diyAudio Forums Archive

diyAudio Forums Archive > Top > Amplifiers > Pass Labs

Pages: [1] 2

AlephJ-X - Click HERE for Original Thread

wuffwaff

Hi,

after thinking about this for some weeks I´ve changed one of my Aleph-X monos to a Jfet input. (To see (hear) what all the fuss was about)

I´ve used a pair of matched 2SJ74 (Idss about 10,5mA) cascoded with a pair of matched 2SJ103 (Idss about 11mA).

The 2SJ74 are connected to the current source with one 100 Ohm trimpot so they have about 50 ohms source resistance and dc offset can be adjusted.

I changed the drain resistors to 680 Ohms from 390 and the current source bias from 20mA to around 11mA.

No other changes were made.

Well the good thing about this is that it works, dc offset is near zero and the absolute offset starts quite low. I didn´t change the McMillan resistors (10k) wich probably isn´t quite right but I have to think about this a bit (suggestions are welcome).

The problem is that at the moment the bias probably is too low to drive the 3 output fets per side.
At 2Volts the bandwidth is about 83kHz (8 Ohms) and 71kHz (4 Ohms)
This changes to 63kHz and 45kHz at 10 Volts
and 33kHz and 20kHz at 20 Volts output

I did expect something like this but not as dramatically. Before the change the bandwidth was 142kHz (8 Ohms) and 126kHz (4 Ohms)

Any ideas if something else could be the cause for this behaviour?

William

P.S. I didn´t listen to it yet and as long as there´s only one I´m not shure I can say something usefull about the sound. Before I modify the second amp I would like to see if I can raise the bandwidth (without disconnecting some output fets!)

jupiterjune

Cool!

Why don't you cascode with mosfets and put the current through the differential pair back to 20 mA or so?

The jfets should be able to take that current when cascoded w/mosfets no problem--I did it with my BOSOZ at 20mA per side (there are two CCS in BOSOZ, if your not familiar with it) with no problems at all.

Even though the gates of the output devices have high input impedence, they do require some current. Sounds like the 5.5 ma per side is not enough to drive the gates.

JJ

Babowana

quote:

Originally posted by wuffwaff

. . . and the current source bias from 20mA to around 11mA.

As JJ indicates, you can keep the higher bias current.
For example, I'm using 2SJ108GR (Idss clasification, -2.6mA~
-6.5mA) for my Babo Zen. But, it handles 12mA of bias current
without any prblem. I had another test and found that
it could handle even 16mA. This much of bias was not possible
simply by the self-biasing. It was, however, possible together with
current source because, in this case, the jfet acted as a simple
resistor.

Hope this info will be useful . . .

:darkside:

GRollins

quote:

Originally posted by Babowana

As JJ indicates, you can keep the higher bias current.
For example, I'm using 2SJ108GR (Idss clasification, -2.6mA~
-6.5mA) for my Babo Zen. But, it handles 12mA of bias current
without any prblem. I had another test and found that
it could handle even 16mA. This much of bias was not possible
simply by the self-biasing. It was, however, possible together with
current source because, in this case, the jfet acted as a simple
resistor.

Hope this info will be useful . . .

:darkside:

Indeed...but aren't you driving the Gate negative?

Grey

Babowana

Hi Grey,

You could find the answer to your question from the attached
sketch. Iwill post pdf also on my thread.

:darkside:

EUVL

William,

What is observed is not unexpected.

A JFET has much lower transconductance at about 30mS. On top of that you have put in about 50 ohm source degeneration. So you open loop gain for the first stage drop from say 25 to 8.5. On top of that, you are using a cascode (2SJ103) which has almost identical Idss to the driver (2SJ108), so that you are not getting much Vds across the diff pair (I doubt if you get much more than 1V).

And you would expect the open loop bandwidth of the second stage to almost half, especially in a high power AX where you have many power FETs in parallel, since you increase the signal source resistance (which is essentially determined by the drain resistor of the diff pair) by a factor of 2.

And since you are already operating at or near Idss, you are asking your JFETs (both the driver and the cascode) a lot to swing 2x bias, especially the cascode which has much large change in Vgs and hence would essentially switch off your driver JFET by making its Vds near zero.

So what to do ?

I would suggest changing the bias to around 5mA and the drain resistor to 910k to start with. Change the source trimpot to 10R which is sufficient to balance currents. You get the JFETs back to a decent working point, and get the gain back up. The open loop bandwidth of the 2nd stage is a problem, but you now also have more negative feedback, which should help. And this is quick enough to try.

If you still want to run at 20mA and use 390R drain resistor, then just put 3 pairs JFETs in parallel and bias each at 7mA. You should use FETs then with Idss no less than 12mA. This will get you back the bandwidth for the second stage. The first stage is now of course 3 time lower in bandwidth, but then a 2SK108 has much lower capacitance to start with compared to a IRF9610. Anything in between is a balancing act between first and second stage. The combinations are endless. You can even add a driver (source follower with IRF610 / 120 ohm) between the diff pair and the IRFP240s to get you even higher bandwidth than before.

And I would use MOSFETs as cascode with Vgs 9.1V to give a decent 5 to 7V Vds to the diff pair, if I were you. Or you use selected J174s as cascode (one for each 2SJ108).

Just my 2 cent.

Patrick

wuffwaff

Hi,

thanks for your replies.

Jupiterjune, Babowana:

How can I bias a JFet with an Idss of 11mA with 15mA? I would have to bias the gate negative and this would cause a 0,7V wide flat spot.

Patrick,

thanks for the answer!

The 50R source resistors are there because I couldn´t get a 25R or 50R trimpot yesterday :rolleyes:

1V across the diff pair is about right (but I´m using SJ74, not SJ108)

Bias was chosen around 5.7mA per JFet. Since Idss is around 10 this would give a 4mA swing. Do you suggest 5mA in total or per side? With 910R this would give 0,025 x 910 =2,27V wich is not enough, I need around 3,9V to get the absolute offset right.

I put the cascodes in because first I wanted to try it in my (32V) Aleph5. This uses 3 Fets and needs only around 3mA of current swing at 100 kHz (measured). With these cascodes I could just plug the diff stage in without making bigger changes to the circuit.

OK,

I will try to get a 10R trimmer today and throw out the cascodes (they are not needed in my Aleph-x) Then see what happens.

I don´t want to make the input stage too complicated as the amps are sounding very nice at the moment (with IRF9610) and this was just an experiment.........

William

Babowana

quote:

Originally posted by wuffwaff

How can I bias a JFet with an Idss of 11mA with 15mA?

