| Giaime |
Hello everybody!
I just finished designing a 211 A2 SE amp, just for the curiosity. I'm not planning to build it: the objective was taking more experience and familiarity with mosfet drive and A2.
What do you think of the design, is there anything obviously wrong, or something that could be done better? The circuit is explained in the article linked.
Also I'm not too sure about the setting of the quiescent current in the mosfet, as you can read in the article I made a supposition about how it works.
I've uploaded it in my site:
http://www.giaime.altervista.org/hivolt.html
Feel free to make any comments!
:att'n: Please note that this is not a project suited to beginners due to high voltage :att'n: |
|
|
| Eli Duttman |
Giaime,
I LIKE FETs serving as "spear carriers" for tubes; so, you'll get no argument from me. Perhaps MOSFET Follies contains information of use to you.
A "stiff" B+ rail seems indicated. Think about a hybrid bridge rectifier using stacked sets of UF4007s and 2X 6AU4GTA damper diodes feeding a choke I/P filter. Given the SLOW B+ rise, you could forego the use of a bleeder resistor across the 1st filter cap. 211s heat up and conduct FAST.
Edit: fixed typos |
|
|
| kevinkr |
Just wondering what dictated the choice of an ECC83 mu follower? I haven't found this topology to sound particularly good with this tube. Looks great on paper, in practice lousy bandwidth and very slow sounding.
Also the gain will very closely approach mu, making this amplifier excessively sensitive unless some global feedback is employed. (No, I'm not advocating this..) Off the cuff about 350mVrms would drive this thing to full output.
If it were me I might go for the 6SL7, probably in SRPP, full output power would be achieved at a little over 1Vrms. |
|
|
| Giaime |
Hello Kevin!
In fact, gain is pretty high, around 85, and it needs 1.2Vpk to drive the 211 to full output. Frequency response is artificially limited due to 6.8k grid resistors, to have a -3dB point of around 150kHz.
And obviously yes, I would use global NFB, as explained in the article, for a factor of 6dB.
That's not hard to implement a 6SL7 in that stage, neither any other high gain tube, maybe we can use the full 6dB feedback only using ECC83.
Edit: further investigations show that only with ECC83 you could use feedback: otherwise, you need a very thought preamp that puts out hefty swings ;)
I'll redraw the amp schematics adding nfb. |
|
|
| Tubes4e4 |
Hi Giaime,
IRF840 has a horrendous input capacitance of 1300pF, so you might check if the peak current swing of your tiny current trickler ECC83 mu follower really is able to drive it to the required bandwidth. I doubt it.
You might consider IRF820 instead, which has only 1/3 of input capacitance (about 360pF).
Tom |
|
|
| Giaime |
Hi Tom,
good point, and thank you very much, I have to dig into datasheets. But since it is a source follower, the input capacitance that datasheet says is still applicable?
I ask because in the article "Mosfet Follies":| quote: | "But wait!" I hear you yell. "The MOSFET has a big gate-source capacitance. Won't that suck all the high frequencies out of the signal?".
No, it won't. The big gate-source capacitor is there, OK, but when the device is hooked up as a follower, the apparent capacitance seen by a circuit driving the MOSFET gate is reduced to a very small level by the fact that the source is following the gate almost perfectly, so the apparent capacitance is reduced by the local feedback to an inconsequential level. Even a high impedance 12AX7 plate is not affected in the audio range by this capacitance. Even the big power, big capacitance MOSFETs do this one OK. I have done this, and what's more, done the necessary measurements to find the effect on the response of the composite stage - and there is no detectable change in frequency response below 30kHz. None.
|
So?
|
|
|
| Giaime |
| Updated schematic... |
|
|
| Eli Duttman |
In the case of a source follower the gate capacitance is not much of an issue. HOWEVER, the reverse transfer capacitance does matter. The reverse transfer capacitance of the IRFBC20 is a low 8.6 pF. The IRFBC20 data sheet here.
BTW, if you CCS load a 12AT7 section, you will get lots of linear gain and better drive capability than the high Rp/high mu triodes offer. |
|
|
| Giaime |
Yes, I see. Thank you very much: however IRF820 has a reverse capacitance of 40pF, I don't think this will be an issue. Given the similar rating, surely one could use the IRFBC20 instead.
About the CCS-loaded 12AT7: surely, but there the gain will be no more than the mu of the valve, instead here I need at least 38-39dB, not 35, to employ negative feedback.
Three ways:
1) ECC83 mu follower with nfb,
2) 12AT7 CCS loaded (or even totem pole)
3) two simple grounded cathode stages, in cascade. There should be substantial harmonical cancellation, however to tweak this and to evaluate the designs one should prototype the whole amp (or at least the input + mosfet stage).
