Actually the tail current is above the optimum for moving-magnet cartridges. Something of the order of 70 uA per pair would be better, but then the loop gain gets even worse.
max dissipation BC560, 625mW, BC860 200mW. fT is different too - 250MHz for the BC560C and 100MHz for the BC860C. Noise, hFE are both worse on the BC860C.
On the plus side they have much better hFE/Ic performance
Anyway they are not part equivalent by a long shot.
On the plus side they have much better hFE/Ic performance
Anyway they are not part equivalent by a long shot.
2SA1312/2SC3324 are surface mount absolute equivalents (the spec sheet is identical) to the 2SA970/2SC2240 . Current Toshiba production with tens of thousands in stock by the usual Mouser-like sellers.
OK, this is getting interesting.
I got the lib files of those transistors, from Toshiba for Pspice, and converted them to what I think should be working models for LTSpice.
.MODEL 2SA1312_BJT PNP(LEVEL=1 IS=2e-013 BF=415 NF=1 VAF=60 IKF=0.2 ISE=1.05e-009 NE=5 BR=10 NR=1.02 VAR=8 IKR=0.0001 ISC=1.5e-011 NC=1.3 NK=0.5 RE=0.5 RB=0.2 RC=2 CJE=10E-012 VJE=0.75 MJE=0.33 CJC=1.0176e-011 VJC=0.79167 MJC=0.319 FC=0.5 TF=900E-012 XTF=10 VTF=2 ITF=500E-003 PTF=0 TR=10E-09 EG=1.11 XTB=1.55 XTI=3 TRC1=0.005 TNOM=25)
.MODEL 2SC3324_BJT NPN(LEVEL=1 IS=1.3e-013 BF=475 NF=1 VAF=200 IKF=0.2 ISE=3.5e-014 NE=1.3 BR=28 NR=1 VAR=10 IKR=0.004 ISC=1.5e-010 NC=1.5 NK=0.5 RE=0.25 RB=2 RC=0.02 CJE=3.0625e-011 VJE=0.75 MJE=0.33 CJC=4.9687e-012 VJC=0.9247 MJC=0.1914 FC=0.5 TF=3e-010 XTF=50 VTF=5 ITF=0.1 PTF=0 TR=10E-09 EG=1.11 XTB=1.5 XTI=4 TRC1=0.05 TNOM=25)
Haven't yet tried them, but they might interesting options for an "SLN inspired" version with high voltage transistors, not sure if low noise enough, that you can find today.
I think I mentioned the NPN/PNP low noise pairs 1 + 1 and 2 + 2, that could also be used on an SLN version, though not sure about the best output pair to use.
What do you think about such options. considering BC550 and BC560 replacements do not exist anymore?
I got the lib files of those transistors, from Toshiba for Pspice, and converted them to what I think should be working models for LTSpice.
.MODEL 2SA1312_BJT PNP(LEVEL=1 IS=2e-013 BF=415 NF=1 VAF=60 IKF=0.2 ISE=1.05e-009 NE=5 BR=10 NR=1.02 VAR=8 IKR=0.0001 ISC=1.5e-011 NC=1.3 NK=0.5 RE=0.5 RB=0.2 RC=2 CJE=10E-012 VJE=0.75 MJE=0.33 CJC=1.0176e-011 VJC=0.79167 MJC=0.319 FC=0.5 TF=900E-012 XTF=10 VTF=2 ITF=500E-003 PTF=0 TR=10E-09 EG=1.11 XTB=1.55 XTI=3 TRC1=0.005 TNOM=25)
.MODEL 2SC3324_BJT NPN(LEVEL=1 IS=1.3e-013 BF=475 NF=1 VAF=200 IKF=0.2 ISE=3.5e-014 NE=1.3 BR=28 NR=1 VAR=10 IKR=0.004 ISC=1.5e-010 NC=1.5 NK=0.5 RE=0.25 RB=2 RC=0.02 CJE=3.0625e-011 VJE=0.75 MJE=0.33 CJC=4.9687e-012 VJC=0.9247 MJC=0.1914 FC=0.5 TF=3e-010 XTF=50 VTF=5 ITF=0.1 PTF=0 TR=10E-09 EG=1.11 XTB=1.5 XTI=4 TRC1=0.05 TNOM=25)
Haven't yet tried them, but they might interesting options for an "SLN inspired" version with high voltage transistors, not sure if low noise enough, that you can find today.
I think I mentioned the NPN/PNP low noise pairs 1 + 1 and 2 + 2, that could also be used on an SLN version, though not sure about the best output pair to use.
What do you think about such options. considering BC550 and BC560 replacements do not exist anymore?
That is absolutely true.
Other options include the 4401/4403 workhorse low noise transistors, or the ZTX E-line devices from Diodes Inc (was Zetex). Again generally lower noise than the BC550C/560C.
But the precise choice here depends on whether it is for a MC or MM catridge. Transistor choice is driven by that.
Other options include the 4401/4403 workhorse low noise transistors, or the ZTX E-line devices from Diodes Inc (was Zetex). Again generally lower noise than the BC550C/560C.
But the precise choice here depends on whether it is for a MC or MM catridge. Transistor choice is driven by that.
My quest is for a MM RIAA preamp. So I wonder which transistor might be better. Opinions please.
