The Kuartlotron - keantoken's simple error-correction superbuffer

Well, everyone has different design objectives. CFA's can use plain miller compensation, TMC, TPC, double EF, triple EF, beta-enhanced VAS, and so on, and all these probably have a different optimal input stage designs. So really you can't ask "which is the best input stage for a CFA".

It's not as if we can think in universal terms for a bunch of completely different amps, just because they were all thrown into the CFA bin. Each amp needs its own treatment.
 
Last edited:
I am planning to use the Kuartlotron just after a jfet common source high gain amp stage. It outputs 10v DC offset modulated by signal. Maybe I can build the stock buffer with symmetrical psu (so I will loose output cap) but would like to increase it's input impedance to 1mega so I can use a small cap in the input.
1 mega input load is a sure fire way to get DC offset out of anything. Therefore, you can try it if you also install an output cap.

On split rail power, Kuartlotron supports input loads from 10k (and lower) to 100k without an output cap. I'm using 56k so that my sound card won't be dull in tone, but this need depends on the source.
 
Originally Posted by keantoken

That works great with low source impedance, and a source that's willing to drive the entire buffer at 1GHz. But it causes negative input resistance which turns your input cable into an RF transmitter, or at the least cancels the loss of the preceeding wiring and boosts RFI reception.

Negative resistance ! Well that lead to some Very interesting stuff indeed, & not just audio.

The RF transmitter aspect could be good for Phone etc Jammimg puropses :D

The wiring cancellation is intriguing ! Maybe the RFI could be eliminated with a simple R/C filter, and/or inductor ?

Thanx for your observations etc :)
 
Member
Joined 2006
Paid Member
What is signal ground referenced to? If it's on the same supply, I need to know where the reference is. If isolated supply, then I can make a single-rail version without coupling caps. It may be that you will end up needing caps in the signal path if you want to power it from the same supply as your Jfet circuit.

Well, the jfet common source high gain stage is powered from a single rail 35v psu and the signal gnd is the same as the psu. There is an output offset around 10vdc.

I would like to use the buffer after that so if I build the stock version with simetric psu, I will need a coupling cap and would like to minimize it ... that is why I wanted a very high input imp on the buffer.

Off course if you can come up with a single rail version without coupling caps, that would be even better.
 
Well, the jfet common source high gain stage is powered from a single rail 35v psu and the signal gnd is the same as the psu. There is an output offset around 10vdc.

I would like to use the buffer after that so if I build the stock version with simetric psu, I will need a coupling cap and would like to minimize it ... that is why I wanted a very high input imp on the buffer.

Off course if you can come up with a single rail version without coupling caps, that would be even better.

Well, you have 10V offset so you'll need a coupling cap no matter what. I would suggest you use the symmetrical supply version with a large input resistor so you can use a small cap and no output cap. The alternative is to use your current supply dropped to 20V with an output cap on the buffer.
 
The wiring cancellation is intriguing ! Maybe the RFI could be eliminated with a simple R/C filter, and/or inductor ?

You could do that, but the RC is not just a filter, but an RF load on the preceeding circuit. If an opamp has noisy output, and you increase the output load, you may reduce the noise, but you worsen the performance of the opamp by forcing it to drive an RC. This is one reason I chose the tiny input RC snubber instead. This is all that is necessary for a stable circuit; the builder can add RC filtering IF it is necessary or advantageous. In this way I am not adding "unnecessary features with side-effects" to the circuit. This is appropriate for a buffer which is meant to reduce the load on the preceeding circuit.
 
Last edited:
Well, there is good and bad and the circumstances which determine which is which. I decided that for my circumstances, it wasn't viable. I can't speak for other circumstances however, and it would be interesting see if it's useful for other applications.

This circuit does seem like it could be useful for instrumentation. It reminds me of the kind of circuits you find in analog oscilloscope schematics.
 
Kean,

You are saying that we dont need input filter (220R with 150p C). What about compensation ? In the Mpp thread, we saw a schematic where 22R with 390p cap compensation was shown. When should we use that? In you latest schematic, that is missing.

I am also thinking about the inductor. If you have the input filter, I think the inductor may not be necessary, am I right in saying that? I am planning on building this buffer since it looks simple (need to order some parts though) and hence these questions.

Thanks
 
We made this measurements on my build.
I thermo coupled the transistors.
They are selected for HFE.
I used fixed resistors and did not do a distortion trim.
I can not remember if the TKD pot was in circuit.
I check that today.
 

Attachments

  • KeanTokenJG_Bode.pdf
    11 KB · Views: 364
  • KeanTokenJG_THD_vs_Ampl.pdf
    9.6 KB · Views: 388
  • KeanTokenJG_FFT(1).pdf
    18.6 KB · Views: 325
  • KeanTokenJG_THD_vs_Freq(1).pdf
    13.8 KB · Views: 373
Kean,

You are saying that we dont need input filter (220R with 150p C). What about compensation ? In the Mpp thread, we saw a schematic where 22R with 390p cap compensation was shown. When should we use that? In you latest schematic, that is missing.

The "latest" schematic is on the Kuartlotron webpage in my signature. I posted one special (but not better) schematic for RCruz and Dan exhumed the old schematic with the suboptimal compensation. Build the one on the site; it is the best so far.

I am also thinking about the inductor. If you have the input filter, I think the inductor may not be necessary, am I right in saying that? I am planning on building this buffer since it looks simple (need to order some parts though) and hence these questions.

The inductor shouldn't be left out, else you get RF peaking regardless of any input filter. It's not just a filter.
 
We will try distortion trimm at 2V 1 kHz next.
Douglas Self found distortion in the TKD pot so we should also measure without.
Notwitstanding the strange step at higher voltage the buffer behaves monotonic, a good sign i think.

Can you elaborate more on the comment that TKD pot has distortion please? How much? why? Should I be worried since I've used one?
 
Member
Joined 2006
Paid Member
Well, you have 10V offset so you'll need a coupling cap no matter what. I would suggest you use the symmetrical supply version with a large input resistor so you can use a small cap and no output cap. The alternative is to use your current supply dropped to 20V with an output cap on the buffer.

Perfect.

Can the buffer be built with a jfet on the input ? That way we could avoid the input cap :)
 
Hi,

@Keantoken
certainly a design I like, as one could use two matched Duals or one matched Quad (That340, though they might generate considerable output offset).
I have difficulties though to follow Your explanations of how its functioning.
For the basic structure (You named it Tringlotron) You claim identical Vbes and Ics.
The only identical Vbes are in fact those for the current mirror.
The Vbes of Q1 and Q2 certainly differ.
Also all 4 Ics differ if You account for the base currents.
Introducing emitter resistors in the CM and their Vbes won´t be identical too.
What difference could be expected when using a PNP-CM instead?
I´d expect even slighty lower THD for that one and probabely lower thermal problems!

With the Kuartlotron I basically don´t understand what You call error correction here?
Is it just another term used for positive feedback or for the partial cancellation of the Vbes?
As the Base of Q4 is modulated signal dependent, and hence the base of Q3 there certainly is (positive) feedback, which might have been the source of Joachim´s circuit instability.

What Ya say?? ;)

jauu
Calvin