| dorkus |
| the borberly DC-coupled FET line amp we were talking about (the one the single-ended guys didn't like :p) is actually a tweaked-out version of the super buffer. the line amp includes source followers for additional output drive capability, and a servo for better DC stability. i suspect it sounds pretty good but i've never tried it, i'm probably going to try the full line amp version. check out the article on the borberly site on the all-FET line amp, from the 5/02 issue of AudioXPress, it mentions the super buffer. |
|
|
| rickpt |
| all cascode.....:cool: |
|
|
| HarryHaller |
Isn't that the same 12 year old circuit that HPotter posted in the previous thread on this topic?
H.H. |
|
|
| Peter Daniel |
| I guess so. That circuit is 10 years old and looks better to me.:) |
|
|
| Ren Hoek |
| Not yet ready.......:eek: |
|
|
| rickpt |
no
this is the all cascode, it has cascodes in the folower stage.
as for the older one posted on the other thread, i dont like the way borbely bias the front stage cascode. wath do you guys think about that?
regrads
rick |
|
|
| HarryHaller |
Stranger..... we don't discuss cascodes in front of the children and women folks 'round here. You take that kinda talk elsewhere......
H.H. |
|
|
| Ren Hoek |
| :confused: I thought the newest one didn't have cascodeed followers? |
|
|
| dorkus |
you are right ren, the latest complementary FET design from borbely does not cascode the output stage, only the diff inputs and 2nd gain stage. the output are traditional source followers using MOSFETs.
i am starting to wonder if the "super buffer," from which the latest design is derived, is worth a try. borbely mentions that the latest one (the EB-1198, we should really refer to these by model # from now on to avoid confusion) was designed with even greater drive capability in mind - he wanted to be able to drive 100 ohm headphones in Class A w/low distortion. this sort of overkill is nice in a preamp but not necessary in most systems. the super buffer is a little simpler than the 1198 (one less active stage) and also does w/o the DC servo that people seem to have a strong distaste for. i wonder why the addition of the output stage necessitates a servo in the 1198, or is borberly just going for extreme DC stability there? in any case i think i may prototype the super buffer instead of the 1198, and compare that to the SE-FET design that the others favor. with very tight device matching, i think the performance could be quite good. thoughts?
oh, and speaking of the SE-FET design... does anyone have any concrete topology changes from the original design? if i can not get suggestions specific enough to be drawn out in a schematic, then i will just go with the original layout.
cheers,
marc
p.s. it seems to me that it is very dangerous to make broad generalizations about topics such as cascodes. analog design and analysis is so dynamic and complex that the effectiveness of a given technique can hardly be isolated, rather its role must be evaluated only in the context of the circuit's implementation and performance as a whole. if you want to talk about cascodes, you should really specify where and how it will be used or it is of little value to us. therefore, comments like "cascodes sound bad" will be summarily ignored from now on. :D |
|
|
| Keithj |
Hello folkus...
I've looked at the thread and, to me, the idea of a preamp has certain ah, attractions. I know I'm coming in late and I know that what follows may be interpreted as an attempt to take over the discussion. It's not.
I'd like to haul the discussion back a little if I can and look at a basic need (for me) at least. That is, the components must be readily available through dealers that are accessible at a reasonable price.
This would seem to me to indicate op amps as many of the bits and bobs that are used in such as Borbely preamps are not readily available at reasonable prices (here in New Zealand at least).
I'd also like to propose an order of discussion which, to me, seems logical.
1. The architecture (number of gain stages, the amount of gain, whether or not a headphone amplifier, tape buffers, position of volume/balance controls and so on.)
2. The structure of inputs and outputs, number of, switching methods (relays, solid state etc).
3. How volume/balance will be done, actively, passively and by wht chips/pots etc.
4. Gain stages and buffers, design etc.
5. Power supplies (series, shunt, etc)
6. Control circuitry and programming, what microprocessor and who does the programming... includes displays
7. PC boards and layouts
8. Case and so on.
I've been involved in a system like this and the project never got off the ground because of a lack of structure to the discussions. Even within such, a loty of fun can be had. What do you think, and let's have some fun!
kindest regards, Keith |
|
|
| BrianGT |
Keith,
Take time to read the first post of this thread, and the previous long thread on this topic, and most, if not all of your questions will be answered.
--
Brian |
|
|
| dorkus |
has anyone built ANY of these circuits we've talked about? i think someone, jam maybe, said they've tried most of them but does anyone still have a working version they've listened to extensively and can comment on? apart from the circuit theory i'd like to have some listening impressions.
also, has anyone tried building the forssell design:
http://www.forsselltech.com/JFET%20Opamp.PDF
yes it's an opamp but hey who knows maybe it sounds really good. |
|
|
| AudioFreak |
| i'll be putting together a testbed in July and i'll do some listening tests on the various designs as have been suggested here and i'll try the various circuit variations such as SE with caps or SE with level-shift etc.......... |
|
|
| dorkus |
| i'll try to build a couple of them by mid-june... anyone know where i can get 2SJ74's? MCM has the 2SK170's but not the complementary J74's... man why doesn't toshiba make those awesome matched-pair single can JFETs anymore, bleh... |
|
|
| AudioFreak |
Dorkus,
Yeah i know July is a long way away ... sorry about that but i dont get holidays till then :D |
|
|
| rendisha |
Has anyone tried
Tip Top Electronics
They seem to carry a lot of japanese semis - Toshiba, Hitachi, Sanken... for very low prices. Download
Semis Original Catalogue
I don't know what grades of semis they have. Just few are specified as GR or BL. 2SJ74 is not specified. |
|
|
| dorkus |
i was just browsing through some old threads and came across an excellent post by AndreasJ on testing of different capacitor types. he was saying that polypropylene film types performed best, but even they had non-linearities at zero crossing. the best way to operate them is with DC bias, which correlates with listening observations people i know have made where all capacitors, regardless of type, seem to perform better with a stable DC bias.
