New FET distortion schematic...sensation?

[Medved]

The need of different types of JFETS with well-defined parameters. This may be a barrier for those who don't have more than one type of JFET available. FETS are also not so easy to get by.  In this respect an equivalence between the russian FETS and the more familiar J201, MPF102, 2N5458, etc here might help to break this barrier.  For instance, I can see that pairs like J201 and J202, as well as 2N5457 and 2N5458/2N5459 are good candidates for the 4-FET cascode.

You've forgotten a great 2sk170 - perfect choise for t1,t3 in cascode. It has ultra low noise... $)

Really, D90xGT has 3.5nv/vHz of input noise on 1kHz and 355nv in 70-10000 Hz. It's amazing result.
It begins to distort the output signal when input voltage is only 100-150 microvolts.

3) Finally, the very large number of JFETS involved in the designs may scare away many people from experimenting, especially beginners, since chances of a mistake and a circuit not working are quite high    I'm not scared at all, however I must confess I'm lazy!    Nevertheless I fell in love with the Magnum MkII speaker sim. Especially the "rectified" mode.

You can start with only 2 jfets and one mu-cascade and get many interesting sounds. Is it so complex?
Pairs of fet's in each cascade in magnum speakersim should have equal I0

I have a question regarding the samples. What was the setup you used? Guitar amp/speaker sim?

Guitar->D90xGT->PowerAmp->2/3 closed back cab with 12" Celestion Classic Lead 80->Shure 565sd (analogue SM58)->Audigy.

I believe you have achieved a great Marshall sound.  I'm more into cleaner sounds, like a Fender amp at breakup (with a bell-like or glass-like sound) and Vox amps as well.  I would like to experiment with this FET circuits along these lines.

In a week or two I'm going to do a clean channel with light crunch on JFET's...

It's good to have a second opinion, however not from doctors, since you end up with two different diseases! 

[stm]

Really, D90xGT has 3.5nv/vHz of input noise on 1kHz and 355nv in 70-10000 Hz. It's amazing result.
It begins to distort the output signal when input voltage is only 100-150 microvolts.

Pairs of fet's in each cascade in magnum speakersim should have equal I0

So here is where you need matching. I found the 2N5461 P-channel (available from RS Components) and I can buy locally the 2N5458 N-channel, which is its complementary pair. They both have I0(min) between 2 and 9 mA.  I think they are good candidates for the buffers on the speaker sim. Also the following couples are suitable for this, if someone can get them:  2N5460 and 2N5457 ; 2N5462 and 2N5459.

I have a question regarding the samples. What was the setup you used? Guitar amp/speaker sim?
Guitar->D90xGT->PowerAmp->2/3 closed back cab with 12" Celestion Classic Lead 80->Shure 565sd (analogue SM58)->Audigy.

So you used here the older D90xGT version? (not the GT4?)

In a week or two I'm going to do a clean channel with light crunch on JFET's...

Great!

Regards.

[ISC]

While trying to digest the information given in the translated form is not that easy, is it correct to say that there are two types of cascaded FETs
One using the same type FET (J201's) and different FET's (J201 and 2N25457/8) and that the rule is that when both the same they must have matched VO or VGS and when different the FETs must meet a set Vo or VGS range so like the Phase90 the FETs must be matched.
Why do you use these different types of cascades in the shematics , are there rules for this as well.
When the ROG amp emulators hit the forum a lot was learned and some famous preamp ccts became great little stomp boxes. Can the cascade thing be taken into those designs and even further ? Im glad you have reignited the spark in this thread.



Thanks
ISC

[Medved]

So here is where you need matching. I found the 2N5461 P-channel (available from RS Components) and I can buy locally the 2N5458 N-channel, which is its complementary pair. They both have I0(min) between 2 and 9 mA.  I think they are good candidates for the buffers on the speaker sim. Also the following couples are suitable for this, if someone can get them:  2N5460 and 2N5457 ; 2N5462 and 2N5459.

You can build this device using only n-channel (or p-channel as well) but in this case mathing pairs are nedeed too.

So you used here the older D90xGT version? (not the GT4?)

No, no! GT4 of course $)

While trying to digest the information given in the translated form is not that easy, is it correct to say that there are two types of cascaded FETs
One using the same type FET (J201's) and different FET's (J201 and 2N25457/8) and that the rule is that when both the same they must have matched VO or VGS and when different the FETs must meet a set Vo or VGS range so like the Phase90 the FETs must be matched.
Why do you use these different types of cascades in the shematics , are there rules for this as well.