2SJ74 is big enuf to be the diff input pair.
And, the CCS can be any kind.
Choky's Babbelfish is an excellent example.
Babbel fish has TR CCS.

I could use the A-X MOSFET CCS as it is. Why not?

I'm sure that the sound with 2SJ74 diff input will
be different in positive way ^^.

:darkside:

Luke

wouldnt the loveltec jfets give the required current to drive the output stage?

EUVL

> Bias was chosen around 5.7mA per JFet.

Sorry. I misread it as 11mA per JFET. 5.7mA is fine.

> With these cascodes I could just plug the diff stage in without making bigger changes to the circuit.

I would leave the cascode in. John Curl was right in saying that you don't want to operate Toshiba JFETs such as 2SJ74 too close to their max allowable Vds. I think your 100W AX would put something like 20V on them ?

Just that you should not use 2SJ103 as cascode. If you look at Borbely's circuits, though he uses 2SJ103, he was using 2.8mA bias. Also he has a lot of open loop gain, which means you swing very little current. Try J174 if you wish to cascode with JFET. Or just use no cascode, as you already planned.

If you have well matched 2SJ74s, I would throw out the trimpot altogether.

Patrick

PS I have never tried Loveltechs at such low current (10mA). So I do not know if you can find a linear loadline at 10mA. Also you will fins that Loveltechs has much higher capacitances, even when cascoded.

wuffwaff

Hi Patrick,

I´ll see if I can get some J174´s.
For now I will try without the cascode as the fets are seeing around 18V and input capacitance will at least be lower than with the 9610´s

I couldn´t get a 10R trimmer but I´ve just put two 10R´s parallel to the 50R trimmer wich should work as well. I think I need it as I "only" bought 30 2sj74 and the matching isn´t that nice if you look at Idss and Vp.

Will inform if I have new results.

William

jupiterjune

I will take a closer look at what I have populated on the boards and post a schematic tonight.

I know I used some very high Idss jfets ( u1897 from mouser, and another) -- these I clearly remember worked with higher bias current. I need to doublecheck on the lower Idss j-fets used--I think they worked as well, operated in the pentode region, but I've got to check. I got rid of the flat spots on top of the sine waves by adjusting the gate voltage on the cascode mosfet. You can't do that with the self-biased jfet cascode.

For now, I've got to install some trim to keep the wife happy!

wuffwaff

Update,

so I removed the cascodes (2SJ103), lowered the source resistors to 8R and plugged the hole thing in.

Results are interesting and a lot better than before:

Bandwidth into 8 and 4 ohms:

2Volts 120kHz / 99kHz
10Volts 120kHz / 95kHz
20Volts 102kHz / --
15Volts -- / 74kHz

This is a lot better than yesterday. I also measured the voltage over the drain resistors wich was 4,7 volts and a bit higher than expected giving 6,9mA bias per fet.

So I changed these from 680 to 820 Ohms to lower the bias to a bit below 6mA.

This changed the bandwidth at 15V / 4 Ohms to 86kHz (from 76kHz)

So for the moment (until I get my J174 and maybe some V spec 2SJ74) I will leave it like this and do some listening.

Still not shure if I shouldn´t raise the McMillan resistors though. Absolute offset starts at around 2V and goes down quite fast.

William

EUVL

> Still not shure if I shouldn´t raise the McMillan resistors though. Absolute offset starts at around 2V and goes down quite fast.

I also use 10k as you. But by all means try 15k or 20k and see whether you hear anything different.

Although V spec 2SJ74s would allow you to lower the drain resistor further, it does not help to keep open loop gain. If you are really into more bandwidth, I would seriously consider putting 2 diff pairs in parallel.

But glad you are on the right track again.

Time for some listening first. :)

Patrick

jupiterjune

OK-

Here is a hopefully readable schematic of the cascoded BOSOZ. I made some changes to increase the gain as would be needed for an amp front end. This really increased the power dissipation across the jfets. These are vishay j105's -- that is all I had populated in the boards. ( I have some u1897's on hand, and also some 2n4416's....)

Ignore the scribbled out lines--the diagram was for testing a preamp circuit operating just above unity gain.

JJ

jupiterjune

Well I checked the schematic -- it is barely readable. I will try to fiqure out a way to post a more readable schematic with what I have on my computer.

Anyway, here is a pix of the input and output. The input is the smaller sinewave, and the major divisions are 1V. The larger sinewave is the output, and the major divisions are 5V. I backed off on the magnitude of the input sinewave until the flat spots showing up on the tops of the output sinewave were just about gone. I think you can see it a little bit if you look closely.

The jfets, vishay j105's are supposed to have an Idss min of 500mA.

Again, this circuit was tweaked for use a preamp circuit that was driven single ended -- I'm sure much better results would be had if it were optimized for an aleph x input. Cascode modulation seemed to work wonders, but it is a bit difficult to implement when using a single ended input.

Any comments would be appreciated.

jupiterjune

quote:

Indeed...but aren't you driving the Gate negative?

Grey- Is there a particular problem with driving the gate negative -- actually, now that I think about it, I don't think I entirely understand the question -- the BOSOZ has n channel devices, and the J105's have Vgs off of app. -7 volts. So those gates are negative with respect to the jfet sources. But these are p-channel devices.......I am trying the understand the problem you were seeing.

Wuffwaff-
If I get a chance today, I'll try re-populating one of the boards with 2n4416's -- they are supposed to have an Idss of 10 to 15 ma -- I guess then I will see what the problem is.
Perhaps it just isn't possible to get the dc voltages needed to drive output stage??

JJ

Babowana

quote:

Originally posted by jupiterjune

. . . , and the J105's have Vgs off of app. -7 volts.

So, your source output is allowed to see the +6.73V?

:darkside:

jupiterjune

quote:

So, your source output is allowed to see the +6.73V?

Babowana, not sure how you got this number, I know the pix didn't come out too well so here are the voltages (DC), (a bit tough to read on the dwg):

gates of input jfets 0 (vishay siliconix j105)
source of input jfets 7.43 V
drain of input jfet 31 V, also source of cascode mosfet
gate of the cascode mosfet 35.6 volts (irf 610)
drain of the cascode mosfet 48.6 V

The voltage on gate of the cc mosfet had to be cranked up to get the roughly 12.5 volts output before clipping, but this dumped much of the load on the j-fet.

JJ

I'm going to mess around with this for a bit to see what I can figure out.

Wuffwaff--please let us know what you end up with and how it sounds compared to the stock aleph-X

wuffwaff

Hi,

changed the other mono this morning and had a bit of a listen this afternoon and tonight.

I´m afraid this isn´t really a succes in this combination. The mids and highs are good to very good but it is a bit lacking in the low bass (< 50-60Hz).
The punch is gone and sustained bass notes are there but not as loud as before.
This wasn´t quite what I expected.

I´ll leave it like this for a few more days, then go for a 10 day cycling vacation and when I return I´ll listen a bit more and return to the former (standard) setup.

Maybe I´ll try a different bias for the input pair (drain resistors from 390R to 330R) and see what happens..........

I noticed another thing wich must have something to do with the active current sources:

if you suddenly raise the output voltage from 10V to 18V (10kHz, 4 Ohm load) the amp will only keep this output voltage for a short while (maybe one second). After that the level goes down and distortion sets in. I think that this has something to do with the bootstrap capacitor and the resistors that feed it and the active current source.
This only happens at higher frequencies when the amp is operated near it´s current limit.
Lowering R14/31 (hifizen) from 1k2 to 820R makes it worse, raising it to 1k8 makes it better.
All earlier Alephs (2,4,5) had a 4k75 resistor for R14 and a 1k5 for R15. A3, 30 and 60 used a 1k5 for both.
Could it be that at high frequencies the bootstrap capacitor is emptied by the higher current (because of the C.S.Fet capacities)?

more in a few weeks.....

William

EUVL

William,

The improvements in mid and high are what you should expect.
The loss in bass is due to your loss in open loop gain. To get that back you need to get rid of the source resistors of the diff pair altogether, and increase the drain resistors to 900 ohm to 1k. Your open loop gain would then be comparable to before. Only now you have a lower open loop bandwidth for the second stage. But that may not be a problem once the loop is closed.

As to the bootstrap cap, put a 2200uF cap in and try to see whether it makes any difference.

Don't give up as yet.

Patrick

wuffwaff

Hi Patrick,

I said the mids and highs were good to very good. Before the change the mids and highs were good to very good too.......

It is always difficult to tell the difference if some part (like the bass) is lacking.

I tried without the source resistors and got 132mV of offset wich is a bit too much for my taste. I will have to search a bit further for some better matched pair with high enough Idss. The 30 pieces I´ve got are a lot further apart as I´m used to with IRF parts

At the moment the drain resistors are 820R and bias is 4,7V/820 = 5,7mA. If I use a 900-1000R drain resistor bias will be 4,7-5,2mA so I could use a diff pair wit an Idss between 9 and 10mA. I´ve got 7 of those so maybe I´ll find some that work without source resistors. (2SJ109 would be welcome.......)

Could you explain why the open loop gain of the first stage has an influence on the bass performance? I must admit that I only looked at 1kHz signals and above cause I didn´t expect any difficulties here.

As for the cap in the active current source. Going to 2200uF would probably change the time from 1 to 10 seconds before distortion will be setting in. I don´t think this behaviour is important for normal listening and I will have a look if my Aleph 5 has the same "problem". It would just be nice to understand.

William

Zen Mod

quote:

Originally posted by wuffwaff
...........

check AC gain of output CCS

Babowana

quote:

Originally posted by Babowana

Babowana, not sure how you got this number

Sorry, I have read "app.-7V" as "app -.7V"

:darkside:

EUVL

> Could you explain why the open loop gain of the first stage has an influence on the bass performance?

It has been a long time since I did this, so I am only recalling from memory.

If I remember correctly, the standard AlephX-100 has about 30-40dB NFB on 8 ohms. This drops proportionally with load impedance. If your TOTAL loop gain is further reduced (since you have changed nothing for the second stage, this can only be a result of the first stage changes), AND you have a very reactive load which drops to say 1 ohm during large amplitude transients, your amplifier will run out of gain very soon.

Anything below 20dB NFB is asking a lot from the JFETs, as you would be putting something like +/- 0.2V at their gates. I think you would be driving one of them over Idss, and the other completely off (and hence deeply into its non-linear region.

Also, your damping factor suffers as the amount of NFB drops, if I am not wrong.

I have relatively efficient speakers (B&W 804s), and I do not need a lot of bass for my taste. So for me it was never an issue. I can recall (this has been years) comparing the Rollins Aleph-X with 4x 044s and 2x 9610s with the 100W AJX and for me it was definitely an improvement in terms of bass.

Compared with the AlephX100 with 9610s, I found highs much cleaner and sweeter. But that is only my observation with my own setup. AND I was (am) using 2SK1529s and NOT IRFP240s.
JFET was 2SJ109, no source degeneration, drain 910R.

Patrick

wuffwaff

Zen Mod,

ac-current-gain: checked and i.O. (51%)
bias: checked and i.O (6,9A)

Nothing was changed at the active current sources so this is as expected.

Patrick,

don´t think my speakers (Thiel CS6) are so difficult to drive at lower frequencies. The lowest impedance (3 Ohms) occurs at higher frequencies.

I will try the 910R / no Rs version when I´m back from holiday.
Do you know how much the gain will change from 820R / 8R?
Do the 2sk1529 have a lower input capacitance? Are they still available?

William

EUVL

The gain for the first stage with Toshiba JFETs (J74s or J109s) is roughly :

R drain / (30ohm + R source).

You can then work out the difference between different versions easily. The 30 ohm comes from 1/Yfs at 5mA bias, which is about 33mS.

Yes 2SK1529 is still widely available. And they have much lower capacitances. Datasheet is available on the internet (too large to post).

If you get 100 pieces, they are only slightly more expensive than IRFs (in Germany). But this was 3 years ago, I have to say.

Patrick

wuffwaff

Hi,

only found some 1529´s at 4,57€ a piece at Kessler.......

So gain now is 820/40 = 20,5 (26dB)
Gain 910/40 = 22,75 (27dB)
Gain 910/30 = 30,33 (29,6dB)

so I could raise the open loop gain by 4dB, and loose some in the second stage.

I´ll just try. BTW the bass loss is not only appearant on high level listening but also at lower levels.

William

EUVL

> only found some 1529´s at 4,57€ a piece at Kessler.......

http://www.fibra-brandt.com/
List price 3 Euros + VAT. But ask him how much for 100.

Anything less than 100 you will not get good match (from experience); so I suggest you try to find out the cause to the bass loss first, which I admit is a bit of a puzzle to me as well.

Or maybe you should wait until Nelson publishes the Aleph J circuit for some hints.

Patrick

wuffwaff

Hi Patrick,

did you compare the 1529 to the IRFP240´s (I´ve got plenty of the latter.....)