maybe this is why AC coupling works so well in SE circuits - because the circuit by nature has a fixed DC offset, this in turn biases the cap, optimizing its performance to some degree. whereas if you stick a cap into a non-biased AC signal, you are more likely to get induced noise and distortion. that may be why i perceive even good film caps as colored, i am usually operating them around zero-crossing.
so in a SE circuit, i may be more amenable to capacitor coupling after all... just a thought.
you can see the original cap thread here:
http://www.diyaudio.com/forums/show...hp?threadid=525 |
|
|
| HarryHaller |
All the Borbely designs with negative feedback are "op amps", whats the distinction?
H.H. |
|
|
| sonnya |
There is no one. Only in the the type of design. Borbely shown in this thread are what i would call a CFB design where the forselltech design is an VFB design.
Sonny |
|
|
| jam |
Marc,
That is why I suggested hooking one end of the buffer to ground and letting one end of the coupling capacitor sit at about half the poistive rail. Most people miss the fact that coupling capacitors work best with some dc bias on them. Black Gates particularly, will compete with the best films.
Jam |
|
|
| HarryHaller |
| Another nice thing about the differential pair using same type of fets is that closely matched pairs such as the K389 and J109 are readily available. Also the K170 and J74 do not have the same transconductance. Gains never can be precisely matched for each half of the circuit. These is somewhat aleviated by source degeneration for the fets in the complementary diff pair. Adjustment of source resistors value with a pot for DC offset nulling can further complicate things. DC Offset trim can unbalance the gains for each half of the circuit. I also hate putting a pot in a location in the circuit where resistor quality is so important. |
|
|
| dorkus |
yeah, i'm not so sure about offset trimming either. pots kind of suck. i would only use them to find optimal values in a circuit, then replace it with proper fixed resistors.
i guess you're also pointing out a fundamental problem with all complementary circuits - lack of symmetry w/devices. i'm hoping with the K170/J74 pair they will be "close enough" but i can see why so many prefer the SE route.
anyway, i just ordered a bunch of JFETs from MCM... K170, J74, K246, J103, K389. the J103's are out of stock but hopefully they will be available soon. i got 100 K170's, gonna have to match them all probably, bleh... i will probably build the balanced SE line amp first, maybe the Forssell too if i have time. when the J103's arrive i'll try the complementary super-buffer. does anyone have any suggested modifications to the original borbely SE design?
i was going to ask another question but i can't remember what it was... |
|
|
| dorkus |
| oh yeah i'm going to try to run the SE line amp off +/- 24V instead of 36V, so i don't have to adjust the supplies to try the other circuits. i'm assumming i just need to adjust the top resistors (R3/R6), the diff pair source resistors (R4/R5), and the current source resistor (R9)? not sure how much i'll need to adjust them, guess i'll have to play around... or i can go hit my old textbooks and do the math i suppose. |
|
|
| HarryHaller |
Voltage equals current times resistance. Go read Nelson Pass's web site and Erno Borbely's for any more detail than that. This stuff is not rocket science but is a little more difficult than balancing one's check book. An evaluation version of Spice of the web might help also. All the jfet models have been posted I think.
J103's? Don't you want J109s?
H.H. |
|
|
| dorkus |
| quote: | | Voltage equals current times resistance |
yes, thank you harry, i was an EE. :p
i was not sure off-hand how the current would change as i changed voltage/resistance. i'm assumming a linear approximation is not good enough because of the characteristics of the active devices. i may try to get PSPICE or something set up so i can do some simple DC modelling.
| quote: | | J103's? Don't you want J109s? |
i dunno, but borbely shows J103 being the complement to K246... did i make a mistake? |
|
|
| mlloyd1 |
dorkus:
Have you thought about experimenting with different cascoding devices:
1. JFETs as you are currently planning
2. MOSFETs as Borbely (and HH ?) advocates
3. Some high performance (e.g. low collector-base capacitance, high Ft, high beta, etc. like the Sanyo 2SA1210/2SC2912) bipolars
I think some listening tests of the same circuit while changing the cascoding device may be interesting.
Too bad nobody has an Audio Precision handy - it would be nice to see how the numbers correlate with listening observations and preferences.
Ooopps; I said the bad word, "numbers":) I'll go hide now..... heh, heh, heh
mlloyd1 |
|
|
| dorkus |
mlloyd1,
nope, haven't thought about tweaking the cascoding yet, but i'm welcome to suggestions. i need to read up first on my cascodes (my analog teacher would start the lecture about how great they are but by the time he got to the details i was usually asleep :o ) , so i don't feel confident coming up with my own schemes yet but after i get the basic circuit up and running i'd definitely like to try modifications. i'll probably wait until i get a handle on the "sound" of the base circuit first though.
cheers,
marc |
|
|
| HarryHaller |
| I thought you might want complement for 2SK389. I personally would not waste time on K246 and J103 for cascodes. I guess even Erno is allowed to have a bad idea occasionally... |
|
|
| dorkus |
| quote: | Originally posted by HarryHaller
I thought you might want complement for 2SK389. I personally would not waste time on K246 and J103 for cascodes. I guess even Erno is allowed to have a bad idea occasionally... |
oh, damn. now you tell me. :D
so just use the same K170/J74 for the cascodes?
also, is there any matching requirement for the cascode device? |
|
|
| HarryHaller |
You could if you reference the gates for the cascode elements to allow at least 5 volts across the gain transistors drain to source. Connecting the gate of the K170 and K246 together only puts only a volt or so across the K170 which is not a good thing. A show of hands from anyone who knows why?