I can't well understand you...

When the ROG amp emulators hit the forum a lot was learned and some famous preamp ccts became great little stomp boxes. Can the cascade thing be taken into those designs and even further ?

Yes, for example i've built Slo100 preamp on JFET's

Thanks

Not at all $)

[stm]

I'll try to clarify things up (based on my current level of understanding):

While trying to digest the information given in the translated form is not that easy, is it correct to say that there are two types of cascaded FETs

Don't forget to check the original article by Xbananov cited in the very first message of this post. Also, go down and you'll find four PDF files with the (automatic) translation. I suggest you print them all and contrast the original with the translation, since the translation changes some acronyms in the original making them unreadable.

Regarding JFET AMPLIFIER circuits, three different topologies have been used extensively so far:

A) Single JFET amplifier (no matching needed for obvious reasons). This kind of amplifier has been widely used for boosters and in ROG's amp sims, oh and don't forget the Fetzer Valve (by ROG) and the Marsha Valve (by Erik Vanhansen).  Clipping on the lower side is soft while clipping on the upper side is hard. IMHO this is a good representation of triodes, however the upper clipping might be too hard for some tastes. This topology has moderate gain (around 20 dB).

B) Mu-cascade (AKA Mu-amp). This uses two identical JFETs. Accurate matching is not a must, however in my sims I found that mismatch changes the upper and lower clipping level a bit, so the upper JFET biasing might need some adjustment around Vcc/2 if you desire clipping symmetry. This explains why swapping JFETs in the circuit DOES alter the sound, but IMHO, it is nothing that you can't fix with a biasing pot. This topology has been successfully used on many designs (Jack Orman introduced its use as a guitar booster), and a good example are the BSIAB I & II. Clipping is soft on both signal ends. This topology has high gain (consider around 40 dB) and it is typically associated with preamp valves. Only very light loading on the output is acceptable.

C) Quad-cascode (I introduced the term QUAD to strengthen the fact that it uses four JFETS). This is the most scary one to implement, since two different kinds of JFETs are needed. This topology can be thought of as an extension of the Mu-cascade. The important thing to remember is that the two new JFETs (T2 and T4) that go over the two original JFETs (T1 and T3) must have larger V0 (VGSoff) and I0 (Idss) in order to operate properly. The matching among the identical JFETs is desirable but not mandatory, just as in the Mu-cascade. This topology has even softer clipping, in a manner similar to the tube screamer.  It has VERY HIGH gain (consider now 60 dB) and it is associated typically with power amplifier valves. An important consideration is that this circuit needs basically no loading to properly function, so a buffer at the output is mandatory.

Now take a look at the waveforms and hopefully you'll get a better idea of what to choose:

(NOTE: image borrowed from Xbananov's article)


The above was basically a rephrasing of what has been said here before and what's in Xbananov's article. Medved, please correct any inaccuracy!

Best regards,

[Medved]

Regarding JFET AMPLIFIER circuits, three different topologies have been used extensively so far:

A) Single JFET amplifier (no matching needed for obvious reasons). This kind of amplifier has been widely used for boosters and in ROG's amp sims, oh and don't forget the Fetzer Valve (by ROG) and the Marsha Valve (by Erik Vanhansen).  Clipping on the lower side is soft while clipping on the upper side is hard. IMHO this is a good representation of triodes, however the upper clipping might be too hard for some tastes. This topology has moderate gain (around 20 dB).

That's right, but! You should always remember that you have to set the working point of the scheme for each JFET individually. Isn't it maching in a sense? You should "match" resistors and JFET $)
IMHO it's bad representation of clipping triodes.

B) Mu-cascade (AKA Mu-amp). This uses two identical JFETs. Accurate matching is not a must, however in my sims I found that mismatch changes the upper and lower clipping level a bit, so the upper JFET biasing might need some adjustment around Vcc/2 if you desire clipping symmetry. This explains why swapping JFETs in the circuit DOES alter the sound, but IMHO, it is nothing that you can't fix with a biasing pot. This topology has been successfully used on many designs (Jack Orman introduced its use as a guitar booster), and a good example are the BSIAB I & II. Clipping is soft on both signal ends. This topology has high gain (consider around 40 dB) and it is typically associated with preamp valves. Only very light loading on the output is acceptable.