I'm not sure if that is something, MM input load, that can be simulated accurately and get realistic results.
I did try the ZTX851/ZTX951 on the sims, but not with several in parallel, as with the original SLN. These transistors are not cheap either.
But those 2SA1312/2SC3324 are cheap, and it might be worth simulating them in parallel.
But there's been so many CCS tail resistors being used, that I do not know which ones to try.
Suggestions, please?
I'm not sure if that is something, MM input load, that can be simulated accurately and get realistic results.
I did try the ZTX851/ZTX951 on the sims, but not with several in parallel, as with the original SLN. These transistors are not cheap either.
But those 2SA1312/2SC3324 are cheap, and it might be worth simulating them in parallel.
But there's been so many CCS tail resistors being used, that I do not know which ones to try.
Suggestions, please?
For MM low Cob & high Hfe should be prioritized I believe. We don't want much input capacitance via Miller for MM, it interacts altering response, not much base current either if for DC input coupling. Because highish Ib into a high Zo generator can develop enough DC offset voltage across it.
Rbb' should also be good but not necessarily in champions self noise territory. The MM signal is high enough. Very low Rbb' types are usually low Hfe & high Cob.
Thus in your case the SMD 2SC/2SAs look like OK picks considering all that. In BL(L) Hfe rank. IMHO of course.
Rbb' should also be good but not necessarily in champions self noise territory. The MM signal is high enough. Very low Rbb' types are usually low Hfe & high Cob.
Thus in your case the SMD 2SC/2SAs look like OK picks considering all that. In BL(L) Hfe rank. IMHO of course.
Use cascode to achieve low input capacitance. Then go supersymmetric to cancel distortions and also get an improved noise performance (because the PNP/NPN input arms are essentially in parallel). Ideally with transistor hFE matching.
If we are talking the original 2006 phono stage at the start of this thread, well yes.
But with a blank sheet of paper, and leaving discrete aside for a moment, the LM4562 has a noise equivalent input resistance of 440 ohms. Which is amply good enough for a MM cartridge (about 80dB below 5mV).
I'll get my hat and coat....
But with a blank sheet of paper, and leaving discrete aside for a moment, the LM4562 has a noise equivalent input resistance of 440 ohms. Which is amply good enough for a MM cartridge (about 80dB below 5mV).
I'll get my hat and coat....
MM cartridges being very high impedance need low current noise, the NE5534A stomps all over the LM4562 for this application.
Noise-wise, the only advantage of paralleling transistors is that it reduces the effect of the base resistance. The base resistances of the 2SA1312 and 2SC3324 are so low that for moving-magnet applications, the base resistance thermal noise will be far from dominant anyway, so parallel connections are not really useful.
The optimal tail current depends on the cartridge inductance, hFE, and what kind of weighting you are interested in (for example RIAA + A or RIAA + ITU-R 468), but assuming a single pair of NPNs and a single pair of PNPs with hFE values around 400, the order of magnitude is 70 uA. You may be forced to use a somewhat higher current to get more loop gain, though.
By the way, the base resistances in discrete transistor SPICE models are often completely off, so noise simulations are not necessarily reliable.
Assuming the cartridge load is 1k (they are all there or thereabouts). This analysis ignores the inductance, which modifies the simple calculation below
NE5534 has vn = 3.5nV/rootHz and in = 0.4pA/rootHz. The net (root sum of squares) effect with a 1k resistive input is 3.52nV/rootHz input referred noise
LM4562 has Vn = 2.7nV/rootHz and 1.6pA/rootHz. That results in 3.14nV/rootHz.
So the LM4562 is 1dB quieter - for a resistive load.
In practice the large cartridge inductance of ~400mH modifies those simple calculations significantly.
When that is taken into account, you are right that the old-school workhorse of the NE5534A turns out to be around 5dB quieter than the LM4562, which is a significant performance advantage.
NE5534 has vn = 3.5nV/rootHz and in = 0.4pA/rootHz. The net (root sum of squares) effect with a 1k resistive input is 3.52nV/rootHz input referred noise
LM4562 has Vn = 2.7nV/rootHz and 1.6pA/rootHz. That results in 3.14nV/rootHz.
So the LM4562 is 1dB quieter - for a resistive load.
In practice the large cartridge inductance of ~400mH modifies those simple calculations significantly.
When that is taken into account, you are right that the old-school workhorse of the NE5534A turns out to be around 5dB quieter than the LM4562, which is a significant performance advantage.
Yes, that would mean complete redesign, something I'm not really interested in.
The most I would like to try is higher supply voltage, which these transistors allow, and a DC servo to cut offset. And see how it sounds.
A better supply than the original SLN might be bring audio improvements too.
Sawyers, as a reasonably accurate rule of thumb, you can take the cartridge impedance at 3852 Hz (RIAA + A weighting) or at 5179 Hz (RIAA + ITU-R weighting) and plug it into the kind of calculation you just did.
There seems to be a lot of claimed Fairchild BC560 on FleaBay.
But they're priced cheaply, so I bought a few.
Perhaps this soon after announced obsolescence
they're manufacturer overstock?
And how do you know they are legit?
You also need BC550 for this project.
I do not like SMD parts, but I'll try these new options if necessary.
You also need BC550 for this project.
I do not like SMD parts, but I'll try these new options if necessary.