| MBK |
There is a lot of activity on the X-amp, but the (to me) very interesting X-preamp gets little attention.
Besides the main circuit which has been covered to some extent, without to my knowledge any complete working circuit though, I am especially interested in the volume control. Nelson Pass gave quite a few hints in various threads: very high precision and low distortion, with bipolar switches, patent pending... He even suggested people back-engineer the volume control.
Given the patent issue we should be thankful and that must be hint enough... yet my expertise is a bit underpowered here. My understanding is too limited to reinvent the wheel from scratch, especially without at least a basic topology to work on.
So, question to the experts - did any of you do some work on the x-preamp volume control? |
|
|
| zygibajt |
I,m interested,too.
Also (this is question to Nelson Pass),is there any chance that we see service manual for X2 preamp in near future?
PassLabs website says it's no longer in production.
I'm looking forward to see any deep conversation on the topc of X-preamps.
Bartek |
|
|
| wuffwaff |
IŽll use a stepped atenuator (elma) placed at the output with around 4k7 -5k for my X-BOSOZ. (ladder type)
ThisŽll give maximum ca. 1100 ohms (2200 balanced) output impedance at -6dB. In the volume-region where I normally listen to, output impedance will be below 400 ohms.
william |
|
|
| MBK |
Sure, a stepped attenuator will work, as will a quality pot or a relay ettenuator... however ...
I was looking for truly elegant engineering though. I see the Pass philosophy as such. Pass uses thoughtful topology to get the best out of the devices, e.g. by keeping devices at best linearity - example, Aleph current source. For remaining errors Pass uses error cancellation between similar devices, in the X-approach. Many topology can achieve this in theory but added complexity multiplies the potential problems exponentially - so Pass keeps complexity low.
Note that ordinary push pull would could qualify here in philosophy if only NPN and PNP devices behaved truly similarly ... note also that the class A power consumption looks "unelegant" to me, though the Aleph AC current gain alleviates the problem, again, elegantly.
The beauty comes from building topologies where errors either matter the least possible, or get cancelled, with ordinary devices, by self adjusting, inherently linear, inherently stable circuits that work with constants that don't need a trim.
Now I expect a balanced volume control to follow the same lines of thought. I know nothing about the X-Preamp circuit (and little about electronics), but I think I get the ideas behind it even though I can't engineer one from scratch. So I speculate about the guiding principles.
Ideally a volume control should vary the gain, rather than getting gain and then squashing it. Linearity and stability issues apparently prevent this.
In addition, component variation makes balanced pairs and channel matching very difficult.
I wanted to use non-mechanical switches to build a better volume control. I dislike mechanical switches. I believe much distortion comes from bad contacts, so,the less the better. Pots probably worst. Relay parameters also got me thinking ("minimum signal"... meaning, below minimum we have a problem???).
Then I read Nelson Pass hinted that he uses bipolar switches in a shunt approach.. Now, the X-approach uses symmetric error cancellation. So, in this line of thought, I would imagine an X-volume control to somehow use a cross-feed between transistors which control the volume shunt on both halves of the balanced pair, in order to improve matching. Maybe even channel matching can be achieved this way. And nonlinearities of the transistors get x-ed out.
So, this kind of approach - which I can "see" conceptually but have trouble calculating - gets me going... :nod: |
|
|
| JasonL |
What are your thoughts.. what do you want in the pre'amp.
Display
controls
Volume
digital ?
balanced
balance left right
input selects?
any of these things or just a imple on off switch with a balance Control and a volume |
|
|
| Petter |
Bipolar device switched attenuators have one major drawback: They cannot switch across phases but have to shunt to a fixed potential (typically reads ground).
I agree that attenuation is suboptimal - and that it would be even better to do gain variation.
With the X model, it is possible to vary the gain by changing the "X" resistor. The higher the value, the lower the gain.
The problem is that this is phase to phase shunting and thus requires something like relays.
The good part is that if you head on over to trading post, you will find Dale and Craig's relay board (also available at www.dipchipelec.com). If you guys are willing to buy, they might just be willing to modify the volume sequence to match - then again, you could do the inthinkable (which everybody else does) and use their SHM unit the way it is designed - as a balanced phase to phase shunter in the normal position (either before the gain stage, or after it).
