Zen -> Cen -> Sen, evolution of a minimalistic IV Converter

hi all,

i remember having asked Patrick (EUVL) some time ago about substituting jfet with mosfet in his I/V circuit but apart from remembering hew did not recommended it, i don,t know exactly why the performance would be inferior.

Is it because of gate leakage current and capacitance linearity?

because, apart from added biasing circuit PSU,probably higher voltage rquirements, and maybe higher capacitance (with lower linearity depending) on the mosfet pair, i don't see what else could deter anyone from trying it.....but then again i'm not a pro :)

PS: i have a pcm1794 with high swing current and i thought about using 2sk2013/j313
 
Hello,

I quickly implemented my first version of the SEN "to try" on a prototype PCB directly (PCB are very cheap nowadays) and using SMD parts I had in my drawers.
My DAC is the PCM1794 as well and I am struggling with the cancellation of the 0dBFS current of the DAC (6.2mA).
I implemented a 2N4391 in current source mode but I am having issues with the thermal stability, it can quite easily drift over time and not very stable.
I am using 3S LiPo battery.

So I would appreciate if someone had a good recommendation about a low tempco stable current source?

Beside I was able to stabilize the DAC DC offset to a value not too far from 0 and did some listening on one channel only as I haven't assembled the second one since I want to complete the debug on the first one.
I would like to say that at a first listening the SEN seems to sound very good.
I appreciated very quickly the dynamics and low response impact. Was very good. I need to debug & complete the two channels for a more complete listening.

So I am looking for some help on this current source as I am quite excited now to enjoy this SEN completely :D

Here is how it looks like:

27301341058_8809eedce5_b.jpg
 
Does your voltage varies across the source resistor of the jfet, thus indicating the drift?

If yes try this one attached with small signal BJTs.

should perform better as well (maybe a bit more noisier depending on bjt).

Vref is ground,
0V is a negative (say -10V)
Iout is the dac current output



(maybe someone more experimented then me has a better alternative/advice?)
 

Attachments

  • bjt_ccs.jpg
    bjt_ccs.jpg
    10.7 KB · Views: 609
Last edited:
Hi,
Coincidentally, I've built a prototype based on this very good idea to null the dc offset on a pcm1792a-same as pcm1794 and I'm testing it these days. I'm using it without the rest of the I/V circuit for now. BF861C seems to work fine.
 

Attachments

  • ItoV conversion schematic.JPG
    ItoV conversion schematic.JPG
    20 KB · Views: 596
Thanks for all your answers.

Patrick, one question since it is apparently a bit difficult to source the 2SK246. As a replacement should I understand that we preferably look for a high Vp / low Yfs JFET here?
Would the 2SK208 be a possible replacement? (it has a lower current but I coud add more devices in parallel).
I could try as well to couple them thermally.
 
Unfortunately the 2SK208-GR grade seems as well difficult to order.
On the other side, the datasheet of the 2SK879 seems strictly identical, not sure why they have a different part number but this one is available in GR.
Is it another SMD version of the 2SK246? Should be fine to use as well right?
 
To be honest i did not tested it for this specific application, but it is a very common CCS used widespread.

Nevertheless it should give better AC rejection than a simple jfet based CCS if stable thermally.

i actually plan to try this setup on my own CEN+pcm1794+bjt css when i get back at it.

The default JFET CCS is a very acceptable performer, especially given it's part count, but not the last word in AC rejection.

before i put my project on the ice some time ago, i have put in a bsp129 depl. mosfet to give me an idea of the importance of AC rejection for the CCS part and it was significantly better, though much more noisier and having more capacitance.
 
Hello,

After several hours of trial I was able to get something quite stable using J111 I had in my drawer as current source (while waiting for my order of 2SK879).
The main issue is actually not too much the temperature drift but more the sensitivity of the SEN to leave quickly 0DC position as soon as a little current offset is present at the input introducing then distorsion.
It is not easy to adjust given that I use the DAC in mono mode and I+/I- bias adjustment have an influence to each other.

Still I was able to assemble a full stereo SEN quite stable and with very low distortion measurements.

I confirm my first impressions, this SEN is really really good. I like it a lot, many thanks again Patrick for sharing this with us ! :):up:
 
are the 2sk170 performing the conversion coupled togheter as pair or joined by the heatsink?

a heatsink will help to keep the temperature stable but won't help compensate anything, which is what is desired as far as i know.

Maybe also a small source resistor could help stability...?

i have tried the SEN as phase splitter and had that same issue, but they weren't coupled and didn't have time to fiddle with it.

However, the CEN (3 pairs) i talked about wasn't coupled and didn't drift (with jfet ccs and pcm1794)
 
Last edited:
Actually this first proto is using JFET I had in my stock.
I am not using 2SK170, I am using 2SK2145, already matched and I use them in parallel to handle the 1794 large current.
You are right the heat sink here is only to couple all the parts thermally and not to compensate anything.
Later on I possibly plan to try with BF862 that I have as well but not matched yet.

You mean a source resistor on the 2SK2145 would help?

The circuit is relatively stable but very tricky to adjust at the moment.
Best setup I found for the current source is following what Patrick was saying paralleling JFET with a large source resistor.
I also increased the negative voltage applied on the current source.
Right now using J111, better than 2N4391 which are better than BF862 (very unstable as current source).

I appreciate your help as I am still learning a lot here.
 
you can try the source resistor on the 2sk2145 it but i'm unsure it would do anything for jfets...better to look for someone else advice. 2sk2145 is already coupled so no need firsthand for heatsink, unless they are running above or close their rating, but should not hurt either

In any case, this circuit seems very temperature sensitive.

Maybe it would be a good idea to attach the CCS transistor(s) to a heatsink too, if not already done.
 
Last edited:
Ok now we have something really working well.
CCS could not do the job well alone, it was always drifting out after some time because the DAC DC current itself is changing a bit over time and temperature.
So I moved to a servo controlling the CCS to constantly track any change in the DAC DC current and compensate accordingly.
After lots of simulations and lots of prototyping here is finally a version that is working very well.
It uses a new JFET from ON Semi with a 0 tempco current equivalent to the PCM1794 DC bias current in mono mode and a servo to track and compensate any changes in this bias.
The CCS is now below the heatshink like the SEN jfet to couple all this thermally.
The servo is below the PCB.

40382779620_b86601106a_b.jpg

41468865244_56441cacb4_b.jpg


This is very stable over time, no more drift, excellent THD measurements and I confirm again very good listening experience.
Thanks a lot Patrick for sharing this with all of us.

Long time I was looking at implementing in the right way this SEN with the PCM1794A. Now I think I finally get it :)