One hint concerning the Aleph J could be the use of 2 output fets instead of 3 (6?)

Let´s wait until I´ve tried the 910R plus no source resistors to see if this helps. If not I can return to the 9610´s and see if the dissappeared bass returns.

I will build a pair of Aleph 30´s for a friend in the near future. Will do further experiments with the JFets in these.......

William

GRollins

quote:

Originally posted by jupiterjune

Grey- Is there a particular problem with driving the gate negative -- actually, now that I think about it, I don't think I entirely understand the question -- the BOSOZ has n channel devices, and the J105's have Vgs off of app. -7 volts. So those gates are negative with respect to the jfet sources. But these are p-channel devices.......I am trying the understand the problem you were seeing.

JJ

My post was in reference to post #3, wherein the poster references another circuit. However, the same principles apply.
There's a reason that you don't see the "big guys" using circuits that drive devices past their limits: it's not a good idea.
Let's begin with the transfer curves that you can find in the datasheets. You'll note that for any device that we would describe as "linear," the chart stops when the control element--in this case, the Gate--hits the axis. True, you can find charts that go past the axis...and they're usually in textbooks as examples as to why you don't want to overdo it. In a nutshell, the behavior of the device becomes grotesquely nonlinear. Going a little bit over the line gives you distortion. Going a little bit further drives the device into catastrophic breakdown.
The whole point of John Curl's complementary JFET follower (which is what the poster was referring to, as I recall) is that each part--the N-ch and the P-ch--is matched against the other. The further you get from that ideal, the more you have to look at ways to rein in the stronger device. The typical way to do this is to add resistance under the Source of the part with the higher Idss in order to bring the pair back into balance.
To allow the Idss of one part to completely override the other is not a particularly good idea. Let's assume for the moment that you're not driving the underdog part so hard that it's drawing Gate current (this is bad), you're pretty much guaranteed that the JFET is in a very nonlinear state; not something I'd call a resistor.
This is DIY, you're pretty much free to do what you want. I choose not to do this particular thing.
Just for giggles, test the circuit for distortion, particularly with a strong input signal.
And yes, using the CCS to overdrive the gain devices in a differential gives you pretty much the same sort of problem. Depletion mode JFETs are not meant to be operated as though they are enhancement mode devices.

Grey

jupiterjune

Grey-
Thanks for the feedback.

I am going to spend some time playing around with the basic j-fet circuits to get a full understand of all the particulars of how they work.
Much appreciated.

JJ

Babowana

quote:

Originally posted by GRollins

My post was in reference to post #3, wherein the poster references another circuit. However, the same principles apply.
There's a reason that you don't see the "big guys" using circuits that drive devices past their limits: it's not a good idea.
Let's begin with the transfer curves that you can find in the datasheets. You'll note that for any device that we would describe as "linear," the chart stops when the control element--in this case, the Gate--hits the axis. True, you can find charts that go past the axis...and they're usually in textbooks as examples as to why you don't want to overdo it. In a nutshell, the behavior of the device becomes grotesquely nonlinear. Going a little bit over the line gives you distortion. Going a little bit further drives the device into catastrophic breakdown.
The whole point of John Curl's complementary JFET follower (which is what the poster was referring to, as I recall) is that each part--the N-ch and the P-ch--is matched against the other. The further you get from that ideal, the more you have to look at ways to rein in the stronger device. The typical way to do this is to add resistance under the Source of the part with the higher Idss in order to bring the pair back into balance.
To allow the Idss of one part to completely override the other is not a particularly good idea. Let's assume for the moment that you're not driving the underdog part so hard that it's drawing Gate current (this is bad), you're pretty much guaranteed that the JFET is in a very nonlinear state; not something I'd call a resistor.
This is DIY, you're pretty much free to do what you want. I choose not to do this particular thing.
Just for giggles, test the circuit for distortion, particularly with a strong input signal.
And yes, using the CCS to overdrive the gain devices in a differential gives you pretty much the same sort of problem. Depletion mode JFETs are not meant to be operated as though they are enhancement mode devices.

Grey

I fully agree that using the complementary matching against the other could be the best. But, this can not stop new way of experiment and fun. The overriding does not necessarily mean that it passes the Idss axis as you say. I’d like to think that the overriding means vertically shifting the Q-point keeping the original transfer curve’s slope. The overriding might not improve its own narrow span of current swing though.

The narrow span of current swing is not a big deal in my specific case. For example, Babo Zen needs very small current swing there. And, I don’t see any symptom of actual problem from my Babo Zen yet, except the very good sound. I will see further if I meet any problem, for your info.

:darkside:

EUVL

> did you compare the 1529 to the IRFP240´s (I´ve got plenty of the latter.....)

Yes. I have published the curve somewhere. They have similar transconductance. It is well known that I prefer 2SK1529s. But that is a matter of taste. Nelson does not.

One hint concerning the Aleph J could be the use of 2 output fets instead of 3 (6?)

How you tell !!!!!
I use 12 Power FETs per channel in total.
Don't be so Dutch. IRFPs cost nothing !!
:)

PS I am not German, onl live here. But you can still hate me nevertheless. (Joke)

Patrick

wuffwaff

Hi,

this morning I wanted to check the cold startup abs. dc offset after setting it to zero yesterday evening.
It´around 2,75 Volts wich means that I can easily raise the McMillan resistors to 20k or so.
I also checked the rel. dc offset wich was 4mV on the right and around 5V:eek: on the left......
Somehow while carrying the amp downstairs a source resistor had lost contact and there was a slight imbalance:rolleyes:
This probably put some tension on the left woofer causing it not to woof very nice anymore.
So after resoldering and reseting the missing bass found it´s way back into my living room:)

The sound is quite nice now.

When I´m back from holiday I will first raise the McMillans to 20k, this had quite some influence on the sound with the 9610´s and probably will with the 2sj74´s

William

Zen Mod

quote:

Originally posted by wuffwaff
Hi,

this morning I wanted to check the cold startup abs. dc offset after setting it to zero yesterday evening.
It´around 2,75 Volts wich means that I can easily raise the McMillan resistors to 20k or so.
I also checked the rel. dc offset wich was 4mV on the right and around 5V:eek: on the left......
Somehow while carrying the amp downstairs a source resistor had lost contact and there was a slight imbalance:rolleyes:
This probably put some tension on the left woofer causing it not to woof very nice anymore.
So after resoldering and reseting the missing bass found it´s way back into my living room:)

The sound is quite nice now.