Jack Orman isn't an inventor of mu-cascade. At least in 1983 russian book was published with the description of mu-cascade.
Here is the first page. Pay attention to the year of publising.
http://www.sugardas.lt/~igoramps/article29/02.jpg
Here is the description of mu-cascade:
http://www.sugardas.lt/~igoramps/article29/05.jpg

You can change the level of assymetry by biasing the upper JFET- it's absolutly right, moreover, symmetric clip is often impossible when upper JFET bias is Vcc/2 even with matched JFET's. Upper clippnig level is practically Vcc, lower clipping level is gteater than voltage on the source of lower JFET which can be 1.5-2v.

C) Quad-cascode (I introduced the term QUAD to strengthen the fact that it uses four JFETS). This is the most scary one to implement, since two different kinds of JFETs are needed. This topology can be thought of as an extension of the Mu-cascade. The important thing to remember is that the two new JFETs (T2 and T4) that go over the two original JFETs (T1 and T3) must have larger V0 (VGSoff) and I0 (Idss) in order to operate properly. The matching among the identical JFETs is desirable but not mandatory, just as in the Mu-cascade. This topology has even softer clipping, in a manner similar to the tube screamer.  It has VERY HIGH gain (consider now 60 dB) and it is associated typically with power amplifier valves. An important consideration is that this circuit needs basically no loading to properly function, so a buffer at the output is mandatory.

No, it isn't an extension of the mu-cascade. It is exactly a mu-cascade but on complex JFET's. It's a big difference. Gain can be as high as 70-75db. It's also very sensitive to input capacity of the buffer.

Here is a new sample of D90xGT4 which contains only acoustic feedback of amp and guitar. I think it's interesting... Singinig guitar... $)
http://webdrive.purga.ru/egorov/d90xgt/d90xgt6.rar

[stm]

That's right, but! You should always remember that you have to set the working point of the scheme for each JFET individually. Isn't it maching in a sense? You should "match" resistors and JFET $)
IMHO it's bad representation of clipping triodes.

Maybe, but it is well known that triodes clip in an assymetrical way, and I don't see this represented by the mu-cascade or the quad-cascode. 

Jack Orman isn't an inventor of mu-cascade. At least in 1983 russian book was published with the description of mu-cascade.
Here is the first page. Pay attention to the year of publising.
http://www.sugardas.lt/~igoramps/article29/02.jpg
Here is the description of mu-cascade:
http://www.sugardas.lt/~igoramps/article29/05.jpg

We are aware of this, this subject has been discussed before. My point in mentioning Jack Orman is that he was (AFAIK) the first guy that reported using that circuit topology to amplify a guitar signal, not meaning he invented it.  I've also seen that circuit in a National Semiconductor Application Handbook which I happen to have at home. I'm not sure of the printing date, but it was during the 80's.

You can change the level of assymetry by biasing the upper JFET- it's absolutly right, moreover, symmetric clip is often impossible when upper JFET bias is Vcc/2 even with matched JFET's. Upper clippnig level is practically Vcc, lower clipping level is gteater than voltage on the source of lower JFET which can be 1.5-2v.

In my sims I found also that changing the Source resistors on the two JFETs (with values between 100 and 1k) you can achieve this as well.

No, it isn't an extension of the mu-cascade. It is exactly a mu-cascade but on complex JFET's. It's a big difference. Gain can be as high as 70-75db. It's also very sensitive to input capacity of the buffer.

This circuit I haven't been able to simulate it properly. I am using 2N5457 and 2N5459 JFET SPICE models provided by Microcap 7, but I only get HARD clipping, much harder than with the mu-cascade    If you have made succesful simulations, could you post the SPICE models used on both kinds of JFETs?

Best regards.

[puretube]

First of all, lets come to an agreement about terminology.

2) About cascades

It's called mu-cascade.


It's called mu-cascade on cascode JFET's or just cascode.

?

[stm]

Puretube, I simulated the two circuits above, with the only difference being that I replaced the diodes with 2k resistors.
I also played around with the values of these resistors, making them as low as 100R.

First I started simulating with a single FET amplifier and got the expected waveform shown above (waveform A).

Then, tried the mu-cascade and got the same waveforms I reposted above (waveform B).