H.H. |
|
|
| dorkus |
| time to hit the books... *sigh* |
|
|
| MRehorst |
Why is a gain stage necessary?
Why discrete devices as opposed to an op-amp?
Seriously,
MR |
|
|
| HarryHaller |
Good question! Maybe we could use T.I. op amps instead. When the design is done we could probably put it on an ASIC. Nice ESL by the way... Can any anyone sugest a good T.I. op-amp to use?
Seriously,
H.H. |
|
|
| dorkus |
| now now harry... be nice... |
|
|
| Peter Daniel |
Dorkus,
I noticed that you tend to favour Broberly complementary designs. I just came across that application which might be of some use to you. As you see the parts count doubles but no caps in the signal path.:) |
|
|
| HarryHaller |
What.....? You guys are reading stuff into a perfectly innocent post. You explain it then.......
H.H. |
|
|
| dorkus |
| haha. right harry. ;) |
|
|
| dorkus |
hp,
yeah, i was thinking of something like that. actually, my latest idea is to have an SE line stage of some sort as an input buffer w/a little bit of gain (maybe 6dB)... this would also convert unbalanced to balanced. this gain stage would feed a Wolfson stereo attenuator chip, except i would use the two channels for the balanced signal. the output of this chip would feed complementary borbely line stages (two of them for balanced operation), also with 6dB or gain or so for a total of 12dB gain. i could also use the SE line stage here i guess but i think the wolfson requires an inverting amp after it (not sure why). i would need AC coupling for the SE input stage, but the output stage could be direct-coupled.
i know some people are going to complain about using the Wolfson IC or 2 gain stages, but i don't really know of a better way to do it practically right now. i would really like the attenuation range and resolution that something like the Wolfson provides. this solution also provides full balanced in/out with a minimum of fuss.
anyway, the borbely SE line stage is first up on the prototyping list, so we'll see how it sounds.
marc |
|
|
| dorkus |
| hmm so i was about to order all the parts i need to wire up the Borbely balanced SE design, then i realized... the schematic on his website (the EB-2000/402) and the one in the 6/99 issue of Audio Electronics (the JFET Part II article) are slightly different... he mentions in the web version that they have "upgraded the basic design new constant current source and complementary JFET output stage..." this new source follower is instead of the version of the White cathode follower shown in the article. should i assume the web version is the "latest and greatest," and build that? i guess if Mr. Borbely calls it an "upgrade" then it must be better... ? |
|
|
| MRehorst |
| quote: | Originally posted by dorkus
hp,
i know some people are going to complain about using the Wolfson IC or 2 gain stages, but i don't really know of a better way to do it practically right now. i would really like the attenuation range and resolution that something like the Wolfson provides. this solution also provides full balanced in/out with a minimum of fuss.
marc |
If I may suggest, Burr-Brown (now owned by my employer, TI), makes a stereo attenuator chip (PGA2310) with performance very similar to the WM8816, only with a much better pinout- digital stuff is on one side of the chip, analog on the other side, allowing the part to straddle analog and digital ground planes making board layout MUCH easier and making it easier to keep digital crud out of the analog circuits. Also, the op-amps are internal (but I guess that limits tweakability). The PGA2310 can be sampled from the TI web site, and can be purchased in onesy-twosy quantities from Digikey.
For balanced I/O, Burr-Brown also makes the INA2134/7 and DRV 134 balanced line receiver and balanced line driver. Also readily available at Digikey.
Back to my original questions- I can see that op-amps aren't so bad after all... I still would like to know why any gain is needed in a "preamp".
MR |
|
|
| HarryHaller |
| Couldn't you at least plug the good BB opamps? |
|
|
| HarryHaller |
Some people still use turntables. Some of us experiment with low output DACs and low gain power amps. What's an op-amp guy got against a little gain anyway?
H.H. |
|
|
| jam |
Yeah! And what about amplifiers that have low gain eg. the Zen.
Jam
P.S. TURKEY SPAM ! Some form of hidden implication Ha. Ha. |
|
|
| HarryHaller |
| Anyone else get thier (spam censored) samples? |
|
|
| dorkus |
MR,
i've ruled out the PGA2310 because it has an internal opamp, and a mediocre one at that (i think it is based on OPA604 or something). the wolfson is a passive part so it gives us more flexibility with choosing our own active circuitry.
harry was being sarcastic. most of us think that IC opamps suck. you should be careful how you use the word "opamp" though. i think you confusing it with IC opamp. an operational amplifier is simply a circuit that amplifies the difference between its two inputs, thus performing an "operation" in a mathmatical sense. you can make an opamp with discrete transistors or you can use an IC. many people think opamps sound bad, possibly due to the tremendous amount of negative feedback they require to limit gain. i tend to agree, though i am willing to try discrete versions. i am also looking at good ICs (none made by your company i'm afraid) but primarily for comparison purposes with the discrete circuits. while ICs can still sound pretty good i don't expect them to hold up to the discretes. in any case, this and your other questions are probably better left for a different thread.
