75 dB SNR with a 3 mV cartridge.
the current noise is as important as the voltage noise (e.g. 500mH+500Ohm = 6800 Ohm @2kHz).
Paralleling transistors does not decreases voltage noise but only decreases the base resistance,
assuming that the total collector current is constant.
If base resistance is already well below the source impedance nothing
is gained by paralleling. And paralleling degrades other important parameters.
Anyway paralleling small signal transistors is unaware engineering as a single medium power transistor
has a much smaller base resistor right from the start.
With very low MC impedances < 100 Ohm current noise is negligible. Here you must minimize the base resistance of the transistor as it is in series to the source impedance (the emitter resistance too is in series).
The only advantage of Jfets is the neglibible current noise. They shine with source impedances > 5-10 kOhm.
A good low noise MC preamp should have an eqivalent input noise of < 0.4 nV/sqrt(Hz) to get 65 dB SNR with a 100 uV cartridge.
For a MM preamp 4 nV/sqrt(Hz) is enough to get 75 dB SNR with a 3 mV cartridge.
Last edited:
That's not quite true. The amplification of the devices adds up directly, while their noise contributions only add up to the square root, because the noise sources are uncorrelated. So for four devices you get 4x gain, but only sqrt(4) or 2x the noise.
But a much smaller hFE, too. That is detrimental for other things like open loop gain (and thus ultimately increases THD). Again, do you have any circuit handy that uses a single medium power transistor in the first stage? I've never seen this done before.
The Holman phono stage looks interesting though and I'm tempted to actually build and try it.
Hi,
Besides the well known SK-JFETs from Toshiba, there are the BF862 from NXP and some other admittedly obsolete parts with en figures below 1nV/sqr(Hz).
Interfet's IF1801 and 9030 rival the best bipolars and the 0.3nV/sqr(Hz) specs of the IF3601 and its Dual the IF3602 are spectactularly low.
The more so as it combines with very high gfs.
What makes the JFETs desirable for useage in MC-Pre gain stages is their missing Gate current, resp. the low leakage current.
A bipolar input stage may require a large valued cap to protect the delicate MC coils from the gate current, which may reach similar orders as the MCs signal current.
JFET inputs don't require a blocking cap.
The combination of low voltage noise and current noise, low leakage and sufficiently high transcondctance makes them rather look superior against bipolars.
If only more, easy to source and less costly devices were around.
jauu
Calvin
I tend to disagree
Besides the well known SK-JFETs from Toshiba, there are the BF862 from NXP and some other admittedly obsolete parts with en figures below 1nV/sqr(Hz).
Interfet's IF1801 and 9030 rival the best bipolars and the 0.3nV/sqr(Hz) specs of the IF3601 and its Dual the IF3602 are spectactularly low.
The more so as it combines with very high gfs.
What makes the JFETs desirable for useage in MC-Pre gain stages is their missing Gate current, resp. the low leakage current.
A bipolar input stage may require a large valued cap to protect the delicate MC coils from the gate current, which may reach similar orders as the MCs signal current.
JFET inputs don't require a blocking cap.
The combination of low voltage noise and current noise, low leakage and sufficiently high transcondctance makes them rather look superior against bipolars.
If only more, easy to source and less costly devices were around.
jauu
Calvin
Last edited:
68 kOhm is not the emitter resistor.....there is no input BJT emitter resistor in this Elektor design.
That marketing argument is getting old....and bal inputs are not "modern" per se.
A balanced MC circuit has its merits, one being that if the complete signal path is true balanced, there is an advantage wrt psu rejection.
Balanced circuit means differential inputs, it doesn't take much to see the advantage of using a balanced connection from coil to input, AND the rejection of the differential input. Without diff input, a balanced connection is not of much use, as any common mode interference is not nulled.
I buy the 3dB noise penalty anytime over a good common mode rejection
Hi,
The 3dB noise advantage applies only to a SE-connected source driving into a SE- or complementary-SE input stage.
jauu
Calvin
Well, most phono-amps working with either discrete or integrated OPAmps feature differential inputs, with the 3dB noise penalty built-in fixed, regardless of their inputs connected in balanced or SE-fashion.
The 3dB noise advantage applies only to a SE-connected source driving into a SE- or complementary-SE input stage.
jauu
Calvin
Medium power bjt's have been mentioned.
Syn08 made some measurements....
http://www.diyaudio.com/forums/analogue-source/154394-hps-4-0-phono-stage.html
Those cheap Fairchild transistors are well suited for the job.
Syn08 made some measurements....
http://www.diyaudio.com/forums/analogue-source/154394-hps-4-0-phono-stage.html
Those cheap Fairchild transistors are well suited for the job.
It depends on the collector current. You can parallel 4 transistors with 1mA Ic or use one transistor with 4 mA Ic....
same result besides the Rb contribution.
One example:
A BC860C has a Rb of 590 Ohm @ 1 mA and en of 3 nV/sqrt(Hz)
A 2N4403 has Rb=17 Ohm @ 10 mA and en of 0.5 nV/sqrt(Hz)
You would need 36 BC860 to get the noise figures of the 2N4403.
The (nonlinear) Miller capacitance would be 90 pF....
In OPV the Ic of the input stage is fixed and you can reduce noise voltage by paralleling BUT with increased current noise.
I know his circuits, but then again he has tried a lot of different topologies with his HPS circuits. Was just wondering why I don't see those medium power bjt input stages more often (in commercial designs, too) than those with small power bjt's, when they're just as cheap and easily available but supposedly perform much better?
As far as I understand, you get about the same result for one transistor with 4mA Ic as for 4 parallel transistors with 4mA Ic (same current per stage), but using 4 transistors in parallel with 4mA Ic each (same current per device) there's more difference than just the Rb contribution.
Do not overestimate commercial builds....
Anyway, there are also small signal BJT's around that have lower rbb than BC550/560.
Word is that BC327/337 are around 30 Ohm, quite high hfe and low early voltage. Only Cob is a bit high, but depending on how it is used of not too much concern. Look up for Joachim Gerhard's findings.
I'm more a 2SK170/2SJ74 guy anyway........
Probably just like most of the guys around here
. Personally I don't like unobtainium or hard to get parts, even if they're the best around. Instead I prefer to make the most out of what's readily available, and the supra circuit seems to accomplish that just fine. Maybe I'll build another one with BC327/337...992 are usually specified @ 30Vce whereas BC are usually specified @ 10Vce.
If you want to compare you need to use the same Vce
The fT of the BC is always higher than the 992 and this may indicate that the Cob difference actually favours the BC.
If you need low input capacitance to suit an MM, then go to cascoded input transistor.
And they look pretty good actually (albeit being three times more expensive than the BCs). 120V VCE would enable the circuit to be run from something like 50V rails for increased headroom . The input capacitance should not be a problem at least with higher damping (ie. a smaller input resistor). So much to try out...
. The input capacitance should not be a problem at least with higher damping (ie. a smaller input resistor). So much to try out...
Andrew is right about the Cob. According to the graphs in the datasheets the ONS BC560C has 2pF at -6.5V, while the Fairchild KSA992 has 2pF around -35V. At -10V they're roughly 1p8 vs 2p8.
Edit: The ONS BC327 has a stated Cob of 11pF at -10V and wouldn't reach 2pF at -100V even if it survived such a voltage...
Edit: The ONS BC327 has a stated Cob of 11pF at -10V and wouldn't reach 2pF at -100V even if it survived such a voltage...
Last edited:
I think that cob isn't such an important issue, as far as it is low enough. Each MM cartridge requires a specific resisitive and capacitive loading to achieve flat frequency response. So *imho* the only thing that matters is that you have to know, or measure, your preamp's input capacity, the capacitance of your cable, then add them - and add a capacitor, if required.
Best regards!
Best regards!
I think it makes sense to try and keep the input capacitance low so we have scope to adjust the loading, both resistance and capacitance have an impact on MM response:
Load the Magnets!!! - [English]
Thanks Andrew, I didn't realise cob varied with Vcb. Using the datasheets, at 15V the cob of the 1845/992 is around 2pf and 3pf for the 550/560, so it would seem we can reduce the input capacitance by around 33%, or am I missing something?
Load the Magnets!!! - [English]
Thanks Andrew, I didn't realise cob varied with Vcb. Using the datasheets, at 15V the cob of the 1845/992 is around 2pf and 3pf for the 550/560, so it would seem we can reduce the input capacitance by around 33%, or am I missing something?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.