For automagic matching, you could Blumlien garter the CCS pair.
If you are planning to fine tweak both for some specific current,
I suppose the garter enhancement would be irrelevent...
Bridge the cathodes with a fine non-polar audio type cap, and
Then like Shoog said, two large polars in back to back series,
with a bleeder to keep em' formed.
Gotta remember when scaling the bridge capacitance, impedance
between cathodes is rather low... And series caps throw half out
the window. They might need to be kinda large.
10M45S internally might be nothing more than a "ring of two"
enhancement mosfets. Just as one might do with bipolars,
abusing VBE as a reference. But with much less predictable
gate thresholds, somewhere in vague neighborhood of 4.5V...
Pass' current source uses a combination of bipolar VBE threshold
as the voltage reference in the ring of two, and then MOSFET for
the current blocker. This makes better sense as VBE seems fairly
repeatable. And the MOSFET can operate the drain very close
to the voltage reference. You can get this working stable down
two Volts or less.
The top device in ring CCS might better yet be implemented as
an IGBT if one were going with discretes all the way. Of course,
you then have a voltage drop VCE to consider, I don't think you
want to be working anywhere near saturation with an IGBT.
But these are readily available to 1700V, if you need that sort
of blocking performance with a very small gate drive.
If you are planning to fine tweak both for some specific current,
I suppose the garter enhancement would be irrelevent...
Bridge the cathodes with a fine non-polar audio type cap, and
Then like Shoog said, two large polars in back to back series,
with a bleeder to keep em' formed.
Gotta remember when scaling the bridge capacitance, impedance
between cathodes is rather low... And series caps throw half out
the window. They might need to be kinda large.
10M45S internally might be nothing more than a "ring of two"
enhancement mosfets. Just as one might do with bipolars,
abusing VBE as a reference. But with much less predictable
gate thresholds, somewhere in vague neighborhood of 4.5V...
Pass' current source uses a combination of bipolar VBE threshold
as the voltage reference in the ring of two, and then MOSFET for
the current blocker. This makes better sense as VBE seems fairly
repeatable. And the MOSFET can operate the drain very close
to the voltage reference. You can get this working stable down
two Volts or less.
The top device in ring CCS might better yet be implemented as
an IGBT if one were going with discretes all the way. Of course,
you then have a voltage drop VCE to consider, I don't think you
want to be working anywhere near saturation with an IGBT.
But these are readily available to 1700V, if you need that sort
of blocking performance with a very small gate drive.
About the bridging capacitor: It goes from one cathode straight to the other right? Am I right in assuming that it alleviates the need to "fine tweak" to match current? That would be great since I stupidly didn't leave the 68-ohm cathode resistor exposed to allow for testing.
What size do you think I might want? At the moment one channel has two 150uF electrolytics and the other has two 220uF types. I have four low-impedance 2200uF caps I'd intended for a Gainclone that I might use: is 2200uF ok? Does the size of the electrolytics affect the size that this bridging capacitor needs to be?
You lost me at "10M45S internally might be nothing more than a "ring of two" enhancement mosfets".. My solid state theory is nothing to be proud of. In layman's terms, what are you suggesting that I could do?
What size do you think I might want? At the moment one channel has two 150uF electrolytics and the other has two 220uF types. I have four low-impedance 2200uF caps I'd intended for a Gainclone that I might use: is 2200uF ok? Does the size of the electrolytics affect the size that this bridging capacitor needs to be?
You lost me at "10M45S internally might be nothing more than a "ring of two" enhancement mosfets".. My solid state theory is nothing to be proud of. In layman's terms, what are you suggesting that I could do?
The bridging cap merely unifies the current sources into a single
tail for music signals. Yet remain as separate tails for purpose of
balancing DC.
Blumlein's garter bias was for eliminating the need to fine tweak
the DC balance...
Its a tube thing, so you should have no probs understanding or
finding older refs to it. Except in your case, applying this balancing
trick to G , K , and current sense resistor of the CSS devices.
"Ring of 2" and variants... Don't drive yourself crazy looking for
Blumlien's garter, not yet in this picture. There's a writeup about
it at Tubecad Journal.
I'll draw a ring of two (or is it four?) garter later, if you havn't
got it figured for yourself by then..
tail for music signals. Yet remain as separate tails for purpose of
balancing DC.
Blumlein's garter bias was for eliminating the need to fine tweak
the DC balance...
Its a tube thing, so you should have no probs understanding or
finding older refs to it. Except in your case, applying this balancing
trick to G , K , and current sense resistor of the CSS devices.
"Ring of 2" and variants... Don't drive yourself crazy looking for
Blumlien's garter, not yet in this picture. There's a writeup about
it at Tubecad Journal.
I'll draw a ring of two (or is it four?) garter later, if you havn't
got it figured for yourself by then..
Attachments
There is a great difference between the Garter circuit and using a true CCS, and that is in the form of the actual cathode impedance and the degree of differential operation it allows. The garter works really well and sound absolutely fine - but it doesn't represent a true differential stage. The CCS with the cap node referenced to earth via a 1meg resistor represents a true differential stage. Horses for courses, but since you have the IXY chips then I think the garter would be a step backwards.
DC balance is only as good as the matching of the IXY chips and my small experience is that that is not very tight (expect it to be 1-3mA off which should stay constantas the valves age). If you are using normal PP transformers this shouldn't present a major issue, and a slight imbalance may even enhance the sound by keeping the transformer core saturated at zero crossing. Hum is your only issue.