no offense, but i think you would best not suggest IC parts (other than attenuators like the Wolfson) in this thread at the moment... :p most of us refuse to touch them so it's probably a waste of time, and we're busy tweaking some other discrete circuits anyway. unless you know of some new miracle IC that sounds magnificent...
cheers,
marc
p.s. yes, we need gain. my sony SACD player has a pretty low output level and i want the flexibility of a wide volume range.
p.p.s. harry: haha, yeah really... :D |
|
|
| dorkus |
| quote: | Originally posted by HarryHaller
What's an op-amp guy got against a little gain anyway?
|
yeah seriously! you op amp geeks (i used to be one of them, sort of :( ), with all your talk about the pie-in-the-sky opamp with infinite open-loop gain, infinite-bandwidth, zero output impedance, blah blah blah... :p |
|
|
| jam |
I would like to know if any of op-amp fans have built a discreet version of their favourite op-amp and compared the two under the same conditions ie. gain, supply voltage ....
I have and gave up on op amps. years ago, granted they got a lot better over the years but........
Also, all you have a black box with fairfy little control of what happens inside. A cop out in my book.
Jam
P.S. Most of you guys are missing what Harry is trying to say. Maybe more hand holding is in order.:D |
|
|
| dorkus |
| quote: | Originally posted by jam
P.S. Most of you guys are missing what Harry is trying to say. Maybe more hand holding is in order.:D |
what IS he trying to say? you seem to be his soulmate, his partner in crime, his butt buddy, whatever you want to call it... so you need to translate for us. :p
p.s. when you say you gave up on opamps, do you mean ICs specifically, or the whole topology altogether? |
|
|
| dorkus |
| i just realized my posts seem to be getting more abrasive and sarcastic... i think i might be turning into another harry or jocko, nooooooooooooooooo... |
|
|
| jam |
I do not by any stretch of the imagination claim to be Harry's best bud. Be he has offered everyone a tremendous amount of information to steer people in the right direction but most people miss it.
I am speaking fom experience because I have gone down that path before and I have to agree with most of his conclusions. He bullds stuff and has the technical background to back it up rather than pontificate (is this a good word) from some spec sheet or text book. He has a hint of sacarsm ( well a lot) in some of his replies but I can understand why....... Try beating your head against the wall a few times.
Most of you will agree that Nelson Pass builds some of best audio gear and I can't seem to find any op-amps in his products. Anyone wonder why?
Jam
P.S. I need a new picture of Harry, I have a new set of horns for him. |
|
|
| dorkus |
just kidding jam. was throwing in a little sarcasm of my own. hope you were not offended. :p
so, i guess you are basically not a negative feedback fan. i don't blame you. i'm looking forward to building the Borbely SE circuit, i have not tried anything SE in my system yet so it will be an interesting experiment. btw should i build the one w/complementary output followers or the "White cathode follower" version? i guess i should try both but i'm just so lazy... so i will build the complementary version to start cuz it has fewer parts :p |
|
|
| jam |
No, I never said that I hated negative feedback! Applied correctly it is very useful especially in power amps.
When you build the Borbely let me know, I might have a few ideas for you to try.
Jam
Me offeneded naw!, By the way your mum wears army boots! :p :D |
|
|
| HarryHaller |
Jam is the guy that put horns, a tail, and a pitchfork on my posted picture. With friends like that....... I am perfectly content to be the man everyone loves to hate. It's a dirty job but someone has to do it. Mr. Pass gets all the hero worship and deserves most of it.
H.H. |
|
|
| HarryHaller |
| I think they would go great with this one but make sure Grey isn't looking........ |
|
|
| jam |
I can't see any details, picture is too dark. How about posting another one. Wear a mask if you have to.
Jam |
|
|
| dorkus |
| all the other folks are gonna lose interest if we just post pictures of harry all day long. no offense harry. |
|
|
| HarryHaller |
Yep....... Like I would dare show my face on this forum. There are people from Texas on this forum. My life wouldn't be worth a plugged nickle. Next thing someone will want my phone number....
H.H.(Happily Hidden) |
|
|
| dorkus |
i vaguely remember a while back, there was a product called a "cable charger" which DC-biased interconnects on one end, and AC-coupled them at the other with a high-quality film cap. supposedly this was supposed to optimize the signal transmission to some degree, though it was never explained to me how it did so. the editor of my audio magazine said despite the added coloration of the capacitor, he felt it often made a worthwhile improvement in sound quality.
i was just thinking about this in connection with my "Son of Dork" preamp... just random thoughts of the Borbely SE circuit, high-pass filtering my main speakers (i have a subwoofer), connecting the preamp to my amp, and dielectric performance. it occured to me that all cables are capacitors to some degree, and most any capacitor seems to perform better when DC biased so as to avoid signal zero-crossing. maybe this "cable charging" stabilizes the dielectric and avoids any nasties that might be associated with zero crossing? there was also some claimed supression of RF interference, though i have no idea how that would work.
