Two weeks ago i came across a Fisher ca-2030 amplifier which happened to have inside a downsized version of the mm phono preamp in Fisher CC-3000 ,but also strikingly looking like the APT-Holman phono preamp, with fet transistors in the feedback path, instead of bipolars.Well everybody can have a quick check on how both APT Holman and Fisher cc3000 measure and they show wonderful specs .I don't have the guts to say that i did a better preamp than any of them.Here we have just some simulations!
I thought of trying to make something that have something that each of the preamps i liked had.
So i got the bipolar feedback of APT -Holman for higher transconductance thus more effective THD reduction, lower noise with a lower impedance feedback network as Fisher's , the cascode and the low impedance feedback network of the Fisher preamp to be able to easily get the VAS transistor in a state of quasi saturation because of the cascode inability of giving enough current to VAS transistor base, the final stage of Kenwood l-02a GOD for lower output impedance and you know who's folded cascode VAS
On top of it i tried to take advantage of Mark's notes on Va and quasi saturation effects in transistors.
Unfortunately i don't own any mpsa18 nor did i have a model for it at hand so i tried something that i have in abundance, 2n3906 for the VAS transistor that will HOPEFULLY allow this preamp to gradually saturate when the cartridge hits the Dust on the vinyl.
I also got an unexpected help from somebody on Linkedin who suggested me how to feed the input from a simulated Shure v15type3 like cartridge as i'm not good enough with simulations.
So the original Fisher ca 2030 preamp wasn't able to sim the same at +-17.5v supply which happens to be common with middle class phono preamps. It used +-22.5v while the full sized version in the CC-3000 preamp was supplied at +-25.2v.
Now my preamp can take easily +-35v which allows full amplification of any tick and pop and doubling the gain without compression or clipping, but needs some components to be changed in values to spec well and 3 other transistors for higher max VCE.
Of course , at +-22 or 25v the behavior is much better with very minor changes, but i was aiming at feeding any 30w amp, not just a 170w amp as a real combo like Kenwood l-02 that can handle anything at its input..
I thought of trying to make something that have something that each of the preamps i liked had.
So i got the bipolar feedback of APT -Holman for higher transconductance thus more effective THD reduction, lower noise with a lower impedance feedback network as Fisher's , the cascode and the low impedance feedback network of the Fisher preamp to be able to easily get the VAS transistor in a state of quasi saturation because of the cascode inability of giving enough current to VAS transistor base, the final stage of Kenwood l-02a GOD for lower output impedance and you know who's folded cascode VAS
On top of it i tried to take advantage of Mark's notes on Va and quasi saturation effects in transistors.
Unfortunately i don't own any mpsa18 nor did i have a model for it at hand so i tried something that i have in abundance, 2n3906 for the VAS transistor that will HOPEFULLY allow this preamp to gradually saturate when the cartridge hits the Dust on the vinyl.
I also got an unexpected help from somebody on Linkedin who suggested me how to feed the input from a simulated Shure v15type3 like cartridge as i'm not good enough with simulations.
So the original Fisher ca 2030 preamp wasn't able to sim the same at +-17.5v supply which happens to be common with middle class phono preamps. It used +-22.5v while the full sized version in the CC-3000 preamp was supplied at +-25.2v.
Now my preamp can take easily +-35v which allows full amplification of any tick and pop and doubling the gain without compression or clipping, but needs some components to be changed in values to spec well and 3 other transistors for higher max VCE.
Of course , at +-22 or 25v the behavior is much better with very minor changes, but i was aiming at feeding any 30w amp, not just a 170w amp as a real combo like Kenwood l-02 that can handle anything at its input..
Attachments
Last edited:
Indeed, nothing can beat a true valve at saturation capabilities
the real world ksc945/ksa992 are just 50ma transistors and they also have some quasi sat capabilities... bc550c and bc 337-40 are very linear and much better for current sinks for the VAS , but i don't know much about the pnp complementary transistors .2n3904 and 2n3906 behave well too as having a H2>H3 pattern.
The simulation algorithms are actually pretty good (although the leave out some things like Zener noise).
The individual models are indeed much more problematic, with accuracy all over the place.
These are the 2N3904 models I have:
The first one appears to have hFE maxed out, while the second one appears more reasonable (although it has an incorrect Icrating)....
The individual models are indeed much more problematic, with accuracy all over the place.
These are the 2N3904 models I have:
Code:
.model 2N3904 NPN(IS=1E-14 VAF=100 Bf=300 IKF=0.4 XTB=1.5 BR=4 CJC=4E-12 CJE=8E-12 RB=20 RC=0.1 RE=0.1 TR=250E-9 TF=350E-12 ITF=1 VTF=2 XTF=3 Vceo=40 Icrating=200m mfg=Philips)
.MODEL 2n3904T NPN(IS=4.1799f BF=215.38 VAF=79.874 IKF=.88689 ISE=4.2698f NE=1.9988 BR=4.8050 VAR=100 IKR=.54262 ISC=2.5346p NC=1.4324 NK=1.3890 RB=14.616 RC=.53092 CJE=9.4850p MJE=.33333 CJC=7.6517p MJC=.33333 TF=481.51p XTF=10 VTF=10 ITF=1 TR=124.93n Vceo=40 Icrating=600m mfg=Philips)The first one appears to have hFE maxed out, while the second one appears more reasonable (although it has an incorrect Icrating)....
Set up a jig to compute working hFE as current is swept. You typically find that working hFE never comes up to the Bf number. Another parameter falls-off gain at high current, and yet another models low-current fall-off/leakage/recombination effects.
Here's how an ancient model plots out. The notional high Bf is 400+, but the low and high current coefficient ranges overlap so that the peak hFE is 181.
I would not trust this too far either!! Processes have improved since the '3904 hit the showroom. I'd expect the whole gain-curve to be much flatter. If I found a '3904 with hFE of 48 at 1uA today I'd set it aside for fuzz-box trials.
Attachments
In the Art of electronics from where Mark started the VA vs quasi sat effect that gave me wings for this small mods, 2n3904 graph is one of the most linear of all at Vce=5v while 2n3906 is the complete opposite.There's also a discussion in the Comparators section over the antisaturation techniques used with LM311 .Rescaling the base currents into 2n3906 to be as low as possible with increase in currents drawn by 2n3904 CCS when VCE of 2n3904 is too low and consequently PNP transistor's VCE is low too, might give me that most sought compression effect at high input generated by dust on vinyl so that i can safely have lower supply voltages.I'm still simulating various combinations in order to understand what i really need to get my effect done.
and something really interesting...with a bias setting the output off the center i get better results with lower input , with centered output i have better results at the highest input but not so great results at lower input signals .
Attachments
tried that multiple times including now...thd worse by 2db for h2 or both h2 and h3 depending on level.
Arto Kolinummi said in his book it's better this way and i just copied Kenwood l-02 output stage.Even tried different resistors values, different transistors, the mix is perfect as it is.
been wrong the whole time on the cart simulation side ...now corrected, but the results aren't much different.The highest distortions appear with an input of 270mv ....For +-17.5v rails that is a lot...
Attachments
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.