Petter |
|
|
| JasonL |
| the boards from dipchip are expensive.. TOO much for some one starting off. there are lots around that have not much money to play with and well there prices are way out of my range.. |
|
|
| Petter |
I agree that the dipchip boards are costly but take issue with the word "expensive". :) State of the art equipment at the prices offered is not normal. Look inside expensive equipment and see what they use - usually cheap stuff. Just one more point: Dipchip offers complete solutions, but you don't have to buy everything.
So what I suggest is that you start off with a simple potentiometer in the range of $5-30 and move to the next level when you are ready :)
Good luck - you can do a lot with cheap solutions. I know the dipchip guys are working up a semiconductor version which should come in at a much lower price. They also have a headless version in development.
Their BEST 2 channel setup costs $270 including display, optical encoders, remote controllable with standard remotes, incredible software capability such as acceleration when you move encoder fast etc. and very clever volume settings which likely are way ahead of the original Pass units which at the very least are likely to have lesser sound due to shunting to ground and using semiconductors instead of relays. The main cost drivers are: Relays, display and optical encoders. I'll tell you this much: I bought it. You might be able to get by with one of their lower end offering and save $50-60 - but hey I have to plug what I did part of the design for and use myself. :)
At the very least, go over there, and look at their work for motivation. BTW, I don't have monetary interests in this.
Petter |
|
|
| MBK |
| quote: | Originally posted by Petter
Bipolar device switched attenuators have one major drawback: They cannot switch across phases but have to shunt to a fixed potential (typically reads ground).
(snip)
With the X model, it is possible to vary the gain by changing the "X" resistor. The higher the value, the lower the gain.
(snip)
The problem is that this is phase to phase shunting and thus requires something like relays.
Petter |
NP pointed out, and I found the same statement from various other sources, that this includes *virtual* ground. Info that I found typically uses FET's in that position.
So - how does the Pass production X-preamp volume control actually work? Does the X-topology not have a virtual ground suitable for strategic between phase shunt operation? |
|
|
| Pete Fleming |
| I'm interested in a good way of being able to incorporate remote volume (and possibly source selection) into the preamp. Many of the digital based solutions introduce an unacceptable level of noise/distortion in my opinion. |
|
|
| Petter |
MBK said:
"So - how does the Pass production X-preamp volume control actually work? Does the X-topology not have a virtual ground suitable for strategic between phase shunt operation?"
Shunt to ground with bipolar devices (one leg per phase) would be my guess.
I doubt very much that virtual ground shunting is viable. How "groundy" is the virtual ground at the source of the input FET's`????
To my knowledge nobody uses X resistor variation (or source degeneration) for volume control but I have seen it used for range setting.
Petter |
|
|
| MBK |
Using FET's as switches at the virtual ground of inverting op-amps seems like a standard implementation from the informations I saw around (I saw it mentioned in Horowitz for instance).
As for the phase shunt - I could imagine a transistor between two equal resistors. Hence the transistor would always remain "virtually" at quasi ground between the two opposite phase halves...
Shunt to ground looks again "inelegant" to me :mafioso: |
|
|
| MBK |
| I mean of course a simgle transistor in the middle of two equal R's which connect the 2 inverted phases. In addition to saving a transistor, this solution means any residual distortion affects both halves... |
|
|
| Petter |
MBK,
Bear in mind that most transistors are unidirectional. For bidirectional transistors such as JFET's you might be able to do it. If the perturbations are small enough, you could use a BJT but bear in mind that the current you inject into the base region will have to go somewhere and that you are unlikely to get perfect symmetry.
I looked into this extensively a year ago, and decided eventually that options such as JFET, FET, optoFET, electronic relays (bidirectional FET's), BJT's etc were not optimal. I backed this up with measurements.
Hence I suggest you consider using relays.
If you are confident that the perturbations are very small at the X resistor, you could try a JFET, BJT or FET I suppose bearing in mind all of these devices are not set up for transmitting significant current in the "wrong" direction and that the FET has integral body diode.
You can try this out with mechanical switches or just soldering in the required resistance for "benchmark" performance.
Petter |
|
|
| MBK |
Petter,
yes, the BJT base current issue troubles me too. Maybe I just assume Pass Labs found a good way of doing it that beats relays (in the real world - if contacts were perfect relays sure would beat transistors every time). After all they have used relays before, so why now this mysterious patent pending BJT shunt volume control if it doesn't improve on existing solutions...? |
|
|
| Petter |
non-mechanical devices have some very nice characteristics: they switch fast and don't break so easily.