so i have an idea for Son of Dork... why not combine the best of every world? i could use the Borbely SE circuit, which has an output offset somewhere around +12V or +15V depending on supply. then, i would have a special DC-coupled output on the preamp that has no coupling cap - just the straight SE output, DC bias and all. i would run this signal through the interconnect to the amplifier, then have in-line RCA/XLR adapters at the amplifier end with blocking caps in them. i could even select the value of the cap to deliberately high-pass filter the signal where desired, say at 60Hz for my main speakers.
using such a scheme, i could use a SE preamp circuit that everyone seems to favor, get the benefits of cable DC biasing, optionally high-pass filter my mains, and still have only one capacitor total in the signal path. the only catch is that i'll have to be VERY careful to label the DC-coupled output of the amp as such, and only use this output in conjunction with the blocking caps at the other end - otherwise i could fry my speakers in an instant. i might consider even using a different type of terminal/cable (BNC or something) to prevent such an accident.
thoughts?
dorkus |
|
|
| Petter |
Please find attached a simple preamp circuit. It is missing trim and detail engineering. I like it for the following reasons:
1. "No" feedback
2. As much or little gain as you want
3. No cascodes
4. No level shifters
5. Room for attenuator between stages which DOES NOT have to go to ground (runs in balanced mode).
6. Fully balanced: converts balanced to single ended and vice versa
7. DC coupled
8. For single ended operation, switching phase is very easy, just choose the other output.
9. This is "basically single ended application" of a balanced circuit. I say this since single ended gain is developed over a single resistor with "single" gain device. Borbely appears quite fond of the other approach, (push-pull) input with complementary JFET's with their gates connected.
Let me just explain briefly what is going on in the output stage. I chose drain output so that voltages can be shifted back to zero without added complexity. The gain of the ouput stage is just over 1, so there is significant degeneration. It can be construed that the output impedance will be very low even given the drain output. Without degeneration, balanced mode output impedance is 616 Ohms as shown. With feedback, it is probably in the region of 200 or so Ohms. Of course an SE output has half that output impedance.
I would argue this is just about as simple as you can make a circuit given a requirement for 2 amplification stages with passive attenuation between them.
What do you think?
Petter |
|
|
| JonMarsh |
Interesting, Petter.
Have you simulated this? With which FETs?
How much do you feel you gain from the second stage being differential also as opposed to a more conventional common source amplifier tied to the rails? (besides higher PSRR). Performance could be quite good without going complementary, but DC balance if you're using it single ended would be a problem. Were you considering output capacitors?
The gain structure is very similar to an FET DAC output buffer I designed, but my circuit is more (too much?) complicated compared with yours, it is complementary, and I used somewhat higher power supply rails and Zetex bipolars for the outputs. It is a NLFB circuit also, and uses some other measures to linearize the first diff amp, which can do single ended to balanced and vice versa; it's designed to work from differential DAC output chip.
I like your concept- hopefully you'll keep us informed of how it develops.
Regards,
Jon |
|
|
| Petter |
I have not simulated this, but I have built something very similar in my Pearl. I would say that the second stage in my pearl (SE version of the above) with 4 JFETs has fairly similar operating conditions as the above -- except it runs with cascode due to the much higher gain.
Balanced operation of output stage costs very little in terms of components and alows a volume control which does not involve ground. I like the symmetry, and it is easy to guarantee absolute output voltage level by trimming the current source. Current sources in diff amps really are very neat for this purpose, as they handle biasing automatically, as well as output DC level :)
No, I did not consider output capacitors. I use the best ones available on the market: No such things. It is true that one could go with n polarity devices all around, but as you state, that means more complicated power supply and/or strategically placed coupling capacitors.
I originally had higher PSU voltage for the bottom side, but decided to reduce it given that I was expecting flames due to higher output impedance this causes. Of course, one could build this like I build my Pearl v2 which is balanced with 8 JFET's per half per stage. Then we could double the low side voltage while keeping the output impedance the same.
Suggested devices: n: 2SK170, p: 2SJ109 or something like that.
Petter |
|
|
| dorkus |
| any comments on sound quality? |
|
|
| Petter |
Well my Pearl which is basically the same sounds very good. It is not really directly comparable, though.
A pro friend of mine says that he feels that symmetric Borbely type push-pull input has a less "coarse distortion structure" whatever that means.
I would think this type of circuit, especially with the high level of degeneration will be extremely linear even without loop feedback. Incidentally loop feedback is easily implented, but you have to "cross" so that you take output on right to input on left and vice versa given the even number of gain stages.
Noise should also be low since all gain is takein in first stage, and since devices are very low noise + paralled (good enough for an MC RIAA stage).
The circuit works with any gain device. Some may prefer BJT's, tubes or a mix thereof.
Petter |
|
|
| gromanswe |
As beeing an amplifier designer since many years.
I must say your entuiasm is fantastic.
I am no fan of Nelson Pass designs, neither can I understand
the "use (MOS)FETs constructions at any price"-spirit.
In audio low noise bipolar is mostly
the most suitable alternative.
And in the output stage Bipolar have no match!
I have seen at your suggestions. They might work well,
but aren't they a little unnessesary complicated?
Why make things harder than it need to be?
Although no fan of Pass Amps, one of Nelson's
guidelines is "KEEP IT SIMPLE".
There I agree totally.
Often this is the most cost effective way also.
Less money spent.
Short signal path.
Fewer components for signal to pass through,
and thereby be colorated.
Few and good components is the way to go.
This is the best HIFI preamp design I have ever seen.
Give it an eye.
Simple is not the same as bad.