When I´m back from holiday I will first raise the McMillans to 20k, this had quite some influence on the sound with the 9610´s and probably will with the 2sj74´s

William

bingo ;)

EUVL

Glad that it was not the JFETs at fault.

But I would still encourage you to try 910R/no Rsource, and perhaps 12 FETs per channel at the same total bias.

I know it is a lot of work though.

Patrick

wuffwaff

Hi Patrick,

yes I will. I will probably use some 5R source resistors with a 50R trimmer in parallel though to avoid having to buy another 50 2sj74´s (yes I know I´m dutch)

I will also change the McMillan resistors to a value as high as possible.

I´m already using 12Fets per channnel (IRFP240). Changing them to 2sk1529 would really be a bit of work as they are all hard wired and not so easy to change. It is probably better to sell mine and build some new ones.....this way I could make them a bit wider to fit a non humming transformer.

William

EUVL

If you have problems getting 2SK1529, let me know.

Also you would need to experiment with the source resistor for 2SK1529s. I have used up to 1k for 4 FETs per channel, but no compensation caps at the feedback resistor. You need to experiment.

Patrick

wuffwaff

Hi Patrick,

what do mean with source resistor? 1k seems a bit high for the power fet source resistors. I´m using 0,33R at the moment (3 x 1R 1%).

William

EUVL

Sorry, typing mistake.
Gate resistor I meant.

Patrick

wuffwaff

Hi Patrick,

because of oscillations or to tune the frequency responce?
What is the bandwidth of your amp?

William

EUVL

Because of oscillations.
My bandwidth is about 350kHz at -3dB.

Patrick

wuffwaff

Hi Patrick,

that sounds good!

I´ll think about the 1529´s while cycling on Mallorca for the next ten days.
This evening the amps sound really nice with the JFet inputs.

cheers,

William

wuffwaff

Hi,

so after listening for a while with the last configuration I was still not happy with the lower bass. On tracks where this wasn´t a problem the sound was very good.
So the next step was to change the McMillan resistors from 10k to 22k.

This seems to improve things quite a lot (like going from 4k7 to 10k with the IRF9610´s). More bass and a nicer sound overall. Absolute DC ofset from cold doubled (as expected) to 7V but goes down quite fast.

Will leave it like this for a while and then change the drain resistors to 910R and source resistors to 5R (for dc offset).
I´ll also have to take a closer look at the high frequency behaviour at high currents.....

William

wuffwaff

Hi,

this morning I did the next step:

Changed the drain resistors to 910R
Changed the source resistors to one 10R pot for setting dc offset (relative)
I also had to change a resistor in the current source to be able to set it low enough.

After this I had a look at the active current sources and C9 and C10. After reading this:
http://www.diyaudio.com/forums/show...10&pagenumber=6
I tried some different values (In my case C9 and C10 where 3n3 for an optimal square wave)

So I started with 150pF:

frequency response 142/130kHz 8R / 4R
square wave has a little edge cut out of the leading edges
10kHz 4R is now better (16,6V) but starts oscillating after clipping

220pF:
almost the same as 150pF, square wave a bit better

330pF:
somewhat better behaviour but still oscillating after clipping

470pF:
no oscillation anymore
frequency responce 153/137kHz 8R/4R
10kHz/4R now OK until 19V (near clipping)

a strange thing was noticed on a pure sinus 10kHz into 8R (see pictures). I couldn´t get this away by raising the C´s nor by changing the feedback C´s from 5 to 10 or 20pF

680pF:
this showed the same behaviour as before (with the 3n3): The sinus into 4R (10kHz) went up to 15V or so and then suddenly broke down to something very awful.
the square wave showes some overshoot (small)

So at the moment I´ve put in the 470pF but although better, the behaviour at high voltages is still not OK. The square waves show some strange edges (most obvious at lower voltages) and the sinus at higher voltages has two dips in it.
I haven´t listened to it but I´m shure this won´t really matter at my normal listening levels but I would like it to go away....

any ideas?

William

EUVL

With so much information, I GUESS you might still have too much high freq gain. What cap do you have parallel to the feedback resistors (100k) ?

It would help the diagnosis if you can do a frequency response sweep from say 100Hz to 500kHz or 1MHz.

Patrick

wuffwaff

Hi Patrick,

the feedback caps are 5pF. I also tried 0, 10 and 22pF.

0pF made the square wave overshoot with some ringing
22pF rounded the square wave and reduced the frequency responce to 65kHz / 8R

I will do a frequency sweep tomorrow (this morning I looked at 20, 100, 1k, 10k, 20k, 100k and the -3dB point. Nothing unusual here)

thanks,

William

EUVL

In my setup I am now using 10p, but I have tried 5p and apart from a slight overshoot at 10kHz square wave, I can probably get away with that, even 3.3p.

Caps at the current source is the same as yours, 470p. But I don't use the bootsctrap cap and I have batteries. So it is not 1:1 comparison.

And I don't do my tests close to clipping. I usually test at 4V from rail max. IMHO, you need to give the FETs a few volts or they get into a region where their capacitances jumps up enormously, and many funny things happen there.

But maybe Nelson would what to comment on that.

Patrick

PS Freq Sweep up to 100kHz won't be enough if you have excessive gain at say around 300kHz. That is why you need to go further to at least 500kHz. But a 20kHz square wave (overshoots) will tell you whether you have a HF problem or marginal stability.

wuffwaff

Hi Patrick,

I did my tests at 1V, 10V and close to clipping. At 1V, 10V there´s no problem to be seen (except for the funny square wave, will take a pic tomorrow)
Only starting at about 20V/10kHz into 8R the funny sinewave starts to show getting worse at higher voltages.

4V from rail max would be 2x (22-4) = 36V peak so about 25-26volts eff.

What I don´t have used is C7 and C8 but I don´t think anybody used them.

William

jupiterjune

William-

How are you measuring the output -- is that a summation of the voltages at each side of the resistor?

JJ

wuffwaff

Hi,

I´m using two methods:

1. with two probes, one for each output, then added at the scope.

2. with a sym. to unbalanced converter (INA134) fed with the +out and -out via a resistor network and then with a single cable to the scope

Both are good enough to look at waveforms, the measuring of the ac voltage is performed by an old Fluke device.

William

wuffwaff

Hi Patrick,

here´s the sweep into 8R:

20Hz 10,19V
50Hz 9,88
100Hz 9,88
1k 9,90
10k 9,91
20k 9,91
50k 9,67
70k 9,29
100k 8,70
150k 7,18
200k 5,32
300k 2,73
400k 1,43
680k 0,5
840k 0,35
1160k 0,16
1500k 0,05

No rising in between points. Measured with a Fluke 8500A and watched on a 20MHz scope. Measured with two probes.

