DIY Front End 2022

Not very much to do a mistake. All 3 were 250r and were inserted in the same socket. Nothing else changed.

Unfortunately I didn’t do any listening tests but lowering so much the 3rd hd convinced me that I have to spend the money for the caddocks in the line stage project that I want to build in the next future.
Hello Frontend2022-builders,

I finished stuffing the original pcbs, which arrived a few days ago. The J113s from the essentials-kit are nicely matched.
Thanks to Nelsons son.
Now I want to compare the original against my own pcbs (although I think there won't be an audible difference).
This will be the plan for this weekend. I am running them from a symmetrical PSU (linear regulated) at +-20V DC.
Have fun building and listening to this nice circuit!
Dirk :)


  • Frontend2022_measurementJ113.jpg
    136.4 KB · Views: 194
  • Frontend2022_boards_stuffed.jpg
    65.7 KB · Views: 191
  • Frontend2022_challenge_own_pcb_against_original.jpg
    98.1 KB · Views: 190
  • Like
Reactions: 3 users
Joined 2012
Paid Member
Re: resistors and distortion

For me, it would depend on the actual amounts of distortion. Was it 0.01%, 0.02%, and 0.04%, or was it 0.1%, 0.2%, and 0.4%? Or something else?

Also phono preamplifier, line preamplifier, power amplifier, or something else?

And of course, was the difference audible?
It was a line stage, and the 3rd went from 0.003 to 0.00075.
To lower the 3rd in an ops 4 times you have to increase the bias a few times.

I didn’t do any listening tests but I know now what I want(lowest possible 3rd) and the graphs were enough to convince me. I will do listening tests when all will be ready and I will adjust the amount and phase of 2nd, yes in the linestage.

So I preferred to run my amps cooler and spend a few more bucks to solve it in the line stage.

If I got it right all these years, you have to aim for the lowest 3rd this permitting you to alter the second as you wish.

Nelson said somewhere that these amps sound good with high second or high third but not with both high. It never went bad when I followed him.
I'm no longer using the input caps that are shown in my build in this post. I was able to get rid of C1 and C2 entirely and keep the offset at the collector of Q5 within reasonable limits (+/- 1 Volt) by rebiasing Q3 so that the current in Q1 and Q2 is roughly equal. For me, this meant changing R7 from 332R to 1K2. The offset moves around a bit as things come to operating temperature, but not too badly. Still within +/- 1 V.

Another reason to DC couple the input is that the LF response will no longer be dominated by R2/C2, but by C3 and whatever follows it.

I don't know if this buggers any intentional harmonic profile shaping...
Further to post #299,
Here is another idea to enhance DC stability.

Assume a KSC1845 has the same tempco as a KSA992, as bipolar transistors should.
Assume also that the current through Q1 and Q2 is largely equal.
That means current through R7 = 2x current through R6.

One can :
1) place a R9 ~ 0.5*R6 between -Vs and CCS_Q3.
2) replace Q4 with KSC1845, and change R8 to 0R.

This way, Vbe of Q5 will track the Vbe of Q4.
If the two are also thermally coupled, then their collector current should also track each other.
Hence output DC should be close to Zero even as temperature varies.
Of course, an additional small emitter resistor for each BJT will enhance this tracking and prevent thermal run-away.
And NFB will reduce any residual DC further.
Bias is adjusted via R7, DC offset via R8.

This is essentially the same design idea as the SOPA published before :
and has been proven to work.




  • NP DIY FE 2022 DC2.asc
    4.6 KB · Views: 42
Last edited:
I see how that would improve the thermal stability. For me, the trick is how to get the offset to zero in the first place.

Looking from a slightly different perspective, the voltage at the gate of Q2 will be whatever it needs to be to support the current through it, which will be the current through Q3 minus the current through Q1, which is equal to (vbe(Q5) / R6). The offset at the collector of Q5 will then be the voltage at the gate of Q2 multiplied by the ratio of ((R2+R4)/R2) (about 92 in my case with R4 910k). Which is pretty much exactly what Spice says it is. So maybe a trimpot at R7 to get that bias current dialed in initially?
Rigged up the 2022 front end amp into my Wayne's linestage preamp to do some listening. Fast easy way to use existing volume, switching and power supply. Wow, listened to Roberta Flack and I think this is the best that I heard out of my system. Once again Master Pass can make a couple of parts, a simple circuit and twenty bucks blow my ears away. :)


  • Image.jpeg
    524.9 KB · Views: 150
  • Like
Reactions: 4 users
Built a dedicated linestage with these and using the VRDN reg pcb set at =/-20VDC to power it - all I can say is this is the best sounding linestage I have built and listened to bar none, either solid state or tube.

Again the master has given us a very simple and cheap design that performs at 110% - amazing.
So just for fun :

All FET, all SMD, probably can be made to fit a DIP8 footprint.
Great performance even at 15V rails (-97dB H2, -117dB H3).
Even better at 20V, limited by dissipation.
No unobtanium.



PS You need to unzip and put the LIB file in the same directory as the asc file.


  • All FET Frontend 2022.asc
    5.7 KB · Views: 34
    1.1 KB · Views: 29
Last edited: