About the common Boss SD-1 bypass problem

[cbugatti]

thanks for all the help

okay so I'm going to order 2 2SK30A's and where exactly do I solder these in??

curt

[cbugatti]

anyone??

[shredgd]

Yesterday, after three months, I had the time/will to work on the problem again.

I did other tests focusing on the few differences between the SD-1 and the TS (as the latter is free from this problem):
- the level pot on the SD-1 is only 10k (instead of 100k on the TS), so I thought the signal could pass through it to Vr more easily and then re-enter the "clean" (bypassed) pathway somewhere, so I lifted one leg of the last resistor before the level pot: no improvement.
- continuing with the supposed Vr involvement, I lifted one leg of C4, the cap which connects to Vr after the gain stage (the TS, instead, has a resistor here, and a cap to ground): no improvement.
- I noticed the 0.047u cap (C3) in the feedback loop of the gain stage connects to Vr, instead of ground as in the TS, and it does it closely to where R4 (the 100k resistor before the gain stage) also connects to Vr, so again I thought the effected signal might make its way into the bypassed signal at this level. Therefore I lifted the leg of C3 and connected it to ground: no improvement.
- I bypassed R7 (the resistor between the gain stage and the tone stage), because the TS has 1k here, and I thought too much resistance after the gain stage might bounce the sound back to the input of the IC (fantasy started to dominate...): no improvement.
- So I even lifted one leg of R7: no improvement (indeed, moving the tone pot doesn't affect the timbre of the bleed through. At least, now it is sure what comes after the gain stage is not involved).
- With R7 still lifted I even disconnected C9 from the jfet and connected it directly to a cable to my amp. By doing so I virtually eliminated the jfet switching and the output buffer from the circuit: no improvement.
- I lowered R4 from 100k to 10k and raised C2 from 0.018u to 1u (the cap and resistor between the input buffer and the gain stage) to have the same values as in the TS: no improvement.
- I bypassed the "extra" (compared to the TS) diode in the feedback loop of the gain stage and set the drive pot to half, in order to reproduce a TS with the drive pot set to max (as the drive pot in the SD-1 is 1M instead of 500k): the bleed through is there anyway.

So I'm officially giving up, I tried anything and I can't find a solution. This is indeed a poser.
The only possible explanation is C3 is too close to C2 in the circuit and "something" happens because of this...

To summarize, I'll add my previous observations, so to have them all together in this post.

- First of all, be sure the drive pot is set to max to clearly hear the bleed through (the best way is to strike a low E or A string and look for the background sizzle)
- The problem doesn't depend on the two jfets: I removed both (and then of course I used an audio probe to pick up the signal after the input buffer) with no results, and yesterday's test confirmed that.
- Therefore, lowering or even jumpering the two 22k resistors between the jfets doesn't solve the problem.
- It doesn't depend on the IC. I tried a TL072: same thing.

Giulio

[cd]

If you removed both JFETs how did you get any signal at all?  

Anyway, if you want to do a shunt you can run it off the flip flop that's already in there.

[cd]

Come to think of it, what the gate/source voltage when the JFETs are off (switch open)?  JFETs work when the gate/source voltage is 0 (low resistance).  If the JFETs are simply being turned off (gate/source voltage made negative, low drain current) but not "off enough" there could be bleedthrough.  

Then again that wouldn't explain why with Q1 removed you still get bleedthrough.  And the TS9 doesn't have this problem.  Hm.  Maybe it's a PCB layout issue?  What happens when you remove C2 AND Q1?  Maybe the audio signal is modulating the gate/source turn on voltage?  Then again the gate diode should be limiting the on voltage.  Hm.

[analogguru]

You most love your stompboxes, then they will love you...

If you love your SD-1, then you will give him a 100n film-capacitor, a Diode 1N 4148 and a FET e.g. 2SK 30A-Y or -GR and then your SD-1 will love you and everybody will be happy. :lol:

You will find a schematic how to do it here:

http://forum.musikding.de/yabbse/index.php?board=8;action=display;threadid=2224;start=0#lastPost

analogguru

[R.G.]

Guys, this is not that hard. It's not magic, it's not mysterious.

The distorted signal leaks because the JFET in the distorted signal path is not turning off fully. Period.

This in turn is because the JFETs do not have a high enough reverse Vgs to fully turn them off.

This in turn is because the sources of the JFETs are not high enough for the ground voltage on their gates to fully turn them off for devices that are within the guaranteed ranges on the device datasheets.

The answer is to either (a) get JFETs with lower Vgsoff or (b) raise the bias voltage on the sources of the JFETs you do have or (c) both of the above.

You may, as analoguru suggests kill the gain of the distortion section while in bypass, but that's an extra credit thing. That just lowers the leakage through the distortion path so it is below the distortion threshold and although it still leaks, it leaks the same signal you're listening to, so you don't notice.

Specifically, Q1 is the problem. Its source is sitting at the bias voltage of 1/2of the battery. This may be as low as 8V easily, and its gate is pulled down by Q4's collector in the discrete flipflop. The series diode means it can't get pulled lower than a diode drop, so you have 4v-0.5V = 3.5V  of back bias on Q1's gate.

The 2SK30A is listed as having a Vgsoff of -0.5V to -5.0V. If you get a -5.0V device, it will not turn fully off in the SD-1 circuit.

You can do a couple of things. One is to stick in a J201. The J201 is notable for having a Vgs0ff of -0.1 to -1.5V. The 3.5V of the existing circuit will turn this one off very well, in all cases. The J201 is often used for a replacement for the 2SK30A in some synth circuits.

The pinout of the 2SK30A is SGD with pins down, flat facing you. The J201 is pinned out DSG in the same orientation.

You can also find other JFETs, such as the 2N5292 or 2N5485, which will work almost all the time, as their maximum Vgsoff is -3.5V.

You can raise the bias voltage from half the battery to some higher voltage by changing R18 and/or R19 to get a bias voltage of maybe 5V or 5.5V. This will probably work just fine. The lowest effort version of this is to change R18 to a 22K or to parallel it with a 51K or 68K resistor.

While I have not done either of these, the schematic and Mother Nature say they have to work - unless there is something even more subtle that Mother Nature is trying to teach me, which does happen from time to time.

[cd]

Specifically, Q1 is the problem. Its source is sitting at the bias voltage of 1/2of the battery. This may be as low as 8V easily, and its gate is pulled down by Q4's collector in the discrete flipflop. The series diode means it can't get pulled lower than a diode drop, so you have 4v-0.5V = 3.5V  of back bias on Q1's gate.

Thanks for explaining that more clearly than my "hmm, this or that" musings, RG

[cbugatti]

what is a J201 exactly and where would I be soldering this in?

R.G.- so by theory, if you lowered R18 and R19 from 33k to 22k it would raise the bias voltage which would turn off the JFET in the distorted signal path?

[cd]

what is a J201 exactly and where would I be soldering this in?

R.G.- so by theory, if you lowered R18 and R19 from 33k to 22k it would raise the bias voltage which would turn off the JFET in the distorted signal path?

A J201 is another FET, like a 2N5484.

Changing R18 and R19 (the 4.5V bias generating resistors) would make no difference.  You have to lower (or eliminate) R11 and R14.  If that doesn't work try a different FET (as suggested above).

[R.G.]

Changing R18 and R19 (the 4.5V bias generating resistors) would make no difference. You have to lower (or eliminate) R11 and R14.

Can you expand on your reasoning for this statement? I may have missed some thing in examining the circuit.

If I read the schematic right, the distortion path signal is fed through Q1, the buffered clean signal through Q2. While I have not examined an SD-1, I have fixed several TW-9's that have this same behavior. The issue was always failure to turn off Q1 completely, and changing either the bias voltage to allow more Vgsoff reverse bias or the JFET to a lower Vgsoff type so that the voltage that was present would turn off the JFET always fixed the problem.

Changing R18 and/or R19 to raise the bias voltage makes more reverse Vgsoff available because it raises the voltage on the sources. You can't change the gate pulldown voltage much, as it's pulled to as near ground as the circuit can pull it.

I don't understand the reasoning that says that you must lower or eliminate R11 and R14. They're not part of the root cause of the problem.

It's always possible that I missed something. Could you expand on your reasoning?

[analogguru]

@R.G.

I´ll try to explain it to you.

To make it easier, lets assume following conditions:
1.) Output impedance of Input Buffer is <1 kOhm
2.) Off-resistance of the FET is 1 MOhm
3.) On-resistance of the FET is 0 Ohm

Now, if the Pedal is in Bypass Mode the distorted signal is still present at FET Q1.  In this state Q1 is comparable to a resistor of 1M which forms a voltage-divider with R14 + the output impedance of the input stage...
For this reason, the distorted signal is only attenuated by the factor of 50.

If you remove R14 the voltage divider is formed by the 1M of the FET and (only) the outout impedance of the input buffer, so the ratio is approx 1000....a difference in damping, eh ?

analogguru

[cd]

Changing R18 and R19 (the 4.5V bias generating resistors) would make no difference. You have to lower (or eliminate) R11 and R14.

Can you expand on your reasoning for this statement? I may have missed some thing in examining the circuit.

Mmm, probably not because your theory is 1000x times better than mine.  My reasoning: OK I missed something with R18 and R19.  If you change both R18 and R19 to the same value (say both to 10k, or both to 100k) it won't make a difference, since you still get 4.5V.  Change one or the other or both to different values (say to 3V) then it makes a difference (right?) but you can't stray too far from 4.5V or else you start to mess with the rest of the circuit.  Hmm as for R11 and R14 - that's my mistake, I thought they were connected to D, not S (I always get that wrong with FET drawings that don't have the G close to the S!)  So there's no drop from 4.5V across either which shouldn't affect the switching.

Am I right in thinking that you could just replace D6 and D7 with two 4148s in seriesto fix things?

[shredgd]

Sorry for not writing before, I've been very busy.

cd wrote:

If you removed both JFETs how did you get any signal at all?