BJT switching to ground works pretty well - and I know how Pass does it but will refrain from divulging info that I figured out (and additional that came through private conversation with Pass Labs).
Switching between phases is hard with semiconductors.
For me (and based on my own research), mechanical contacts is the way to go. Relays ended up as the most practical solution once the decision to go with mechanical contacts was made.
Petter |
|
|
| MBK |
Thanks Petter, at least that gives me an idea about the trade offs in the original Pass approach.
Maybe I try a simple op amp circuit with JFETS at virtual ground for starters. I'll have to rely on my ears though for measurement. But I darn hope it will beat that **** pot I now use in my home brew op amp pre... Tracking, linearity, it's hopeless
:goodbad: ... even though that pot also sits optimally at wiper to virtual ground - and I think before I buy some $200 DACT there's gotta be something more elegant and advanced. |
|
|
| Petter |
Dipchip offers a more elegant and advanced solution than the DACT.
What I recommend you do is simply solder in the resistor values you require to verify that your scheme works. I am extremely suspicious of trusting the voltage level at the sources of the transistors.
I once thought JFET's would work since they are more or less symmetrical, but in my measurement jig they turned in terrible performance (which was verified as asymmetry on a scope). I was kind of unhappy about that because I really thought that would provide the best results. Perhaps I could have done better in a balanced setup ...
Petter |
|
|
| Petter |
I thought about it some more and am quite confident that BJT will not work well in the X position given the relatively high current you are passing (in that position it is a current attenuator rather than a voltage attenuator and the current levels will be much more demanding than in a voltage position - depending on bias of course).
It is possible that a JFET might do it but bear in mind my experience with JFET's to date.
You should also be able to get NFET in source to source configuration to work.
No harm in trying though.
Petter |
|
|
| MBK |
Thanks Petter, I looked up data sheet of a JFET I have lots of lying around (J309) - low Ciss of 12 pF, RDS(on) of ca. 35 Ohms, *should have* (haha) benign behaviour if kept at VGS < +-0.2V and IDS <3 mA.
Now with signal "source" R in that configuration of 10k and for typical line level signals of aroud max. +4 dBU we keep IDS < 3 mA, for a VDS of < 0.1 V in that combo at worst. Could work. Plus the schematic I saw for this topology in Horowitz and Hill had a second JFET, that one in the feedback loop with G shunted to S - so remaining nonlinearities *might* be reduced as well.
I will try that one, first on an inverting op amp virtual ground. |
|
|
| Hans B |
Hi all
I thought I would try to start this thread up again.
I am designing an X-preamp and was hoping for some help on this volume control (maybe a hint :nod: ).
First of all the circuit:
The basic concept of the design is taken from the passlabs D1 service manual, except I use JFET 2sk389 as the input differential and the fact that I have X-ed the circuit.
Actually come to think of it - it looks like a mix of an XBOSOZ and the D1 circuit.
Specs are (according to spice -I havent build it yet, need the volume control still):
Vin max balanced : 7V peak
Vout max balanced : 28V peak
gain (balanced) : 4 (12dB)
Input impedance : around 20k balanced
output impedance : around 300 balanced
Frequency range at full scale (7v peak) : 0.03Hz-500kHz
THD @1kHz @ full scale 28V balanced out: 0.18%
THD @1kHz @ 4V balanced out: 0.0023%
Minimum load impedance (dependent on IL2/IR2) : 800 Ohm
I have tried to use shunt of bipolar switches (blue square area) and it seems to work quite well. As you can see a resistor (they are not really pots) is coupled to the collector of a bipolar transistor, and is turned on by voltage applied to the base. For simplicity I have only drawn 1 of these bipolar switches, but the idea is to have a number for bjt's each connected to their own resistor, to form the full volume control.
When using this method it is not possible to get 0V out with a direct shunt, because of the internal Rce, but with the used transistors I can get 800uV balanced out @7V balanced in, which is low enough - if you want lower, increase the input resistance. These BJTs might not be the same in the final design, if they are very expensive - I just picked some with low Rce.
Sofar so good, but initially I wanted to place the volume control after the amplification, that takes place in the differential pair. Either at the output or at the MOSFET gates.
If placed at the output, I would need a resistor in series with the signal (after the cap) and then shunt to ground, but this changes the output impedance quite a bit.
If placed at the MOSFET gates they should be shunt to a "virtual ground", which in this case should be 16V, but I have no idea how to do so.
