The Kuartlotron - keantoken's simple error-correction superbuffer

Ok, using the 1M trimmer without the series 1.8M also works, so I suspect I’m just at the end of the travel for the pot, probably at low R. Would it make sense to try something even higher, like 2.2M for the Series resistor to R9? Is Avery low resistance for R9 a problem?

Sorry, that post should have said “without the 1.8M series resistor...”
 
Ok I think my benchtop psu might have been the issue. I switched over to the intended psu, 10v bipolar using Nisbeths standalone whammy psu boards. Added another 500k in series with the 1.8M one and offset is only a few mV without any trimming. I can play music through it so I guess it works, but I will try the fft to check distortion (though I doubt I can trim much as my interface won’t go below 0.006%.

Sorry for the confusion, not sure why the benchtop psu was giving odd results, it has +/- and gnd taps, but measuring gnd to either + or - showed 0V so something must have been off there...
 
I tried the Kuartlotron in the simulator with AC127/128 models, it does surprisingly well. Unfortunately not all Ge transistors are as useful as these.

The OC70 has a transition frequency of 15KHz. If I heard of a modern transistor with a Ft of 15KHz, I would assume it weighs about 100LB and is used only in special super high power applications.
 
I tried the Kuartlotron in the simulator with AC127/128 models, it does surprisingly well. Unfortunately not all Ge transistors are as useful as these.

The OC70 has a transition frequency of 15KHz. If I heard of a modern transistor with a Ft of 15KHz, I would assume it weighs about 100LB and is used only in special super high power applications.

Thanks for answering Ken :)
(I need Ge buffer, t match with old restored Ge amp. Just to stay in a same "mood"...)
Westrex 2404-C Germanium Amp Restored
But I dont have experience how to transcribe to Ge. Biasing and other things. I am from vacuum tubes world mostly. :(
...
I will try to find some AC127/8
.
I have next Ge-s
.
Type Designator: 2G339
Normally replaceable-slightly different: AC127
Material of Transistor: Ge
Polarity: NPN
Maximum Collector Power Dissipation (Pc): 0.15 W
Maximum Collector-Base Voltage |Vcb|: 15 V
Maximum Collector-Emitter Voltage |Vce|: 10 V
Maximum Emitter-Base Voltage |Veb|: 2 V
Maximum Collector Current |Ic max|: 0.3 A
Max. Operating Junction Temperature (Tj): 100 °C
Transition Frequency (ft): 5 MHz
Forward Current Transfer Ratio (hFE), MIN: 150
Noise Figure, dB: -
Package: TO5
.
Type Designator: 2G381
Polarity: PNP
no datas
.
Type Designator: OC83
Material of Transistor: Ge
Polarity: PNP
Maximum Collector Power Dissipation (Pc): 0.6 W
Maximum Collector-Base Voltage |Vcb|: 32 V
Maximum Collector-Emitter Voltage |Vce|: 20 V
Maximum Emitter-Base Voltage |Veb|: 3 V
Maximum Collector Current |Ic max|: 0.5 A
Max. Operating Junction Temperature (Tj): 80 °C
Transition Frequency (ft): 0.4 MHz
Collector Capacitance (Cc): 100 pF
Forward Current Transfer Ratio (hFE), MIN: 40
Noise Figure, dB: -
Package: TO1
...
thanks again
cheers
.
 
I found some datas for small Ge already have
.
Type Designator: 2G381
Polarity: PNP
Maximum Collector Power Dissipation (Pc): 0.25 W
Maximum Collector-Base Voltage |Vcb|: 20 V
Maximum Collector-Emitter Voltage |Vce|: 20 V
Maximum Emitter-Base Voltage |Veb|: 3 V
Maximum Collector Current |Ic max|: 0.5 A
Max. Operating Junction Temperature (Tj): 85 °C
Transition Frequency (ft): 1 MHz
Collector Capacitance (Cc): 50 pF
Forward Current Transfer Ratio (hFE), MIN: 75
Noise Figure, dB: -
Package: TO1
 
Push-Pull Kuartlotron

Hi Kean,

A young starting DIYer here. I am intrigued by your circuit. Have you ever tried a push-pull buffer like in the attachment? I guess referring to the enumeration from your site, the different Vce of Q1 and Q3 in this arrangement could be suboptimal for the error correction, because for transistors with limited VA the current through R1 and R2 and hence the Vbe of Q2 and Q4 would be different and signal dependant. But on the other hand the input capacitance of Q1 is now bootstrapped, which might help with high frequency linearity, when a high impedance source (read pot) is used. In any case, the buffer can now source and sink high currents by transitioning in class AB and - if I see it correctly - when Q2 and Q4 are thermally coupled as intentioned, the bias can be set by the current through the new input resistors and the ratio between input and output resistors plus, importantly, the bias is temperature stable. The only downside I can see in this regard, is that we will likely suffer from gm-doubling in the class A region due to the new output resistors, since the output impedance of each kuartlotron half is probably near zero from the error correction scheme.
What do you think about it?