Finally, the 4-FET cascode produced much harder clipping instead of the soft curves previously posted (waveform C). I started with V0=-1V for T1 and T3, and V0=-3V for T2 and T4. I also moved these values arounds, but could not reproduce the waveforms. I started with the 2N5457 and 2N5459 SPICE models supplied by the sim SW, and then adjusted V0 (VGSoff) accordingly.  If you invert the positions of the 2N5457 and 2N5459 the circuit won't amplify, which is expected.

I also tried bypassing the lower 2k resistor (diode in the schematic) with a 100uF capacitor, and also with smaller resistors up to zero ohms. I wonder if someone has been able to obtain these waveforms in a simulation or only observed them with an oscilloscope.

[puretube]

sorry for the disturbance,
but my questionmark was related to the images (pics)
that appeared when I looked 2 pages back,
while I had 2 of the russian sites open in other windows...
(those 2 pictures showed changing fotos,
of teddybears and of showering ladies...  )

I had strange appearances/experiences before, with east-european sites earlier,
but those pics really didn`t fit into this forum!

Must be a funny provider, to show wrong pics,
while looking at "properties", it said to be the relevant
(mu-cascade) pix...

[Medved]

Maybe, but it is well known that triodes clip in an assymetrical way, and I don't see this represented by the mu-cascade or the quad-cascode.

Heh, 3 JFET cascades are also exists $) (see "Bsiab Extrem" for example) You can use a 2/3..3/4 (1/3..1/4) of vcc to bias the upper JFET- there are many ways to get asyymetrical clipping if needed.

My point in mentioning Jack Orman is that he was (AFAIK) the first guy that reported using that circuit topology to amplify a guitar signal, not meaning he invented it.

The discription of the schema begins with words:
"On pic.72 is the scheme of a distortion box on two JFET's..." And so on.
AFAIK- as far as i know... Now you know further      
Ok, it isn't really so important today who was the first

This circuit I haven't been able to simulate it properly. I am using 2N5457 and 2N5459 JFET SPICE models provided by Microcap 7, but I only get HARD clipping, much harder than with the mu-cascade    If you have made succesful simulations, could you post the SPICE models used on both kinds of JFETs?

Try to use 2N4117 and 2N3819 models in simulator if you have ones. Later i can post models of russian JFET's made by xBananov himself.

[Medved]

Must be a funny provider, to show wrong pics,
while looking at "properties", it said to be the relevant
(mu-cascade) pix...

The ftp server which i use seems to have such bug- once it showed me strange images which clearly belonged to somebody else...

[amz-fx]

Jack Orman isn't an inventor of mu-cascade. At least in 1983 russian book was published with the description of mu-cascade.

Jack Orman has never claimed to have invented the mu-amp and has always been straight forward in listing the original source for this basic building block...   the National Semiconductor AN-32 applications note that is dated February 1970, or 13 years before your Russian source.

Get your facts straight.

regards, Jack Orman

[lovekraft0]

I wonder if someone has been able to obtain these waveforms in a simulation or only observed them with an oscilloscope.

I got waveform results similar to that shown above in Circuitmaker 6(student) with the schematic as-is, using the 2N5457 model for T1 and T3, the 2N4416 model for T2 and T4, and a +/-100mV 1KHz sine input, if that's any help to anybody.

[stm]

I got waveform results similar to that shown above in Circuitmaker 6(student) with the schematic as-is, using the 2N5457 model for T1 and T3, the 2N4416 model for T2 and T4, and a +/-100mV 1KHz sine input, if that's any help to anybody.

Thanks, using the models you suggested I got similar results. I checked this circuit has a gain around 65 dB, while a mu-cascade with two J201's give 38 dB in my sim. This is in accordance with the GAINS.

What I cannot get to match is that my mu-cascade amplifier has softer clipping characteristics in comparison to the quad-cascode. I expected the opposite based on the graphs posted before by Xbananov, and reposted by me some replys above.

I was working with an input level of 100mV. I then reduced level to 10mV and the mu-cascade went into the linear region, while the quad-cascode was still clipping, now even harder than before.

The following image shows the waveforms for 100mV input, J201/2N4416 for the quad cascode with two 2k resistors on the sources, and J201 on the mu-cascade with 1 ohm resistors on the sources.



The quad-cascode has definitely more gain, however the mu-cascade has softer clipping.

[Medved]

Jack Orman isn't an inventor of mu-cascade. At least in 1983 russian book was published with the description of mu-cascade.