My questions are:
Is there a way to place the bipolar shunt before the MOSFETs and in this case - how?
Is shunting between phases a good idea (Petters mentioned the problems with using BJTs to shunt phase)??
Am I (at all) in the right direction when thinking of "Waynes balanced passive masterpiece" - A hint perhaps Mr. Pass or Mr. Colburn??
Any thoughts, comments would be appriciated.
/Hans
PS. unfortunately I can't make orcad put out a decent picture - you will have to settle for this zipped version - sorry. |
|
|
| UrSv |
| And what does that shunt attenuator say when the input signal is negative? Seems like swithing something with an NPN where the collector is at a lower potential than the emitter would be hard. |
|
|
| Hans B |
Hi UrSv
That is the reason why the BJT transistor is placed after the resistor (with emitter close to ground or virtual ground) and not before. When connected as on the diagram, the voltage over the BJT transistor will always be very small (in the range uV). If the BJT was placed before the resistor - the situation you are referring to occurs.
At least - thats how I view it
Hans |
|
|
| MBK |
| hmmmm... I still wonder if that is the famous Wayne circuit... I mean is this implementation a concept that's patentable? The shunt to ground thing alone can't be it, that's not novel in itself... |
|
|
| Hans B |
Hi again
I think I've got progress!!!
I have found as section in the aleph L owners manual about a patent pending volume control. Could it be possible that passlabs has 2 volume control patents pending. I would assume that they only have 1.
As I read it, the volume control is a passive resistive network, when the gain of the preamp is less or equal to 0dB. When the preamp has to amplify (from 0dB to max gain) a switch determines that the signal goes through the active preamp circuitry. Simple (in principle) and very very smart (as always).
Think about it - for preamp with a 70dB volume control range of 1dB steps and 6dB of gain, the active circuitry are only used in 6 of the 70 steps - the rest (64 steps) is pure passive resistive network. And I don't know about you guys, but I usually don't have my volume control above the 3'O clock position (refer to Aleph L manual, where 3'O clock is 0dB gain - I'm not saying that all preamps has 0dB at 3'O clock!).
Could it be that the great Pass and his crew have been smiling this whole time, when Wayne Colburns famous attenuator has been brought up, knowing that some of this information already existed in some form on the passlabs web page???
The unbalanced version of the volume control is shown on page 10 in the aleph L manual. But how does bipolar shunt to ground come into play in this circuit???? If you choose to shunt to ground, your going to mess with the input/output impedance, and I can't see how it is possible to achieve smalle output resistance and high input resistance in this case. So I am a bit lost on this one - any ideas??
Do you think that this could be the patented solution - having passive attenuation below 0dB and active above 0dB????
Anyway - gotta get some sleep, I will look at it more tomorrow.
/Hans |
|
|
| Petter |
I looked into this topic extensively about a year ago. I reverse-engineered said volume control and did measurements on it based on BJT, JFET, MOSFET and mechanical Relays.
When push comes to shove, the result was that even cheap relays significantly outperformed the only real other alternative (BJT).
The other advantage of relays are that they do not have to be referenced to ground.
The downside is power consumption, switching speed (potential for glitches), size, weight and longevity.
Petter |
|
|
| macka |
Hans,
I like your design,
Have you gone any further at this stage and listened to the circuit?
Ian |
|
|
| Petter |
| quote: | Originally posted by Hans B
Hi all
I thought I would try to start this thread up again.
I am designing an X-preamp and was hoping for some help on this volume control (maybe a hint :nod: ).
[snip]
My questions are:
(1) Is there a way to place the bipolar shunt before the MOSFETs and in this case - how?
(2) Is shunting between phases a good idea (Petters mentioned the problems with using BJTs to shunt phase)??
(3) Am I (at all) in the right direction when thinking of "Waynes balanced passive masterpiece" - A hint perhaps Mr. Pass or Mr. Colburn??
|
1. Yes you can do this. It will be similar to a potentiometer on the input.
2. I think phase shunting is a good idea for a number of reasons :). The main issue is you have to be near ground level (or an alternative rail). There is also yet another alternative - shunt (or "deshunt") between sources of input transistors. When the resistance there is infinite you get zero gain. It is also even possible to change the feedback resistors. In principle the "de-shunting" seems promising but I have only seen it used (by the master) as a gain range setting. Not sure how well it will work in X mode though, depends on how much open loop gain you have and whether you are kicking into that before the feedback resistors take effect.
