Hi John,
Thank you.
Since JFETs actually don't care about reversing D and S, can I replace both 2SK147Vs and 2SJ72Vs with 2SK170Vs so that there will be 4 2SK170V in each channel?
Let's find the problem, first. The 'freeze spray' (instant cold) technique is one that we use all the time for intermittent semiconductors. First, when everything else has been checked, like the caps, the board thoroughly cleaned with alcohol or an equivalent polystyrene recommended wash, then with the circuit making noise, just spray the whole stage, first, to see if things improve. If so, then let the cold dissipate and touch each active component with the spray, and hope to find the bad one. Personally, at this point, I think your problem is somewhere else. Clean first, check the caps, THEN worry about the fets.
It is difficult to be sure about 'doubling up' on the jfets to create either a j72 or a k147. A careful reading of the data sheets, shows that I had guessed wrong about the Gm differences between the k147 and the k170, etc, and you will have to replace BOTH complementary devices with doubled up jfets, just to keep the circuit in reasonable balance. Speculation is not very useful here, so clean the circuit boards and check the caps first.
It is difficult to be sure about 'doubling up' on the jfets to create either a j72 or a k147. A careful reading of the data sheets, shows that I had guessed wrong about the Gm differences between the k147 and the k170, etc, and you will have to replace BOTH complementary devices with doubled up jfets, just to keep the circuit in reasonable balance. Speculation is not very useful here, so clean the circuit boards and check the caps first.
I use the fixture for fet quad (sockets) + VAS and I measure noise by spectral analysis - to identify fakes.
No, the transconductance parameter goes with the geometry, just look at John's plots (the lower right on the first one and the middle right on the second one). For any Idss at a given Id the transconductance is the same especially at lower currents. The point at which the input pair will null seconds and have zero offset will be at different Idss values for the N and P (with a slight trim of the degeneration resistors) and can be computed with a little simple algebra.
The 2S170 plot is most telling, at 2mA all 2SK170's will have 20mS of gm within a small distribution.
Last edited:
Hi John,
Thank you.
All you suggested doing prior to replacing FETs is what I planned doing, even before posting the initial post here.
First, all the electrolytic and polystyrene capacitors will be removed and the said small PCBs will be thoroughly cleaned.
When the capacitors and rectifying diodes will arrive from Mouser, within about a week, all electrolytic capacitors in the PSU and in the whole preamp will be replaced. On the said small PCBs, also all polystyrene caps and a MOSFET with broken leads will be replaced.
Next I'll operate the preamp and see whether the phono input is still noisy; if yes, I'll try to locate the noisy FETS with freeze spray.
Should there be some noisy JFETs, I'll replace SK and SJ pairs, or consult here again.
Hi Pavel,
Thank you.
A fixture to test JFET quads noise is an excellent idea for new ones.
In my case, I'll try locating the noisy ones without removing all JFETs from the PCBs and possibly solder them back. Should the freeze spray wouldn't pinpoint the noisy ones, I'll try to pinpoint them by using a scope at the drains of the left arm and the output, which is the sources of the right arm.
Only if both the above measures will fail to pinpoint the noisy ones, I'll set a fixture to test each one of them individually for noise.
Hi Scott,
Thank you.
Please assist poor me, who is mathematically challenged…
How do I compute so that N and P pairs will have equal transconductance?
Selecting N and P pairs (actually, quads in my case, since I'll replace each V grade JFET with 2 BL grade ones), with slight Idss difference should be doable for me, having about 100 of each. Also, changing, or trimming, the degeneration resistors is a piece of cake for me.
All I need knowing is how to select the proper Idss difference and how to compute the proper values of the degeneration resistors.
Different people (including different designers and manufacturers) have different engineering and sound quality demands and standards.
Indeed, below 1% of second harmonic distortion may not be audible, or may not degrade the sound quality much - especially in a phono stage.
