Ive been doing a little PSpice simulation work with various amplifier and preamplifier types, looking at the differences in harmonic distribution between various topologies and using this information to decide which circuit to prototype. I realize that the simulator won't tell the whole picture about a design, but I found that with folded cascode topologies , the harmonic spectrum was rather strange, with the odd harmonics predominating 10-20dB over the even. This difference is rather persistent, and holds true both with simple and much more complesxdesigns.
I also remember reading about the development of the Pass A75. It seems a folded cascode topology was tried for a while with the A75, but abandoned for sonic reasons. (Mr. Pass, can you comment?) Has anybody here had the chance to measure the harmonic spectrum of a folded cascode amplifier? I'm curious to see if the simulations correlate in any way with real-life results. I am definitetly going to try it some time, but first, I have to learn to master the Audio Precision analyzer at my work place...
I also remember reading about the development of the Pass A75. It seems a folded cascode topology was tried for a while with the A75, but abandoned for sonic reasons. (Mr. Pass, can you comment?) Has anybody here had the chance to measure the harmonic spectrum of a folded cascode amplifier? I'm curious to see if the simulations correlate in any way with real-life results. I am definitetly going to try it some time, but first, I have to learn to master the Audio Precision analyzer at my work place...
As I know, mr. Pass uses current folded cascode in X amps. I think that it is the best amnplification stage yet. But you have to fine tune this circuit, differential stage have to be biased quite high, and on the output side of circuit you need as low impedance as possible to get good results.
I use current folded cascode in my amplifier and I'm biasing differental stage at 12mA, while using only 2mA for signal and impedance at the output is 10kohm, I also use current mirror at the output.
If you can use very high bias at differential stage with at least 20mA for signal, you can drive output mosfets directly (if you will use mosfets) with very low dostortion.
I use current folded cascode in my amplifier and I'm biasing differental stage at 12mA, while using only 2mA for signal and impedance at the output is 10kohm, I also use current mirror at the output.
If you can use very high bias at differential stage with at least 20mA for signal, you can drive output mosfets directly (if you will use mosfets) with very low dostortion.
I thought the A75 sounded fine with folded cascode - it was
just one of several options we presented.
Properly done, folded cascode behaves just like a regular
cascode and sounds the same. The real difference is that
you can use a folded cascode simultaneously as a level shifter
to get a direct coupled output with a single-stage (or more
complex, if you like) amplifier.
The emphasis is on properly done; The CCS which supplies
both the input gain device (the one being cascoded) and the
cascode device has to be high quality, and there has to be
adequate voltage margins provided for it - which is not always
done.
Folded cascode has been most popular in very fast op amps,
typically video stuff, and is often found in conjunction with
current feedback (also known as low impedance feedback, where
the feedback network attaches to the Source/Emitter pins of the
input devices instead of the Gate/Base).
just one of several options we presented.
Properly done, folded cascode behaves just like a regular
cascode and sounds the same. The real difference is that
you can use a folded cascode simultaneously as a level shifter
to get a direct coupled output with a single-stage (or more
complex, if you like) amplifier.
The emphasis is on properly done; The CCS which supplies
both the input gain device (the one being cascoded) and the
cascode device has to be high quality, and there has to be
adequate voltage margins provided for it - which is not always
done.
Folded cascode has been most popular in very fast op amps,
typically video stuff, and is often found in conjunction with
current feedback (also known as low impedance feedback, where
the feedback network attaches to the Source/Emitter pins of the
input devices instead of the Gate/Base).
I'm certainly no expert on circuit topologies but NP's comments
confirmed my dimmed with time observation (which I wasn't
now sure about) that a folded cascode is effectively a re-
arrangement of an LTP pair.
And in a circuit simulator this with give nearly perfect cancellation
of all even order harmonics, leaving odd order dominating.
As I understand it an LTP arrangement reduces even order distortion,
but not to the extent typically suggested by a simulator.
🙂 sreten.
confirmed my dimmed with time observation (which I wasn't
now sure about) that a folded cascode is effectively a re-
arrangement of an LTP pair.
And in a circuit simulator this with give nearly perfect cancellation
of all even order harmonics, leaving odd order dominating.
As I understand it an LTP arrangement reduces even order distortion,
but not to the extent typically suggested by a simulator.
🙂 sreten.
Cascodes (straight, folded, or kinky) don't necessarily have to be part of differentials. The
Differentials, however, do tend to suppress even harmonics.
Grey
Differentials, however, do tend to suppress even harmonics.
Grey
I did a simulation of a symmetrical complementary common source design using current source loading and symmetrical folded cascode output. The harmonic signaure from the simulation was pretty disturbing, with prominent odd order components out to the 7th harmonic, and the evens suppressed by 10X. Maybe the symmetry here is also cancelling the even component. However, if I substitute another common source stage for the output instead of the folded cascode transistors, so that the design now looks like a symmetrical compound pair, the distortion signature shows a smooth tapering off of harmonics as a function of frequency. If anyone is interested, I can post the schematics and an output file for each, so that you can see the difference.
Obviously, the next step is to build up both circuits to see if the difference manifests itself in real life. I also have a folded cascode power amplifier that I also want to evaluate when it comes out of surgery ( higher bus voltage, higher voltage drivers, bigger output FETS). I'm really posing this question to see if anyone has noticed a different sonic character for the folded cascode topology.
Obviously, the next step is to build up both circuits to see if the difference manifests itself in real life. I also have a folded cascode power amplifier that I also want to evaluate when it comes out of surgery ( higher bus voltage, higher voltage drivers, bigger output FETS). I'm really posing this question to see if anyone has noticed a different sonic character for the folded cascode topology.