(3) The last item requires connection to ground and cannot be used phase to phase easily.
| quote: | Originally posted by UrSv
And what does that shunt attenuator say when the input signal is negative? Seems like swithing something with an NPN where the collector is at a lower potential than the emitter would be hard. |
This is what baffles most of us, but has been replied to before. discrete BJT's provide the "best" performance after JFET's (terrible asymmetry) and MOSFET's (don't even think about it). There are of course quite expensive options which may be very good (such as semiconductor relays for the telecommunication industry manufactured by companies such as International Rectifier and Vishay.
Petter |
|
|
| Petter |
You can also shunt between R4R and R4L is using relays, mechanical swithces or potentiometer (or through capacitor if going for ground).
Also, there is not all that much cost (albeit some work and complexity) to add to make this a real X.
Petter |
|
|
| Hans B |
Macka:
I have not yet made the design and can't comment on sound quality. I have been waiting for the wayne's PRBLM, and I posted
here to see if I was on the right track with the shunt to ground method. Right now I don't know what I am going to do. I am down to 3 choices:
1: wait for the PRBLM description.
2: use X-pro's: Constant impedance relay-resistor logarithmic attenuator (balanced)
3: use part of the APOX circuit (relay shunt between phases)
Petter:
Thank you for your inputs. I have 3 comments.
1)
You mention that the resistor between the input JFETs (R1) could be used as a volume control, and I thought of this, but the resistor have to be really big before you hit 0V (infinity). The simulation shows that a 100k resistor gives you 0.4V at the output, which is too much. Of cause you can increase to 1M and decrease the gain further, but I don't like the idea of using very big resistors. Also, as I understand it, the resistor between the JFETs control the "amount" of X in the circuit (the distortion is cancelled through this resistor), and if you increase R1 you decrease the X-effect (although I could be wrong).
2)
I mentioned that I would like to shunt before the MOSFETs to ensure a constant input constant output impedance, however shunting before the MOSFETs doesn't seem to be a good solution in this circuit. I don't have enough current running through the differential to shunt between R4L and R4R, so I kinda scratched that idea now. I would now prefer to do it at the output
3)
you say:
Also, there is not all that much cost (albeit some work and complexity) to add to make this a real X.
Does this mean that you don't consider this an X circuit??
If so I please explain why - As I mentioned it is pretty much a cross between the XBOSOZ and the D1 circuit, and it should be an X circuit. The principle is basically the same as Henriks XBOSOZ.
http://www.diyaudio.com/forums/show...=&pagenumber=17
Thanks again for your comments
/Hans |
|
|
| Petter |
| quote: | Originally posted by Hans B
Macka:
1)
You mention that the resistor between the input JFETs (R1) could be used as a volume control, and I thought of this, but the resistor have to be really big before you hit 0V (infinity). The simulation shows that a 100k resistor gives you 0.4V at the output, which is too much. Of cause you can increase to 1M and decrease the gain further, but I don't like the idea of using very big resistors. Also, as I understand it, the resistor between the JFETs control the "amount" of X in the circuit (the distortion is cancelled through this resistor), and if you increase R1 you decrease the X-effect (although I could be wrong).
/Hans |
It is true that you need an infinite resistance to get to zero volume this way. However, the "X'ing" resistor is basically equivalent to 2 source degeneration resistors in an ordinary long tailed pair. The larger these source degeneration resistors, the more linear (before feedback) and lower gain do you get. My personal opinion (no flames, please) is that too much is made of the "X'ing" resistor. Of course as your gain is lowered, the effect of the other feedback is reduced.
I suspect some of the benefit of the X unit is that it likely has high open loop gain with significant reduction through the feedback network. The way it is set up now, you will have low gain before feedback.
| quote: | Originally posted by Hans B
Macka:
2)
I mentioned that I would like to shunt before the MOSFETs to ensure a constant input constant output impedance, however shunting before the MOSFETs doesn't seem to be a good solution in this circuit. I don't have enough current running through the differential to shunt between R4L and R4R, so I kinda scratched that idea now. I would now prefer to do it at the output
/Hans |
Yeah, it is painful to have a single gainstage - you have to do the volume either before or after the stage, not between. I have a fix for that too, but it is non-X.
| quote: | Originally posted by Hans B
Macka:
3)
you say:
Also, there is not all that much cost (albeit some work and complexity) to add to make this a real X.