In this case quite the opposite.
http://www.mitedu.freeserve.co.uk/C...Audio/gmpre.htm
Groman from Sweden, Amplifier expert
gromanswe homepage:
http://www.geocities.com/gromanswe/:) |
|
|
| dorkus |
just wanted to let everyone know that SoD is still alive and well. it took me longer than expected to get the parts i needed together (and i still wound up going to Radio Shack for a few things) but i am ready to start prototyping. here are some goodies i've assembled for the first phase of the project... nothing too fancy but some nice Nichicon Muse/Panasonic FC power supply caps, Black Gate coupling caps, Panasonic P polypropylene film bypasses, some leftover Solens, Noble pot, Holco/Yageo/RS carbon film resistors (just for prototyping, Caddocks can come later), the necessary MOSFETS and JFETS, etc. etc. you can also see the power supply circuit board i etched last night, i will post details later on how making that went (i tried regular glossy paper this time instead of "special" toner transfer paper i used in the past).
first up... the Borberly JFET buffer (based on White cathode follower), to be followed by the SE JFET line amp. after i finish the power supply of course (the simple pass regulator design used in BoSoZ). |
|
|
| dorkus |
i just finished my Borbely JFET buffer tonight, for a background of the circuit check out this thread:
http://www.diyaudio.com/forums/show...p?threadid=2281
my circuit is the same as shown at the bottom, with Harry's last corrections. i've just plugged it into my system and the sound is very promising... can't say it sounds exactly like a straight-wire bypass but it is impressively clean, with great dynamics. i used really cheapo-quality parts too so the circuit is performing nowhere near its potential. more details and pics to follow soon... |
|
|
| dorkus |
here's a picture of the (messy) circuit i wired together. as i mentioned earlier, parts quality is pretty ordinary so i am not extracting the "ultimate" potential of this circuit, but just trying to get a feel for its capabilities. the resistors are cheap Yageo (Digikey) and a couple Radio Shack carbon films, except for attenuator series resistor which is a 15k Holco. a Noble 25k pot is used for attenuator shunt. the feedback cap is a .15uF Solen metalized polypropylene, the closest thing i had to the .22uF specified by Borbely. local supply bypassing is a 1000uF Pana FC and .01uF Pana P film.
i did not even bother matching devices (too lazy, and i haven't put together my matching setup yet) so there is 30mV of offset in both channels. the output of the circuit is therefore coupled with a 10uF Black Gate nonpolar, followed by a 100k shunt resistor. |
|
|
| dorkus |
the sound of the circuit is quite good... in many respects, very transparent. tonally it is just very slightly bright, but not in a sort of "white light" irritating way, just a slight brightening of the overal tonal shading. i've compared with my passive preamp (just a simple shunt attenuator), and the only drawbacks i've heard thus far are the slightly forward, less relaxed presentation, very slightly less resolution at the very lowest levels, and reduced soundstage depth, which could probably be improved by using a dual mono circuit instead of the sloppy stereo one i put together. interestingly, even though i find it somewhat forward and bright, it still sounds very even and neutral in presentation, and the dynamics are fantastic - definitely an improvement over the purely passive design. what i like best though is the ease with which leading edges and impulses come across, giving the sound a crisper, more defined character... probably because it is relieving the upstream source (cd/sacd player) of load. the buffer's dynamic characteristics seem much better than that of opamps i've heard, transients come across extremely crisply and convincingly, no overhang or ringing. triangles and cymbals sound very clear, maybe just a tad hashed compared to passive pre but crisper. bass is also excellent, so far i have not noticed any aberration despite the capacitor coupling, and the lower midrange down is slightly fuller and more powerful. i have the slight sensation that micro details are somewhat obscured compared to the passive (after all no active circuitry is the best), but at the same time i don't feel i'm losing much detail. instrumental timbres are a strong point as well, oboe timbre is dead-on. oh yeah, and it is absolutely dead quiet - not a trace of hiss or hum.
if it were a little smoother-sounding and did not reduce soundstage depth, it would be near perfect. i think many of these characteristics could be improved with better parts. the Solen coupling cap i used tends to be a little bright and flat i think, so using, say, a good Hovland there would improve matters greatly methinks. nice resistors (Caddocks, Vishays?) would improve matters further. in short, this little circuit sounds great, and has much potential with better layout and parts.
oh, i forgot to mention, you can find the power supply i used here:
http://www.diyaudio.com/forums/show...27&pagenumber=4
here's a picture of the box the circuit is in, just a plastic Radio Shack project enclosure. |
|
|
| Peter Daniel |
| Good info. What is the sound of Black Gates? |
|
|
| dorkus |
i'm not sure how to describe the "sound" of the BG caps, as i haven't done a formal comparson w/a straight bypass, but i have used them in a couple circuits and here is what i've found: compared to other electrolytics, they are far more transparent, but they do not sound like films either. i think the sound is robust, and definitely on the fuller side compared to films. in general i think 'lytics as coupling caps tend to be less thin and bright compared to (most) films, but their problem is usually muddiness and lack of definition. the resolution of BG's is excellent, but they give up perhaps just a bit of HF resolution to the films for a smoother, warmer sound; they sound slightly "massaged" compared to films, but i would hesitate to call them blatantly colored... all caps are, but the BG's keep up with good films overall. i think they are very worthy coupling caps, and i am not even operating them with a proper DC bias.
in the JFET buffer circuit, i currently have a .15uF Solen for the feedback cap, and it is biased 20V or so. perhaps i will experiment with Hovland film and BG caps here to see how they compare. |
|
|
| carlmart |
What capacitors are you talking about?