I took a closer look at the two circuits I mentioned in the previous post, and gave the folded cascode stage a little more voltage drop, so that the bases of the cascode transistors were about 5V away from the rails. This gave the current source load transistors a little more voltage to work with. The nasty harmonic character went away, and the folded cascode stage is now about 0.05% distortion at 10kHz, or about ten times that of the symmetrical compound pair at the same frequency and drive level. This is to be expected, as the folded cascode circuit won't have as much open loop gain. I still will be doing both circuits for test, to see if PSpice really can give a reasonable prediction of circuit behavior. I also want to see if I can tell any diffrence between the two in listening tests. Thanks all, for giving me a tip to straignten things out. When I get some results, I'll post in a separate thread.
FYI, here's an older thread, dealing with the sound of cascode:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=3398&highlight=
http://www.diyaudio.com/forums/showthread.php?s=&threadid=3398&highlight=
wrenchone said:
This would be expected, as matched pairs operated out of phase
with each other (as with a diff pair) suppresses even order
harmonics. If you load them with a common source stage, the
output of that stage is likely to contain a high percentage of
even harmonics. A lot of this is dependent on the specifics of
the circuit and parts themselves.
The cascode transistor imposes very little of its own character
on the signal, but the common source device does by
comparison.
Just for reference, the schematics below show the circuits I was simulating. The one on the left is a symmetric compound pair, while the circuit on the right is a symmetric folded cascode using current source loading. This is all part of an examination of simple current feedback gain blocks for use in an RIAA preamp. The first is a simple version of a circuit used long ago by Erno Boberly as a MC preamp. The second has probably been done by someone else too, though I haven't heard about it. The simple FET cascodes used for the circuit on the left are to get a little voltage margin on the second stageFETs because of the 25V breakdown voltage of the 2SK74. If someone could suggest some complementary Japanese devices that would serve better for cascode FETs than the J175 and PN4393, I'd be happy to hear about it. The source follower in the circuit on the left also needs FETs with higher breakdown voltage than the J110s. I'll probably end up using PN4393s, as I have scads of them.
In practice, I'll be selecting source resistors based on the matching results for the 2SK170BL and 2SJ74BL I have in my lab stock. This will all happen when I find out where I packed my FET matching jig (just moved).
I'll let folks know the outcome when I'm finished playing...
In practice, I'll be selecting source resistors based on the matching results for the 2SK170BL and 2SJ74BL I have in my lab stock. This will all happen when I find out where I packed my FET matching jig (just moved).
I'll let folks know the outcome when I'm finished playing...
Attachments
Re-arrangement of an LTP
With respect to a re-arranged LTP, were you thinking of an emitter-follower/common-base cascade? That would result when driving one half of the LTP (only) and taking the output from the opposite side.
You'd avoid the Miller capacitance-multiplying effect in a similar way to a cascode, too.
Erik
sreten said:
With respect to a re-arranged LTP, were you thinking of an emitter-follower/common-base cascade? That would result when driving one half of the LTP (only) and taking the output from the opposite side.
You'd avoid the Miller capacitance-multiplying effect in a similar way to a cascode, too.
Erik
This isn't directly on subject, but what do people think of the vacuum tube LTP differential crosscoupled cascode amplifier?
Please don't hijack my thread. It is so easy to start a new thread in the vacuum state section - you might even get some replies...
That is a very good thread, but unfortunately it stops without any answer to the original question. The guru's are debating about the definition of cascodes instead. I wonder what will be the conclusion of that thread if it was continued to finish.
After reading that old thread, it shakes my believe again. First, I learn that distortion comes from fluctuating voltage and fluctuating current.
The fluctuating current is remedied by putting big standing current (bias current), so the percentage of signal is small compared to bias current (class A).
The fluctuating voltage is remedied by cascodes, that is to lock the operating voltage of the gain transistor.
I tought that is it. But that old thread discuss about how cascodes gives "FATIGUING" sound for long time listening. Actually I still have questions, why Mr. Nelson Pass seldom use cascode, inspite he has his own patent on cascodes? He only uses it for small FETS or dividing thermal dissipation.
Is it true, while cascode can help with measurement figures, it is bad for listening? What is the reason behind it?
What is a high quality CCS looks like? How much is adequate voltage margin?
wrenchone said:
probably because positive feedback in linear amplifiers fell out of fashion as pentodes became popular - the positive feedback (known as "regeneration") was useful to cancel the internal negative plate resistance feedback of triodes when larger external gains were desired
its going to be difficult to do a “apples to apples” comparison of folded cascode vs current mirrors given the number of inversions necessary to create a negative “current feedback” connection not being the same
you could try adding a common source output amp to circuit #2 but then you are also testing nested vs global feedback schemes as well
The source follower buffer on the first circuit did little if anything to the simulated distortion figure. I would only use it when the circuit has to interface with the cold, cruel world of uncertain impedances. The current mirrors in the second circuit are only there for the bias, to make sure that the two current source loads are reasonably equal, and as benificial fallout, also to provide bias for the cascode transistors. The same thing could be done in countless other ways, but bipolar transistors are cheap...
So sad, I took a closer look at the circuit on the right, jcx, and it does have positive feedback. For some reason, it didn't flop over to one side in simulation like it should. Shame on me for not staring at it harder...It could be fixed by adding another inverting stage or by returning the feedback to the input, but I'm not too interested in either approach. It looks like my folded cascode measurements will be confined for the time being to my power amplifier, once I get all the mods done and coax it back into its case. Now that one has proper negative feedback...
- Status
- Not open for further replies.