Does this mean that you don't consider this an X circuit??
If so I please explain why - As I mentioned it is pretty much a cross between the XBOSOZ and the D1 circuit, and it should be an X circuit. The principle is basically the same as Henriks XBOSOZ.
http://www.diyaudio.com/forums/show...=&pagenumber=17
/Hans |
I missed IL1 and IL2 due to their non-standard placement :). However, I don't like it all that much because M1 and M2 are not acting as cascodes - they need to be p-channel to do that. If they are p-channel, however, you need to turn them upside down and fix the source potential problem which means it does not draw from top rail, but from R4L,R. Then again, the output capacitors ... Check out the X100 thread, first posting .... Remember the top cascode does not get driven at it's gate ...
Now if you put the volume control at the output, you can still add class A emitter followers (if BJT, emitter to emitter with emitter degeneration) at the output.
Hope I am not discouraging you with what is not intended to be a negative posting.
Petter |
|
|
| macka |
Thanks Hans,
I have asked about Wayne's remote pre amp also.
I am unaware of the delivery on that project
Ian |
|
|
| Hans B |
Hi Petter
I am glad somebody is responding :), and you have a lot of good points so I appriciate your response - If nobody replied I would consider this a sign, that I was on the wrong path.
I like your comments, and they are all true, and you have got me to look at the circuitry differently now. I should have stated that I was not trying to make a replica of the patent, but a JAXP (just another X preamp), and that the design wasn't a two stage with a folded cascode approach, but a more simple JFET input differential followed by a "source follower" mosfet, and that the design principle was low open loop gain and modest feedback.
What I think you are suggesting is to go for the folded cascode (like the patent and your x100 backengineered - great work btw) and have more open loop gain. This might be a good idea, I have to look at it some more, but you got me interested in going that way, so I will pursue it more in the following days.
Now this X-thing is really pusseling me.
For instance - the open loop gain:
I thought the idea of X circuits was that it cancelled the distortion through the differential, and thereby removed the idea of using a lot of feedback to remove distortion. And Nelson Pass states in the XSOZ thread that:
"Just about any amplifier with an input diff pair can be turned into an X, but it only seems to work well with simple circuits and low open loop gain. (too bad, huh?)"
And you say that: "I suspect some of the benefit of the X unit is that it likely has high open loop gain with significant reduction through the feedback network".
So my question is, why you think that more feedback is good for the X'ing - As I see it more feedback will lead to less distortion in the "traditional" feedback way, not by X. Anyway it might not be that much of a problem when using a folded cascode or my attemp, since the open loop gain is relatively small.
Another thing is the differentail pair. You mention that the change in the source resistor in the differential will lead to less distortion but also lower gain, and since I think I read somewhere that the distortion in the differential does not get cancelled by Xing, then you have a very valid point. In the next version (will be a folded cascode) I will ensure that the differential in itself has low distortion, and that the common gate FETs are in charge of most of the amplification. Anyway I might be getting ahead of myself here - I have to look at it more before stating such things.
So thanks for the comments :nod:
/Hans |
|
|
| Petter |
Thanks for your detailed reply. I don't agree you had source follower at stage 2 - that would not work well into a current source IMO.
| quote: | Originally posted by Hans B
Hi Petter
And you say that: "I suspect some of the benefit of the X unit is that it likely has high open loop gain with significant reduction through the feedback network".
So my question is, why you think that more feedback is good for the X'ing - As I see it more feedback will lead to less distortion in the "traditional" feedback way, not by X. Anyway it might not be that much of a problem when using a folded cascode or my attemp, since the open loop gain is relatively small.
/Hans |
If you look at the X100 thread, the input gain devices are IRF610's which have a pretty huge current gain. Now that those are loaded against the outer bottom current sources, you have very high gain.
The gain is attenuated by two things: First the "X'ing" resistor. This is more or less equivalent to having a single central bottom current source with source degeneration resistors on the input FET's.
The second item that sets up gain is the combination of input resistors and feedback resistors. If the fundamental gain is high enough, the total gain is exactly Rfb/Rin. (Rin includes output impedance of prior stage if you are getting picky)
So, why is this setup supposed to sound so well? Well, the input transistors are shielded and operate more or less optimally. The output is more or less at the same level as the input, and there is only one gain stage around which the feedback is placed.