I have had success with bog standard 1000uf caps in the bridge, and that is with 6AS7 triodes which have much lower output impedance than you are likely to encounter, so your 2200uf will do fine. I would personally be interested to hear if you get any bass extension when moving to bigger caps.
Shoog
DC balance is only as good as the matching of the IXY chips and my small experience is that that is not very tight (expect it to be 1-3mA off which should stay constantas the valves age). If you are using normal PP transformers this shouldn't present a major issue, and a slight imbalance may even enhance the sound by keeping the transformer core saturated at zero crossing. Hum is your only issue.
I have had success with bog standard 1000uf caps in the bridge, and that is with 6AS7 triodes which have much lower output impedance than you are likely to encounter, so your 2200uf will do fine. I would personally be interested to hear if you get any bass extension when moving to bigger caps.
Shoog
This is what I meant, about the garter matched CSS's.....
And the caps I show at the bottom aren't needed if you are
bridging a cap (or caps) across the top. Unifying the tails...
Note: that I am showing internal details of 10M45s that are
entirely my guesswork... Not based upon any known fact....
I don't have a clue how the upper gate might be pulled up,
Nor exactly how its protected from the excessive voltage.
And the caps I show at the bottom aren't needed if you are
bridging a cap (or caps) across the top. Unifying the tails...
Note: that I am showing internal details of 10M45s that are
entirely my guesswork... Not based upon any known fact....
I don't have a clue how the upper gate might be pulled up,
Nor exactly how its protected from the excessive voltage.
Attachments
Thanks Ken, I'll give it some thought. For now I substituted in the 2200uF caps and fixed my wiring error. I also equalized all the CCS resistors to within .1 ohms. There's a night and day difference between the amp yesterday and today, in every aspect: bass, clarity of treble, etc. etc.
Obviously I can't link the change to one factor or another but needless to say, I'm quite happy. Both channels are now pitch black with no input as well
Obviously I can't link the change to one factor or another but needless to say, I'm quite happy. Both channels are now pitch black with no input as well
Well, the equalization I was concerned about wasn't the resistors...
Its MOSFET gates. And if 10M45s uses VGS-On threshold of one or
more of those gates as a reference to compare the sense resistor,
then that threshold is subject to a large manufacturing variance.
The actual figure might be 3 to 5 volts. Then again, they might be
very identical, especially if cut from the same batch... Who knows?
If you measure the "K"athode voltages to be the same, then the
currents in your 1%'s are the same, and you got no such issues.
If not, then Blumlien's garter would be one alternative to manual
trimming of resistances to assure those currents always match.
Sand won't change much with time, but it does with temperature.
Its MOSFET gates. And if 10M45s uses VGS-On threshold of one or
more of those gates as a reference to compare the sense resistor,
then that threshold is subject to a large manufacturing variance.
The actual figure might be 3 to 5 volts. Then again, they might be
very identical, especially if cut from the same batch... Who knows?
If you measure the "K"athode voltages to be the same, then the
currents in your 1%'s are the same, and you got no such issues.
If not, then Blumlien's garter would be one alternative to manual
trimming of resistances to assure those currents always match.
Sand won't change much with time, but it does with temperature.
Well, I can't find any model for the 10M45S, but do find 10M90S.
Unfortunately, this model isn't giving clues to the internal detail
of the error amplifier. Treating it as-if a single depletion device.
Still suspect business end of this stick is a ring of two MOSFETs.
* PSpice Model Editor - Version 9.2.1
*$
*DEVICE=IXCP10M90S
.SUBCKT IXCP10M90S G A K
*** G: Gate
*** A: (+) Positive Terminal
*** K: (-) Negative Terminal
M_M1 A G K K IXC_10M90S
D_D1 K A 10M90S_D
* updated using Model Editor release 10.0.0 on 02/20/06 at 12:40
* The Model Editor is a PSpice product.
.MODEL IXC_10M90S NMOS
+ LEVEL=3
+ L=2.0000E-6
+ W=19.000E-3
+ KP=1.0547E-6
+ RS=10.000E-3
+ RD=69.978
+ VTO=-4.4878
+ RDS=1.0000E9
+ TOX=2.0000E-6
+ CGSO=526.32E-15
+ CGDO=5.2870E-9
+ CBD=85.184E-10
+ MJ=1.5000
+ PB=2.6056
+ RG=10.000E-3
+ GAMMA=0
+ KAPPA=0
.MODEL 10M90S_D D
+ BV=900
.ENDS
*$
Unfortunately, this model isn't giving clues to the internal detail
of the error amplifier. Treating it as-if a single depletion device.
Still suspect business end of this stick is a ring of two MOSFETs.
* PSpice Model Editor - Version 9.2.1
*$
*DEVICE=IXCP10M90S
.SUBCKT IXCP10M90S G A K
*** G: Gate
*** A: (+) Positive Terminal
*** K: (-) Negative Terminal
M_M1 A G K K IXC_10M90S
D_D1 K A 10M90S_D
* updated using Model Editor release 10.0.0 on 02/20/06 at 12:40
* The Model Editor is a PSpice product.
.MODEL IXC_10M90S NMOS
+ LEVEL=3
+ L=2.0000E-6
+ W=19.000E-3
+ KP=1.0547E-6
+ RS=10.000E-3
+ RD=69.978
+ VTO=-4.4878
+ RDS=1.0000E9
+ TOX=2.0000E-6
+ CGSO=526.32E-15
+ CGDO=5.2870E-9
+ CBD=85.184E-10
+ MJ=1.5000
+ PB=2.6056
+ RG=10.000E-3
+ GAMMA=0
+ KAPPA=0
.MODEL 10M90S_D D
+ BV=900
.ENDS
*$
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.