Input and output caps? They are not in the circuit then.
You mention a 150n Solen, but the one on the diagram is 220n. Did I miss something?
Is there a way to use this circuit and get away without input/output caps? It seems there is not, except if you add a servo.
Carlos |
|
|
| HarryHaller |
HA HA HA HA HA HA HA HA HA HA HA HA HA HA HA HA HA HA HA HA
HA HA HA HA HA HA HA HA HA HA HA HA HA HA HA HA HA HA HA HA ! |
|
|
| Peter Daniel |
| Black Gates are actually sitting on output jacks, because Dorkus didn't match FETs and got 30mV of DC offset. When you match, you should be OK with output and no caps are needed. Solens are in a feedback. I don't think it matters if it's 150 or 220nF. BTW it might be not possible to obtain such small value from BG. |
|
|
| dorkus |
ok harry, i gotta admit i am laughing w/you on that one... :p
the value of the feedback cap is not critical, i'm not sure what kind of impedance it's seeing but i'm guessing it's pretty high and even with .15uF the LF cutoff is probably pretty darn low. Black Gates actually come as low as 0.1uF, but i would probalby use larger values and put them in series to get more voltage out of them, according to my sim it is at least 40V across that capacitor but i haven't verified in-circuit yet...
yeah Peter (what happened to HPotter?), you can definitely get offset negligible with proper matching, however there is still the risk of some drift. i know we hate servos but i was wondering if i could build one into the -24V rail, to vary the negative rail slightly to keep DC balance. i could keep the simple pass-transistor regulator i'm using but instead of feeding the gate of the MOSFET w/voltage references i would use an opamp or something to vary it according to offset. does that sound like a really bad idea? |
|
|
| HarryHaller |
| Combine the following JFET circuits. One of the source resistor values can be adjusted to achieve zero volt offset. |
|
|
| HarryHaller |
Add source resistor to the to upper Jfet in following......
The cap value is not critical and can be smaller with 1 Meg resistor value increased to a couple of Meg for lower freguency cutoff. I would put 221 ohm gate damping resistors on both Jfets. |
|
|
| dorkus |
harry, i tried adjusting the source resistor on the top JFET by putting another resistor in parallel, and it did help lower offset but not by a lot, and i was a little afraid of deviating too much from the 22 ohm nominal value... but i just realized it doesn't really matter what the exact value of that resistor is as long as it's not too high, right? i will play around with the values to see if i can get the offset nice and low.
i think i would rather increase the feedback capacitance than increase the 1M ohm resistor at the bottom to keep the LF cutoff suitably low, in general i don't like to make any impedances too high (more noise?).
oh yeah, i am using a 1k resistor on the gates... too high? 220 ohms seems to be a popular value, but i remember you talking about how you can "voice" the sound of the circuit by changing gate dampening. i will have to check it out...
p.s. what's a good amount of bias current for this circuit? borbely specifies BL-grade in his schematic but i have GR. i guess i will have to adjust the top drain resistor. |
|
|
| HarryHaller |
The bias current is a function of the source resistors. I think you could go down to 10 or 15 ohms for green bias group Jfets. 1K is probably fine for gate resistors. The drain resistor controls open loop gain and making it larger will increase the amount of negative feedback and making it smaller will decrease the amount of negative feedback. Closed loop gain is 1. I would keep the DC voltage across the Jfets between 10 and 20 volts for low noise and good bandwidth. The circuit really doesn't need 24 volt supplies and should wourk fine with 15 to 20V supplies. The negative supply could be 5 volts less than the positive if desired since the bottom JFet doesn't have the voltage drop from a drain resistor.
H.H.
A commitee of one. (Hey it makes as much sense as an army of one.....) |
|
|
| carlmart |
What happened to also monolithic duals 2SK240 and 2SJ75? Were they discontinued? Weren't they cheaper?
Are there any USA sources for dual MOSFETs? The only I know is Audio Lab of Georgia, but they have very small stocks.
Carlos |
|
|
| AudioFreak |
| quote: | Originally posted by carlmart
What happened to also monolithic duals 2SK240 and 2SJ75? Were they discontinued? Weren't they cheaper?
Are there any USA sources for dual MOSFETs? The only I know is Audio Lab of Georgia, but they have very small stocks.
Carlos |
the k240/j74 are no longer in production ... also they were not monolithic but rather 2 individual matched fets mounted in a single aluminium casing. |
|
|
| HarryHaller |
I don't think this will work since the dual fets have a common substrate electrical connection. You really should be able to get very low offset by adjusting the source resistors with the K170s.