To get signal after I removed both jfets, I connected one leg of a 0.1u cap to the tip of a guitar cable (plugged into my amp) and the other leg to various testing points, for example C2 and C9. Of course I also connected the ground of the pedal to the ring of the cable jack. This method is what I previously refered to as "audio probe" (copyright R.G. Keen).
This way I could pick up the signal of my guitar right after the input buffer. The leakage of the distorted sound into the clean path was already there!

analogguru wrote:

If you love your SD-1, then you will give him a 100n film-capacitor, a Diode 1N 4148 and a FET e.g. 2SK 30A-Y or -GR and then your SD-1 will love you and everybody will be happy.  

You will find a schematic how to do it here:

http://forum.musikding.de/yabbse/index.php?board=8;action=display;threadid=2224;start=0#lastPost

Very, very good idea! That will work for sure!
I'll let you know about the results (they can't be other than good) when I have the time to materially do the mod! Bravo!


R.G. wrote:

Guys, this is not that hard. It's not magic, it's not mysterious.

The distorted signal leaks because the JFET in the distorted signal path is not turning off fully. Period.

Thanks for your explanation about jfet switching voltages but, please, read previous posts before answering.
Jfet switching is clearly not involved in this problem (read above), at least in the case of the SD-1.

Indeed, in my opinion, this discussion about jfet switching can be useful, but not for this particular problem.
Ciao!

Giulio

[analogmike]

If he removed the ON JFET, and still has the noise, FETS will not help. If it does remove the noise, it will help. I'm not sure which it is. Easiest way to test is just pull out the ON JFET, and play in the OFF position. Can you try this and see what you get, Giulio?

[shredgd]

If you mean I should try to remove Q1 only (the jfets which lets pass the distorted sound or not), that's of course the first thing I tried before pulling out the other jfet as well: yes, the leakage is always there.
Go through this thread from the beginning and read my posts to see how many tests I did, most of which firmly exclude the role of the jfet in this problem!

[cbugatti]

i know for a fact that removing r11 and r14 do not do anything....I'm going to try changing r18 to 22k and if that doesn't help than r19...should get this done tonight and will post when finished

[R.G.]

I´ll try to explain it to you.
To make it easier, lets assume following conditions:
1.) Output impedance of Input Buffer is <1 kOhm
2.) Off-resistance of the FET is 1 MOhm
3.) On-resistance of the FET is 0 Ohm

Ah, thanks for the analysis.

The problem with that is that 2.) is not true. 3.) is not either, strictly, but it has no effect on the issue at hand.

The off resistance of a JFET biased off is not 1Mohm, but a few orders of magnitude higher. For the 2SK30, the datasheet says that the specified cutoff voltage will reduce the drain current to under 100nA at 10Vds. Doing the math, Rds at cutoff is 10V / 100nA, or 100Mohm. That makes the division ratio 100M/22K = 4545, or -73db if I punched the correct buttons on the calculator.  While is it possible to hear sound 73db down, that's pretty good isolation for a simple switch.

And the 100na of leakage is only what the maker guarantees. The typical device is usually better.  Switching JFETs specify this at 10na.

So if you are not getting at least 73 db isolation in this setup, *you are not turning the JFET off hard enough*.

We also find confirmation of this:

i know for a fact that removing r11 and r14 do not do anything....

But it's possible, as I said, that Mother Nature is trying to teach me something different...
[quote="shredgd"Thanks for your explanation about jfet switching voltages but, please, read previous posts before answering.
Jfet switching is clearly not involved in this problem (read above), at least in the case of the SD-1.[/quote]
I'm sorry. I'm getting on in years and my mind isn't as quick as it once was. Can you point out the part I missed, please?

[quote="cd"]Change one or the other or both to different values (say to 3V) then it makes a difference (right?) but you can't stray too far from 4.5V or else you start to mess with the rest of the circuit. [/quote]
Yep, that's correct. You want to RAISE the bias voltage a bit. Get it up to 5V or 5.5V if you can by reducing R18 or increasing R19. Don't change them both to some different but equal value.

Am I right in thinking that you could just replace D6 and D7 with two 4148s in seriesto fix things?

No, that is incorrect. Doing that decreases the off-bias to the JFETs and makes it worse.

[cbugatti]

I've got some 22k resistors...so if I put one in r18 this should raise the bias voltage to around what...5..5.5V?

if this doesn't cure it completely, I should rasie R19, but I don't have any higher value resistors right now...if put a 22k in series with the 33k in R19 would this raise or lower that value?

thanks

[cbugatti]

*update*

So I put the 22k resistor in R18.  No improvement.  Then put two 22k resistors in series in R19.  No improvement.  I've got no idea.  I just know that I absolutely love the sound of this mosfet/germanium modded SD-1 and I can't turn the drive up at all to get it.

Does anyone have a definite solution for fixing this problem without building a bypass box? If so, how, and can you explain it at an easy level to understand?  

I don't know where I would solder a J201 in (r10/q5/c3...etc) and I have no idea where I would put the parts that "analogguru" was talking about

thanks
curt