Feedback is good. Feedback around several gain stages is normally shunned in audio circuits. Feeback compensates for the fact that there are cascode devices through which the signal has to pass + other adverse effects.
| quote: | Originally posted by Hans B
Hi Petter
Another thing is the differentail pair. You mention that the change in the source resistor in the differential will lead to less distortion but also lower gain, and since I think I read somewhere that the distortion in the differential does not get cancelled by Xing, then you have a very valid point.
/Hans |
There may well be an Xing effect that the horizontal resistor has over ordinary degeneration. Also, when the Xing resistor is zero ....
Also, if you want a low gain unit, select JFET's which I seem to recall you had already done, and you are pretty much there.
Let me know how you get on. |
|
|
| Hans B |
Hi
Petter:
I still don't understand why you don't consider the output MOSFETs of my first attempt a source follower (common drain). To me, the output is just a single ended class A follower driven by the voltage that appears from the differential. Erno Borbely uses followers with JFETs in one of his articles. If I am fundamentally wrong about this I would appriciate an explanation as to why this is :scratch:
To the new design:
The design is now based more on your X100. It is now an all JFET amplifier, and uses the folded cascode like the X-circuit and a follower output (follower is taken from Erno Borbelys article 2 about JFETs), and I included the follower in the feedback loop. This made me wonder why you did not include the output stage in the feedback loop in your x100 - is there any particular reason?
Also does anyone know if the real passlabs X-preamps uses folded cascode??
(well, I know of at least 1 who does :cool: )
Any comments would be appriciated!
/Hans |
|
|
| Petter |
| quote: | Originally posted by Hans B
I still don't understand why you don't consider the output MOSFETs of my first attempt a source follower (common drain). To me, the output is just a single ended class A follower driven by the voltage that appears from the differential. Erno Borbely uses followers with JFETs in one of his articles. If I am fundamentally wrong about this I would appriciate an explanation as to why this is :scratch:
/Hans |
They are source followers. What I don't like is the ac coupled output + driving the source of the follower through a current source.
The topology is a single gain stage with source follower with feedback around the source follower + a capacitor to level shift. I am not sure how the current source will work in this application.
One of the main reasons for using followers is to lower output impedance. The capacitor in series with the output is not something I am particularly fond of.
| quote: | Originally posted by Hans B
This made me wonder why you did not include the output stage in the feedback loop in your x100 - is there any particular reason?
/Hans |
No reason not to, apart from the fact that NP and others suggest it is a bad idea. In a power amp in particular, the load is typically complex which can pass (no pun intended) various types of signal back. NP uses very low impedance output stages to control the load regardless of what it is. In a less expensive output stage, people do the same with feedback. Then they also often use an output inductor to reduce the load's effect on the input stage through the feedback.
Try both, pick what fits you best. You will have less distortion if you put the feedback around the output stage.
| quote: | Originally posted by Hans B
Hi
Also does anyone know if the real passlabs X-preamps uses folded cascode??
(well, I know of at least 1 who does :cool: )
/Hans |
Many of Pass' recent designs that have been in the public domain use folded cascode. All X series units by definition need to use it (with the possible exception of the XA range). If you think about it, how else can you make a fully DC coupled single gain stage where the input DC level is the same as the output DC level?
I also assume you want comments on current design which I assume to be X oriented. If that is the case consider:
1. RL1 is not necessary per se but the value is so high it will likely have little impact.
2. Interesting output circutiry albeit unconventional. I would have put a resistor between J3A and IL3 to set up a direct representation of the same signal at different DC levels. The value of this resistance and IL3 would set the current. You might even want a capacitor across it. Then, you could drive the output buffer directly and control the current at the output stage by setting up source resistors (and varying the first resistor in the normal way). You could also use a Vbe multiplier in this position if you like but you are not likely to need negative tempco in a preamp.
3. The beauty of this design means that you don't need to have (nor want to have?) output capacitors CL1.
4. Note your gain will only be 2 per phase with said input/feedback resistors. You may also check if you really need two current sources at the bottom, but hey - this is totally getting somewhere.
Petter |
|
|
| Nelson Pass |
I'm in the midst of writing up the Wayne PRBLM project.
Does anyone remember what that acronym stood for? |
|
|
| Nelson Pass |
| quote: | Originally posted by Petter
Many of Pass' recent designs that have been in the public domain use folded cascode. All X series units by definition need to use it (with the possible exception of the XA range). If you think about it, how else can you make a fully DC coupled single gain stage where the input DC level is the same as the output DC level? |
Actually not true. Folded cascode is not essential, and appears in
in two commercial Pass Designs, the X600 and X1000 (which have
exactly the same circuit board. The lesser X designs are seen to
use ordinary cascoding on the front end.