However, distortion of each stage is accumulated at the output, so it's better to take reasonable measures to minimize the distortion of each and every stage.
What may be more important is that the THD and the harmonic spectrum, or dispersion (and gain and frequency response) of the L and R channels will be as close to one another as practically possible.
So, being able to select perfect Idss in each N and P pair and between N and P quads for minimum second harmonic distortion and better gain match between channels, why not do it?
I, for one, would like to do it and am willing to 'pay' in the time it will take to measure Idss of a couple of hundred devices.
BTW, selecting different Idss values between N and P pairs, plus having different degeneration resistors values, in order to adjust for equal Gm in the N and P side of each arm isn't new to me. It was discussed few times in the forums here. I only don't know how to calculate it.
Joshua, this is a simple gain circuit. Matching Idss to 1-2ma is enough. However, pair matching is important as well as absolute range of Idss. I suspect that the original devices are V select. This is about 2 times Bl select. This must be considered to get enough working current in the input stage, without changing the load resistors as well.
Hi John,
Thank you.
The existing JFETs are definitely V grade. Should they need to be replaced, I'll replace each N-P pair with N-P quads of BL grade, in N and P pairs connected in parallel. It seems like I have enough devices to attain fairly close Idss match.
I don't intend to change any resistor value, unless the source degeneration resistors will need to be trimmed between the N and P sides in order to match Gm, which will happen should I'd be assisted in how to calculate the necessary trim.
P.S.,
Indeed, it is a simple gain circuit, only, beautifully executed…
(Folded cascode with the same devices may be better, however I'm not going to redesign anything)…
Once that stage will be as quiet as it should be, I'm going to compare it (sound quality wise) with fairly good MC step-up transformers I have.
Also, I'm going to compare the sound quality between those modules and the existing line stage.
Just to satisfy my never ending curiosity…
However this is entirely different topic, unrelated to the present thread.
Indeed, it is a simple gain circuit, only, beautifully executed…
(Folded cascode with the same devices may be better, however I'm not going to redesign anything)…
Once that stage will be as quiet as it should be, I'm going to compare it (sound quality wise) with fairly good MC step-up transformers I have.
Also, I'm going to compare the sound quality between those modules and the existing line stage.
Just to satisfy my never ending curiosity…
However this is entirely different topic, unrelated to the present thread.
Joshua:
Patrick (Euvl) has shown some results of applying slightly mismatched source degeneration resistors and selecting slightly mismatched Idss to help out with transconductance difference between N & P jfets if you have concerns about it. I think it's in his thread on a balanced F5 design?
mlloyd1
Patrick (Euvl) has shown some results of applying slightly mismatched source degeneration resistors and selecting slightly mismatched Idss to help out with transconductance difference between N & P jfets if you have concerns about it. I think it's in his thread on a balanced F5 design?
mlloyd1
Hi Lloyd,
Thank you.
Actually I have on my computer for a long time a pdf file by Patrick, dated August 20, 2010, where he shows that for F5 input stage, using there 2SK170BL and 2SJ74BL, where in the original F5 schematic both have 10R degenerative resistor, when choosing 2SJ74 with slightly higher Idss than 2SK170 and adding 5R to the 2SJ74 degenerative resistor, which is now 15R instead of 10R, the 2SK170 and 2SJ74 match in forward gain.
This is really thrilling, however I do need to find a way how to find the differences in Idss and the degenerative resistors in other cases, of different Idss ranges and different degenerative resistors values.
I don't mind doing it either by calculating or by measuring, only, I have no clue how to do it.
Blindly copying a solution that works in certain case, under certain conditions, to another circuit operating under different conditions, isn't going to work properly, or may not work properly.
Also, choosing arbitrarily certain difference and Idss and certain value of the degenerative resistors doesn't look to me very smart.
Either I'll make changes knowing what I'm doing, or live with mismatch.
It looks like all I can do now is pray for some assistance that may enlighten me.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.