Nothing I can see here.

Also checked the input signal wich is kind off flat until way over 1MHz and doesn´t show the distortion.

William:xeye:

wuffwaff

Hi,

I also checked at 22.5V where the distortion already shows.

With rising frequency the two dips walk towards the 0V line and reach it at around 40kHz. Checking up to 1.5MHz there were no signs of oscillation.:confused:

William

P.S. I almost forgot: The sound is very good at the moment. The mids and highs have gained quite a bit in the way different textures (sound colours and nuances are played).

EUVL

William,

Can you post a square wave response ?
Just use the 1kHz square wave from your scope calibrator.

Also do you have a means to generate a balanced input ?
And try to measure one input (say +Vin) and one output (say +Vout) with your scope and see how the input looks like when you have this distorted sine.

I am a bit puzzled this time as well.
But still, I am glad you like the sound. :)

Patrick

Manu

Hi William,
Could you post a schematic of your setup? I am a little lost , without.
manuel

wuffwaff

Patrick,

square wave is possible but not spectacular (and I have to carry the amp up two stairs again......).

It looks very normal at 1kHz / 10V .
It still looks good at 0,5V but here you can see a very small edge cut out of the square waves edges (ca 2-4% of the height and width).

I do use a balanced input (DRV134) and checked the waveform when connected to the amp (was OK) but I will have a look again.

Manuel,

it´s a standard Aleph-X with 12 Fet´s/ channel and 100w / 8R. The only thing changed at the moment are the input fets from IRF9610 to 2sj74.

William

xavier1000

Hi William,

Did the behaviour you are describing appeared after you changed the frontend to jfets or it could be something that was there even before you did that change, as we are experiencing me and Paolo (sinuko)?

Did you check the output of each side of the monoblock refered to ground when the sinus wave is like the one you show in your last picture? In my case, I noticed that each side of the amplifier were clipping only at the positive pole, but not at the negative.

The cap just improved things but the behavior is still not perfect for high frequencies. I really don't know what is causing this distortion, but it seems that it is comming from the CCS. Please keep us updated on your progress.

Xavier

wuffwaff

Hi Xavier,

yes I looked at both outputs. They showed the same behaviour.

An Aleph never clips on the negative pole because if it weren´t for the current limiters (Q4, Q9 plus resistors) it would sink as much current as your load allows for.

This behaviour wasn´t appearant before I changed to Jfets (although I´m not 100% shure)

Maybe I also have to look at R14/15 and R31/32 again.....

William

RoboMan

Hi William,

Referring to the sine wave in your post 46, I think the dips on the falling edge is caused by over-Aleph current and the distortion at the tips is caused by the slow response of the CCS or not enough bias.

Try these if you have not done so:
(since you did not post the schematic, I'm using Grey's diagram dated 5/24/02)

1. Replace R12 and R34 with a higher value (say 1.8K or use a trim), this will lower the Aleph current.
2. Remove any cap on B-C of Q3 and Q8.
3. If no effect on step 2, increase bias by increase VR1 and VR3, this will cure the distortion at the tips.

wuffwaff

Hi,

my ac-current gain is set at precisely 50%. The power into 8 ohms is voltage limited, not current limited. (100 watts 4R is the limit). I´ve tried removing R12/34 but this way the amp will be current limited into 8R.
I also think the cause is somewhere in the active current source but I´m not shure where.

The bias is around 7A wich should be enough

Removing the caps causes oscillation.

BTW I´m referring to the HifiZen schematic,

William

RoboMan

quote:

Originally posted by wuffwaff

my ac-current gain is set at precisely 50%. The power into 8 ohms is voltage limited, not current limited. (100 watts 4R is the limit). I´ve tried removing R12/34 but this way the amp will be current limited into 8R.
I also think the cause is somewhere in the active current source but I´m not shure where.

The bias is around 7A wich should be enough

From your spreadsheet, you'll need 8A for 8R (64W) without ac-current gain. Have you tried 25% and a little higher bias?

wuffwaff

Hi,

I can try but this is no option for my speakers.
As this distortion is not appearant on 4R I think it is more voltage dependant as current.

William (who is just taking pics of some square waves)

wuffwaff

OK,

some more information:

1. with no load the distortion is the same at the same voltage.

2. with the active current source deactivated the distortion stays the same (removed R12/34)

3. with R14/31 from 1k5 to 2k7 the distortion stays the same

4. as soon as the distortion sets in you can see the dips on both power supply voltages as dips over 200mV big. This means that there are current peaks wich can only be caused by bias peaks wich are influenced by the whole current source

So somehow above around 22V output ( 31V peak, 15,5V per side) something happens causing the bias to peak. This happens twice per cycle.

William

wuffwaff

Here´s a pic of the output at ca.25V and the V+ supply, the peaks are almost 400mV

wuffwaff

Here´s the 20kHz square wave 11V into 8R

wuffwaff

and 1V, 20kHz 8R

EUVL

William

The square waves look fine to me, so I don't think you have an oscillation problem.

The thing I find strange is that the rail dip occurs half way on the downward slope. If you have a differential probe, maybe you can check the output current. The fact that you still get this with the active current source deactivated implies that the latter may not be the cause.

The other measurement I would suggest, just to trace where it might come from, is to measure the voltage across the JFET drain resistors. Again you need to measure with a diff probe, since you need at least one channel at the output to show the phase relationships.

:confused:

Does it get worse with 16R ?

Patrick

wuffwaff

OK:mad: I think I´ve found it.....

looking at the voltage at the drain resistors there´s a strange peak when the distortion sets in.
Looking at the voltage before the source resistors there´s a sinus with almost 0,15V PP just before the distortion sets in. this means that Vgs is changing all the time and when the voltage reaches 0.25V it can´t go any higher, sort of clips and causes this distortion.

Taking ot R46/47 (I tried:D ) is no option as the absolute dc offset get´s a bit unstable.

Changing them to 33k (from22k) makes things better at least until about 27V are reached (wich is almost at clipping)

If you look at the waveform at the jfet sources they start distorting a few volts before the output starts distorting because of the feedback keeping things together.

This also explains why changing the McMillan resistors to bigger ones made such a big sonic change.

So what to do next?

Put in a servo and very big (>50k)McMillan resistors?

Change all output devices to toshibas with lower input capacitance?

Put in the IRF9610´s again?