In the XA designs we have no cascoding, and that would be consistent with the Aleph circuitry. That does not mean it
might not eventually make an appearance somewhere, folded
or otherwise. |
|
|
| moe29 |
PRBLM: Wayne's Passive Remote Balanced Line Masterpiece.
(and that's a quote) :nod: from the DIY progress report thread |
|
|
| Henrik |
Hans B
You are raising som very good questions, all though the volumecontrol has left the forground.
Just one question, your latest creation uses output buffer. Your diffrential output impedance without buffer and DC coupled is around 3K (and if you raise the feedback and short the resistor between the sources then you would get as low as 600 Ohm diffrential outputimpedance), so IMO you dont need the buffer with regard to the output impedance.
Any special reason why you want this buffer?
Have you heard / expierienced the difference in sound quality between folded cascode with buffer and just one good quality cap at the output directly at the drains of the diffpair?
I have been looking into your approach to the volumecontrol, which is valid some how and very interesting.
Regards |
|
|
| Henrik |
| quote: | Originally posted by Hans B
Now this X-thing is really pusseling me.
For instance - the open loop gain:
I thought the idea of X circuits was that it cancelled the distortion through the differential, and thereby removed the idea of using a lot of feedback to remove distortion. And Nelson Pass states in the XSOZ thread that:
"Just about any amplifier with an input diff pair can be turned into an X, but it only seems to work well with simple circuits and low open loop gain. (too bad, huh?)"
/Hans |
As I recall, Nelsons statement about openloop gain were about complex circuits with many gainstages with much higher open loop gain than a single diffrential gainstage like the one in question here. As an example, the diffpair in the Aleph-X is coupled without resistance between the sources.
In the patent Nelson did applie low (how low is that ?) resistance between the sources of the diffpair, but only to regulate the "hall of mirrors" effect when patent is used in more complex circiuts. |
|
|
| Hans B |
Hi
Petter:
First of, I really appriciate your inputs.
To your comments:
1) you mean RL2, right? - and you are right, it is not really necessary - I just gives me a reference to ground when no input is connected.
2) I need to look at this some more to really comment, but it looks promising :)
3) You are right - I don't need the output cap, but I build my aleph-X (70W) without an input cap (because I was going to use a BOSOZ at the time). MY AX absolute DC drift starts at 4V and decreases over the next hour to +/-50mV, which is why I need the output cap.
4) Point taken - I only use 1 current source now.
Henrik:
I wanted to use a buffer at the output (originally I wanted a MOSFET), because I wanted this preamp to work if I decided to reduce the input impedance of my power amp. Also at the time I wanted to place the volume control at the output. If the volume control is placed at the output (using shunt to ground og shunt to other phase), then you would be driving relatively small output impedances.
But if you place the volume at the input, the output follower would probably not be necessary.
I wish I could tell you that I have listened extensively to differently topologies, but I haven't. It would be interesting to see if a folded cascode is better than your XBOSOZ. I actually build a BOSOZ a while back (before my AX'es were working), but decided not to use this preamp - reasons: first of the PCB was too small because I didn't print the circuit board correctly (acrobat - DO NOT check the box that says "shrink oversized.....". Secondly I could actually smell my amplifier. The resistors got so hot that they put out a smell, which scared me a little bit :) - count me out of the "fearless amplifier builders society"!!
Regarding the volume control, I think I am going to wait for the waynes masterpiece (unless I find myself with a lot of time on my hands)
/Hans |
|
|
| mlloyd1 |
drudging up a very old thread ...
was this one ever released? maybe i missed it ...
mlloyd1
| quote: | Originally posted by Nelson Pass
I'm in the midst of writing up the Wayne PRBLM project.
Does anyone remember what that acronym stood for? |
|
|
|
| Nelson Pass |
| quote: | Originally posted by mlloyd1
drudging up a very old thread ...
was this one ever released? maybe i missed it ... |
It was not, for reasons that no one can remember. Probably
lack of time. |
|
|
| jam |
Mr.Pass,
Now might be a good time ;) ...........well probably not for Wayne as he might be busy with all the new Pass offerings.
Jam |
|
|
|