Many greetings,
Jason
 

Attachments

  • 100DFC26-8DE9-439A-AF64-BA3D4812D66C.jpeg
    100DFC26-8DE9-439A-AF64-BA3D4812D66C.jpeg
    609.3 KB · Views: 629
Another Kuartlotron lives

Hey there, just finished my second Kuartlotron using Project16s boards. The first one is living with a friend who is reportedly enjoying it alot! I unfortunately forgot to take pictures of that one on completion, but the one i've just built is basically following the same overall layout.


I used a fixed 56 ohm resistor instead of the pot. Offset was 2 and 3 mV without any adjustments of the last pot, so i left it at that. 100 K input impedance, with a 10K pot in front (this was mentioned in a blog post by Nisbeth for this).



The case is a modushop galaxy maggiorato 2U (the largest size). The PT is an old 2x13V from a now-defunct Danish company. I did have some hum from it while testing but am happy to say that in the case it makes no sound. The PSU is a Whammy PSU (circuit from wayne and adapted to single board by Nisbeth) using CRCRC in front of LM7x10 regs.



I used another ebay find for the input relay selector. I bought the bare PCBs and used 5V omron relays. The pot is an ebay DACT type stepped potentiometer.


I built it to serve as preamp to an LM3886 amp also in the works. The power amp will take a while to finish, but i don't mind, because it sounds great with my F6 powering Pensil 10ps! I am using it instead of a BA-3 preamp at the moment but really not missing gain at all. Very nice! Big thanks for Keantoken and also to Project 16 for sharing his layout.


Silas
 

Attachments

  • kuartlotron1.jpg
    kuartlotron1.jpg
    659.2 KB · Views: 489
  • kuartlotron2.jpg
    kuartlotron2.jpg
    699 KB · Views: 476
  • kuartlotron3.jpg
    kuartlotron3.jpg
    522.8 KB · Views: 462
Hi Kean,

A young starting DIYer here. I am intrigued by your circuit. Have you ever tried a push-pull buffer like in the attachment? I guess referring to the enumeration from your site, the different Vce of Q1 and Q3 in this arrangement could be suboptimal for the error correction, because for transistors with limited VA the current through R1 and R2 and hence the Vbe of Q2 and Q4 would be different and signal dependant. But on the other hand the input capacitance of Q1 is now bootstrapped, which might help with high frequency linearity, when a high impedance source (read pot) is used. In any case, the buffer can now source and sink high currents by transitioning in class AB and - if I see it correctly - when Q2 and Q4 are thermally coupled as intentioned, the bias can be set by the current through the new input resistors and the ratio between input and output resistors plus, importantly, the bias is temperature stable. The only downside I can see in this regard, is that we will likely suffer from gm-doubling in the class A region due to the new output resistors, since the output impedance of each kuartlotron half is probably near zero from the error correction scheme.
What do you think about it?

Many greetings,
Jason

Sorry I missed this.

I've tried this. As you say you will have Gm doubling, except unlike a real EF, the Kuatrlotron has no antilog characteristic and so the transition between class A and AB will be a discontinuity that generates a spray of high order harmonics. This is much like many past active AB circuits which are rarely used. You also have each transistor turning on and off at the same time which creates a lot of uncertainty in potential glitching behavior. So it is only interesting really as a class A buffer. You can add keep-alive current sources in various ways but this always disturbs the log-antilog error correction.

So if we settle for it as a class A buffer what we find is that the output bias resistors increase the output impedance, THD might or might not still be vanishingly low, and the parts count has more than doubled. I didn't go this route because it was a lot more work and cost for what may actually be a loss in performance. Keeping the Kuartlotron simple makes it accessible to more builders and gives it a smaller footprint that can fit in more places.

As for the input transisor bootstrap, I considered that too. Notice that the input transistor still sees the output transistor's antilog voltage at it's collector voltage. It's Vce is now also mismatched with the upper PNP transistor, which might not necessarily be that much of an issue but it will run cooler in addition to it's Vbe changing with Vce as a side effect of Early effect. I weighed the effects and decided to err on the side of matching. What you can do if you want some bootstrapping here is to connect the collector's PSU capacitor to the output point instead. Then what you get is matched Vce, but unmatched voltage swing, along with the desired bootstrapping. I wanted to avoid any potential stability or latching issues this might cause as a multitude of builders are likely to uncover any such issue if it exists. There was no clear benefit so I preferred to keep it simple and try to err on the side of reliability as I didn't think the hypothetical performance advantage would matter in the long run. Sacrificing reliability for fractional performance gains is a sure way to lose your mind, and doesn't necessarily even sound better.

Edmund Stuart's ABII output stage is the closest I have seen to a perfect active class AB circuit but even it still has a mathematically imperfect transfer curve with a slight discontinuity, and it still uses feedback to correct the antilog characteristics rather than error correction. But then it gets lenience because it is a power output stage and much harder to linearize in the same way as the Kuartlotron.

Auto Bias part II
 
Hello @ keantoken, I have adjusted the TP terminal to half of the positive voltage, R1 and R2 are 56R, when using Onsemibc550c/560c, the output voltage can not be adjusted to 0, there are tens of MV, and the output voltage will jump around. And the TP terminal voltage will change from time to time 5.8-5.3v. When I add R9 to 4M, the output can be adjusted to 0, but it is also unstable and will change. But when using PHILIPSbc550c/560c transistors, the TP terminal voltage and output voltage are very stable and will not change, but R9 also needs to be added to 4M to adjust to 0v.