H.H. |
|
|
| dorkus |
| oh whoops, right. source resistors, not drain, determine the bias. i have not measured anything yet but i will check soon and post my results. i still need to check the possible oscillation in my regulator too, but alas my girlfriend is visiting this week so i'm not sure when i'll get to it... |
|
|
| dorkus |
| this design-by-commitee thing is not so bad. you call the shots, i try stuff out and arbitrate the discussion. not exactly a commitee but whatever. :p |
|
|
| dorkus |
| harry was right... for green bias rated SK170's, 22 ohm source resistors are too big. i measured measly bias currents of 3.5mA in one channel, 2.7mA in the other. i am playing with some the values now to try to boost that up a lot, i figure i will want at least 10mA or so, and in the process i will try nulling the output offset as well. the voltage across the JFETs is kinda high right now too, on average around 21V. i'll keep tweaking til i get a good balance of values. |
|
|
| dorkus |
this may sound dumb, but... don't work on your circuit while it's plugged in. :(
yes yes i know, i violated safety rule #1 of electronics work, but i was fooling around with some source resistor values while the circuit was powered so i could immediately meaure the results... little did i know a stray strand of solder drifted towards my power supply, not only shorting out the negative rail (and taking a bit of the PCB trace with it), but also sending a nice 45V spike to some part of my circuit. it burned out the center trace of my Noble potentiometer and at least one JFET gate... oops. i managed to salvage the pot by soldering a litz wire from the center pin directly to the wiper, and i'm replacing the suspect JFET now... still, not a pleasant experience. |
|
|
| dorkus |
ok so i changed source resistor values to null the output offset and increase bias current.. i wound up replacing the resistor on the top JFET with a short, then decreasing the bottom one to around 10 ohms... not sure what the exact value is, i put a small trimmer pot in parallel w/the original 22 ohm resistor so i could vary it. bias current is now around 4.5mA.
the sound is in some ways better, but not quite right. it is smoother and less forward, i guess some of the harshness previously was from not enough bias current. however there is a tradeoff in that it sounds a bit muted and less dynamic now... maybe a little overbiased? the devices run a little warm now but not hot.
the larger problem i'm having is DC stability. i had to change my input shunt pot from the 25k Noble i busted to a 10k alps. however, when i change the volume, i think the DC offset is going all over the place. i measured it with no load connected, before the output coupling cap, and it was pretty stable, under 1mV. then i connected my power amp, and it shot up to 80mV, and changed as i varied the input shunt attenuator. the weird thing is, i am still using the Black Gates on the output, so i don't see how hooking up my amp would affect DC stability. i also did not notice this problem before i started fiddling with the source resistors. any ideas? |
|
|
| mirlo |
1) Are you looking at it with a 'scope or with a meter? COuld it be ultrasonic oscillation, whose average value depends on the connection.
2) Is the capacitor still OK, or could it have been damaged as well? |
|
|
| mhennessy |
| quote: | | this may sound dumb, but... don't work on your circuit while it's plugged in. |
Ouch! Sound nasty :(
At the risk of stating the obvious, I always do my prototyping with a bench PSU with an accurate current-limit. If it's something beyond the voltage limit of my supplies, I knock up simple 2-transistor current sources on Veroboard. Set them to just over the expected current demand. I used to have a + and - pair mounted on a small heatsink (for when they were limiting - they sometimes got hot!)
Obviously, you'd remove them before critical listening tests :) |
|
|
| HarryHaller |
Have you measured the DC offset of your input source?
Small measured DC offsets can be measured as an artifact of high frequency oscillations. I would not run a jfet in saturation in a circuit except as constant current source. A small source resistor
insures the the Jfet is not in saturation. It also provides local feedback which should make changes in DC offset with temperature smaller. If you are running long or fairly capacitive cables a small series resistance of 50 to 200 ohms may be in order as the circuit could be going into oscillations.
There is little more involved in the design of a two transistor circuit than one might think.........
H.H. |
|
|
| dorkus |
thanks for the suggestions guys...
i was measuring w/multimeter, i will check on my scope next. this DC offset is with no source component connected, just the load. i will check if my output coupling caps are still ok. the cables to my amp are only 1m and i don't think the capacitance is very high, my amp is 100k input impedance. still, i may have some oscillation here, i will have to look on my scope. you are right harry, there is a lot to discrete design... the dynamics of even a simple circuit like this can make things pretty harry... er, i mean hairy.
btw, i am still getting that 9mhz oscillation on my scope, WITH NOTHING CONNECTED. i.e. i shorted out my probe, connected ground to the tip, and i still have the interference. i have a PC nearby but no other electronics in the immediate area. i wonder if my 'scope is kaput, i got it on ebay for $250. maybe i will take a picture of the noise and put it up... |
|
|
| dorkus |
you guys were right on the money, check this out.
850kHz oscillation, 2.5V (!!!) peak-to-peak. ugh. :(
time to stick the source resistors back in i guess. any other tips for getting stability out of this circuit? maybe decrease the open-loop gain by adjusting the top resistor? hmm, is there somewhere i could stick a compensation cap? |
|
|
| mirlo |
Shooting from the hip,
I think that 850 kHz is too low a frequency to be related to not having the gate resistors in, but you may want to add them anyway.
One place for a comp cap is across the 500 Ohm drain resistor. Maybe you could try values in the range of 1 nF.
Like I said, I'm shooting from the hip. I'll try to think about it more later.
What do you suppose the cable capacitance actually is? 50 to 100 pF?
What about the AC input impedance of the amp? Often the amp will have a high DC input impedance in parallel with a low pass filter that might be a 1 kOhm resistor +several nF capacitance.
The compensation idea I mentioned might do more harm than good; a 1 nF would add a 320 kHz pole to the system. Maybe the cap has to be much bigger, but if it is much lower then you get into phase issues in the audio band. It depends on where the problem is arising. So consider it well before proceeding. Maybe you will have to put a R on the order of tens of Ohms in series with the C.
Just ideas.
I don't remember, do you have a simulator to play with?
-- mirlo |
|
|
|