I could up the source resistors but I don´t think this is a good sollution.

Any ideas???:confused: :confused:

William

wuffwaff

Hi Patrick,

you answered while I was answered but I sort of did what you wanted Although my explanation might not be very scientific it certainly has something to do with Vgs reaching 0V and getting into the not so linear area.......

William

sinuko

Hi, William
“Put in the IRF9610´s again?” ..
At the moment with 12 fet per channel (21.5 rail voltage) I have the same problem using irf9610. (In my case Mc Millian resistor are 4.7k).

Ciao Paolo
:bawling: :bawling: :bawling: :bawling:

EUVL

William,

I have done over-bias of JFETs as Borbely suggested to up to -0.3V for P-type JFETs and they do just fine. So going beyond Vgs=0 is not a problem for me. There are I am sure different opinions on that.

If I understand correctly, you are getting a high common-mode dynamic error from both outputs which got fed back through the so-called McMillan resistors. That implies you are operating the power MOSFETs in a region with relatively large even order harmonics. Changing to 2Sk1529 would not solve your problem. And the fact that Nelson is using JFETs with 40 FETs per channel means that the solution lies somewhere.

I remember that you are using 0R33 source resistors at the output stage. Try 0R47, as that would sort of linearise your output stage a bit and reduce this common mode error. Reduceing the 100R to ground at the outputs would probably have similar effects.

Another solution to cancel the even order harmonics (on each side of the X) is to drive the current source out of phase with the driver MOSFETs with a phase splitter, a la Linsley Hood. But then it would not be Aleph-X anymore.

Patrick

wuffwaff

Hi Patrick,

I´ll try to explain the problem again:

looking at the source of the input fets everything is fine up to around 0,15V sinus.

At this moment this voltage sort of starts to clip but only at the sources, the output looks fine even when distortion at the sources (Jfets) is already heavy. For me this means that the McMillan resistors can´t deliver anymore current cause nothing is passing through because maybe Idss is already reached?

Raising the McMillan resistors, meaning less feedback to the sources makes things better. What has this got to do with the source resistors at the output fets?

William

wuffwaff

Hi Paolo,

I had some great sonic succes raising those from 4k7 to 10k. Try this and see if the problem gets less (or maybe even is gone)

William

jupiterjune

William-

jupiterjune

Sorry about that last one--some of the keys on my keyboard stick!

What I was going to say is that I find it very interesting that you are getting dips in you power supply voltage.

I think it would be safe to say the power supply's ability to output current at that point is being exceeded. But -- is it because of some fault in the power supply, or is it because the circuit is drawing a lot of "extra" power at that point? It would seem it must be one or the other.

You seem to be focusing on the amp circuit -- I am going to spend some time looking at the McMillan resistors to understand what you suspect may be the problem.

Very interesting thread......

JJ

jupiterjune

quote:

So somehow above around 22V output ( 31V peak, 15,5V per side) something happens causing the bias to peak. This happens twice per cycle

If I understand the x-topology correctly, wouldn't the output on the other side be drawing peak current from the V+ rail? So it might only be once per cycle, shortly after the positive going peak?

JJ

EUVL

> What has this got to do with the source resistors at the output fets?

Measure the total current through the diff pair when playing music, and you will see part of the music signal being feedback as current to the diff pair through them, as well as absolute DC. This is because the Id / Vgs of the output FETs are nonlinear. Although the diff amplitude is largely linearised by the bridged configuration, there is still a common mode component.

But I cannot visualise your description as yet, so I cannot be sure this is the cause of what you see. Give me a bit more time to think about this again.

Patrick

wuffwaff

Hi Patrick,

also thought about this a few times this night. I know understand what you mean with common mode.

0,15V of 31V is about 0,5% right?

Could this common mode error also be caused by the R46/47 not being perfectly matched (they are "only" 1% resistors)

Could the fact that I´m using different source resistors (10R pot adjusted for minimal rel. dc offset) have something to do with it?

thanks,

William

EUVL

> Could this common mode error also be caused by the R46/47 not being perfectly matched (they are "only" 1% resistors)

I doubt. I mean not too long ago, we only have 1% resistors. A circuit has to be designed to work with 1% component variations.

> Could the fact that I´m using different source resistors (10R pot adjusted for minimal rel. dc offset) have something to do with it?

Also not if you think about the diff pair as a monolithic components (black box). When driven by a CCS, it always deliver the same dynamic current at the output for a certain differential input voltage.

Send me a private email tonight. I'll send you some of my analysis of the Aleph-X circuit and how I deal with those issues. Not that you should follow my approach, but at least it might help to understand the current 'problem' better, I hope.

If we suspect that the JFETs are causing the distortion, it can only be two scenarios :

1) It hits a negative Vgs < -0.7V, and the gate conducts (the FET goes into reversed bias).
2) It bottoms out at threshold Vgs, i.e. not enough dynamic range.

That is why it puzzles me so much that you need so much dynamic range in the first place. I assume that you have at least 30dB NFB at 8ohm ?

Patrick

AMV8

William

I am learning so much about the aleph X on this thread. But I guess you want to repair the amp so.....

Have you tried to replacing the input pair with your original semiconductors to check if the new input pair is causing the problem. Changing back seems to be the only way to know for sure if they are causing the problem.

Other than that it sounds as though you are just starting to operate a device in its non linear region which can sometimes cause a kink in the curve.Possibly just one semiconductor not quite to spec. Can you check the current accross a few of the 0.33 ohm resistors to see if you get the same problem with each 0.33 ohm resistor/ output device.

Other than that it just seems that the amp is running out of power. Possibly a connection in the power supply?

Don

wuffwaff

Hi,

the output is not the problem, the input pair is. Both amps have the same problem so I don´t think there´s a defect component .

I will try the original pair. I only need to change the drain resistors, Mcmillan resistors, current source resistor and the input pair........

Like I said the problem also exists without a load!

William

EUVL

> Like I said the problem also exists without a load!

That is the most puzzling part. That kills my common-mode hypothesis.

I need to sleep on this for a few days.

Patrick

PS What is your current source ? JFET ?

wuffwaff

Hi Patrick,

no the current source is a 9610 with 3 2.9V zeners (for a negative tempco.)
Sleep on it!
I won´t do anything today cause I´ve installed some new software for my cars ECU wich ofcourse means that it won´t start anymore (computers are a lot worse than amp!)

thanks,

William

wuffwaff

Hello and good morning,

just refitted my 9610, Rdrain 390R, R46/47 10k and another R in the current source.

The good news is that the distortion at higher voltages/ frequencies stays the same meaning that the Jfets are not to blame.