Jack Orman has never claimed to have invented the mu-amp and has always been straight forward in listing the original source for this basic building block...   the National Semiconductor AN-32 applications note that is dated February 1970, or 13 years before your Russian source.

A key word in my phrase is at least. I just said what i knew. I didn't claim that it was invented in Russia $) Nevertheless in this book mu-cascade was used as a distortion cascade for a guitar.

stm, lovekraft0
Guys, slow down a bit!
Some comments from bananov:

So here is where you need matching. I found the 2N5461 P-channel (available from RS Components) and I can buy locally the 2N5458 N-channel, which is its complementary pair. They both have I0(min) between 2 and 9 mA.  I think they are good candidates for the buffers on the speaker sim. Also the following couples are suitable for this, if someone can get them:  2N5460 and 2N5457 ; 2N5462 and 2N5459.

Place at least 0.5k resistor in sources or you'll ruin yourself on bying batteries

The important thing to remember is that the two new JFETs (T2 and T4) that go over the two original JFETs (T1 and T3) must have larger V0 (VGSoff) and I0 (Idss) in order to operate properly. The matching among the identical JFETs is desirable but not mandatory, just as in the Mu-cascade.

Matching among the identical JFETs is not needed at all- you should make shure that t1 has vo in 0.7-1v, t3 vo is less than 1v, t3,t4 vo in 1.8-3v.

Now about sims. Simulators have VERY poor mathematical model of JFET. They practically don't show the dependence of JFET's mu from VDS and different drain currents. That's why images of clipping are far from real, espessialy of 4FET cascode as modelling errors because of incorrect dependence of mu from vds and Id are multiplying instead of summing in mu cascsde.

And second. Please, don't forget that 4fet cascode has 10times more gain than mu-cascade! Pictures in xBananovs article was given for an input voltage nearly 10 times the voltage of clip beginning of each cascade. All voltages are different! For 4fet cascod it shold be nearly 10-20mv... And cascode shuold be real not virtual... Then you'll see what you want:)

[Medved]

Heh, again no answers...
Ok, I wrote sounds of mucascade and cascode cliipping of 1k wave with variety of input signal's level. It's ~7Mb rar archive and i can't upload it to ftp- i have modem.
Here is only two sounds:
http://webdrive.reborn.ru/egorov/d90xgt/Clip.rar
Anyone can open them in sound editor and see real clipping images:) Nothing similar to sims...

And as I promissed I public the scheme of clean channel on jfets:

Here is a sample on max gain:
http://webdrive.reborn.ru/egorov/clean/Clean1.rar

Some comments:
KP303G- any lownoise JFET with cutoff voltage near 2V
KP301A~J201
First mucascade starts to clip nearly at 1v input amplitude. It's sensivity is set by R3 and it shouldn't clip while normal playing. I use a mucascade in a first stage for ability to use a booster- clipping will be soft and pleasant. I use linear gain pot. With audio pot "clean" is up to 85-90% and only then overdrive begins- it isn't very convenient.

[stm]

Thanks for sharing.

What's your setup for recording?

[Medved]

Setup is the same:
Guitar yamaha rgx-112-->Clean channel-->Power amp tda7294-->Cabinet 1*12 2/3 closed back--> Mic Shure 565sd -->Audigy

First of all, you should try supply voltage near 25V with adjusting R3 and maybe R11. I have only 15V, but it's sounds very good. Soon, i'm going to add a tremolo %)

[MR COFFEE]

Hi Medved,

Thanks for sharing your work and experimentation here with us. Your exploration of some rarely-used circuits for amplifying guitar is commenable and valued by many of us who share your interest in this area.

Medved said

Now about sims. Simulators have VERY poor mathematical model of JFET. They practically don't show the dependence of JFET's mu from VDS and different drain currents. That's why images of clipping are far from real, espessialy of 4FET cascode as modelling errors because of incorrect dependence of mu from vds and Id are multiplying instead of summing in mu cascsde.

Well-said!

Spice is good for designing filters under linear operating conditions, and for designing circuits using a few specialized chips that have good models, but most Spice models of JFETs (and most other linear) devices are so oversimplified that their modeling of device behavior under clipping conditions fails miserably to model their behavior with any accuracy.

Thanks for reminding us all of this as well !

More breadboarding and less computer simulations - yeah!