The bad news is that the distortion is still there!

:bawling: :bawling: :bawling: :bawling: :bawling: :bawling: :bawling:

Help!

William

P.S. I won´t kill myself, put back in the Jfets and just don´t listen to 10kHz tones at 26 volts and 8R until I find a sollution or somebody else has a nice idea........

Netlist

William,
Dumb question: Aren’t the protection transistors kicking in?

Furthermore, I think your diff pair current source has a problem. If you still have a pot in there, try a few different settings while looking at the waveform.
Bear with me, after all that’s been said and done, this is about the only thing I can come up with.

/Hugo

wuffwaff

Hi Hugo,

I´ve removed the protection network (R39/40)

no difference

I tried a few difefrent currents through the cs as I changed the drain resistors. I can´t realy play with it as it also changes DC offset.

I think it is a designed in problem as both my amps have it and also Sinuko ahs the same problem. Some others may not have noticed......

It will probably only affect A-Xes with more than 22V power supplies

William

sinuko

Hi William,
very bad bad news

:bawling: :bawling:

me too at the moment have not any good news

Paolo

wuffwaff

Well,

let´s see it realistic. As long as it sounds good (because normally you will never see such high amplitudes at those frequencies) and you don´t have to give it away for a magazine test no one will notice......

But when somebody has an idea I´m open for this. Shouldn´t it be possible to simulate this?

William

wuffwaff

Hi,

I must correct myself. The distortion is not exactly the same. It starts a bit later (at a somewhat highr voltage) than with the Jfets.
I suppose this is because there´s a bit more open loop gain and thus a bit more available feedback?

William

P.S. I think I will go and wash my car as normal people do on saturdays (must be living in Germany for too long.....)

Netlist

quote:

Originally posted by wuffwaff
It will probably only affect A-Xes with more than 22V power supplies

I'll have a look at mine. It has +/-30V rails.

/Hugo

xavier1000

quote:

Originally posted by wuffwaff
It will probably only affect A-Xes with more than 22V power supplies

I don't think so, mine is 17,5v. It might not affect the original one with 15v.

At this point, I am running out of ideas also. May be we should try to find similarities between the three identified circuits that are presenting this behavior and we should also find a circuit that works fine. Anyway, I also think that a lot of people are having this behavior without knowing.

Any happy user wants to be the first in checking whether his amplifier is presenting this behavior or not :devilr: ?

Xavier

xavier1000

I forgot to mention some relevant characteristics of my AX so we can compare with sinuko and wuffwaff:

Voltage at fets: 17,5v
Number of fets per channel: 8 (IRFP044)
Bias: 7,3A
AC Current gain 50,6%
McMillan resistor: 10k

Xavier

EUVL

William,

This is my reasoning and it seems to fit what you described.

Your AX should have an open loop gain of about 300 at 8 ohm (and more at no load), meaning that you have NFB of a factor of 30.

At an output of 30V pk-pk, your input is 3V pk-pk, and I expect you to have 0.1V pk-pk across the gate of the diff pair.

Assuming that you really have poor linearity (i.e. lots of 2nd harmonics) at the diff pair, and 10% of that 0.1V appears as common mode voltage at the source-junction (or CCS out) of the diff pair, you should not see more than 10mV there.

But you were seeing somthing like 0.15V, and I suspect that this can only be common mode being fed to the loop via the McMillan resistors.

This large common mode voltage should not be there in the first place (at least not such large amplitude).
Because it is common mode (my hypothesis), the diff pair will do nothing to that and just pass it on.
This probably happens at no or light load, because you have a slightly excessive HF gain.
This is damped when the load increases (or load impedance decreases, e.g. at 4 ohm).

If the above hypothesis (and I stress that) is true, then the following may be the cure :

Connect the two outputs using 2x 20k resistors (as in the McMillan resistors).
Instead of feeding the node mid-way of the 2 resistors directly to the diff pair common source, connect this to the common source via a single, third resistor of say 5k.
Add a cap (say 22uF, best non-polar) to the junction of the 3 resistors and tie the other end to ground.

This way, you still feed back common mode error (absolute DC) at low frequencies, but prevent any HF interaction between the diff pair and the output (grounded by the cap).

I admit I have not tried because I do not have the problem, but the experiment is simple enough to do.
You can probably get away with a small 22uF100V electrolytic for the purpose of the trial.

Hope you get what I mean.

Patrick

Zen Mod

appropriate valued resistor between drains in diff pair?

rtirion

William,

I remember struggling with similar trouble (simulation first, and
later with the amp). See original Aleph-X thread. Fred AKA .....
did help back then in his special way.

Now, to reproduce the results was a little bit difficult.
But to make a long story short, I used source degeneration of
the input pair as the solution.
I have been able to reproduce the behaviour (to some extent)
of your amp in a ALEPH 3 sim.
See attached document.

Sorry for the short answer.
I hope this helps a little.

wuffwaff

Hi,

Patrick,

I will try this (sounds simple enough).

Rtirion,

this behaviour is there before the amp clips.

William

sinuko

Hi William,
I tried with a variac to change the rail voltage, at lower voltage it seems to be better (as you suggested)..(but I need more tests)

Paolo

I would cancel my post here http://www.diyaudio.com/forums/show...456#post1172456

EUVL

> appropriate valued resistor between drains in diff pair?

A resistor between drains acts as an attenuator. This reduces gain at all frequencies, just like adding source degeneration. While it might cure the problem, it also affects overall performance (i.e. amount of negative feedback). You might like this, or you might not.

There are at least 2 more ways to reduce gain at HF (for both diff & common mode) :

1) Put a (polystyrol) cap, optionally in series with a resistor between the diff pair drains. This attenuates any HF signal coming out of the diff pair.

2) Add a Sobel network (e.g. 10R, 0.1uF) at each output to ground, (or just one between the outputs). This adds load at HF.

But if my hypothesis is correct, then trying to break the common mode feedback loop (due to the McMillan resistors) at HF would just cure the problem without changing the diff mode performance (which again might or might not be better or worse).

Patrick

wuffwaff

Hi Patrick,

(no I didn´t try yet). Wouldn´t the cap sort of compromise the CS function (as do the McMillans) in that it´s possible to have more current through the diff pair than the 10mA?

William

Page generated in 0.62236094474792 seconds with 17 queries,
spending 0.09688759 doing MySQL queries and 0.52547336 doing PHP things.

Powered by: Search Engine Indexer and vBulletin
Copyright ©1999-2008 diyAudio.com