I don't see other options, because I don't generally like cascode circuits nor interstage transformers. |
|
|
| tubelab.com |
Been here built that! Look at the 845SE on my web page.
http://www.tubelab.com/845SE.htm
Look at the second schematic on that page.
I started out building an Ongaku clone, and after several itterations wound up with the amp that you see on that page. The power supply makes a BIG difference in the sound. I have done two power supply designs, and I am still not happy with it. I have a power supply from an old Motorola police radio base station that puts out 1500 volts at 1/2 AMP! It gives the best sound, but weighs 120 pounds and the transformer buzzes loudly.
I use mofet drive. I wanted to be able to drive 211's and 845's interchangeably which requires at least 200 volts peak to peak! I have since used this circuit to drive an 833A, which can draw grid current of over 100mA. You can't get this from a single stage. Even if you could, I don't think that an ECC83 has the slew rate capabilities to swing near this much voltage at say 50 KHz. For that you need current. I think you might get there with the 12AT7 at about 10 mA (or both sides in parallel at 20mA). I have been using them (CCS loaded) in my latest amps because 5842's have become scarce and expensive.
I have called the CCS, mosfet drive combination PowerDrive. I spent a couple of months experimenting with this when I first figured it out. I thought about the large capacitances associated with power mosfets (gate - source and reverse transfer) and didn't think that they apply in a source follower. There is however a significant gate to drain capacitance. This is not always specified. I work in a building full of CMOS IC designers, so I picked a few brains. I was told that this capacitance is not constant, and depends on the "channel depth" and "channel width". These are big words that are related to the voltage and current across the device. OK, you just can't look this up on a data sheet, it depends on the application. So, I ordered a bunch of high voltage mosfets and set up a test fixture. I drove them with a CCS loaded 5842 and measured the frequency response. I wasn't smart enough to document everything, but I did find that the Toshiba 2SK2700 has no problem hitting 500KHz at 100 volts P-P. I use this one in all PowerDrive designs. It can take 900 volts.
| quote: | | I don't see other options, because I don't generally like cascode circuits nor interstage transformers. |
Just a suggestion, but use 1/2 of a 12AT7 CCS loaded for the first stage. A low Mu triode or triode connected pentode that can handle a plate voltage of 500 volts. I would try a 6BL7, 6BX7, or a triode wired EL-34. Use a CCS load and mosfet followers between each stage. In other words copy my design without the expensive tubes. It works with 211's AND 845's only by adjusting the bias. |
|
|
| Tubes4e4 |
Hi Giaime,
| quote: | | I ask because in the article "Mosfet Follies" |
1. How small is a small part from 1,3nF and is it small enough?
2. Have a look which MOSFET type is suggested there :)
Tom |
|
|
| Antonio Tucci |
The idea of driving an output tube with mosfet/s is a very good and intriguing one.
The reasons for such temptation are (in my mind):
- reduction of costs
- reduction of complexity
- reduction of dimensions
- good performance.
In this view, I do not understand the tube-based input stage. If we choose to sin, then let's sin all the more.
Why not an input-drive stage completely based on FET ?
Yet ... Could a step-up transformer connected between the FET and the grid of the output tube solve many problems corrlated with the great voltage swing required to drive the power tube? |
|
|
| Giaime |
Thank you all very much George, Tom and Antonio!
I'm thinking about a new input stage.
| quote: | | Why not an input-drive stage completely based on FET ? |
Ehehehe... It could be done. However, while my tube circuits works (I'm not saying that they work good), I'm not so good at silicon... :D
And maybe because AFAIK tubes are better voltage amplification devices, and here before the mosfet we need only voltage amplification.
| quote: | | Yet ... Could a step-up transformer connected between the FET and the grid of the output tube solve many problems corrlated with the great voltage swing required to drive the power tube? |
When you have purchased a GOOD interstage, you have already missed at least one of your points:
| quote: | | - reduction of costs |
;) |
|
|
| mark_titano |
Giaime, did you think about a cascode as voltage amp?
With an e88cc, 400V and a lod of 20K you can provide a gain of 43dB, with high bandwidt and high input impedence.
Tha output impedence should not be a problem cos the mosfet isolate the driver from the output stage.
Mark |
|
|
| Giaime |
So, my choiches:
- 12AT7 CCS (solid state) loaded. Gain will be close to the mu of the tube, at least 50.
- the 6S45 triode, 2 of them in SRPP would give almost the same gain as 12AT7, but with extremely low Zout (45ohm)
- the complex one :D : input is a differential amp, to accept negative feedback signal (maybe 6SN7), driver is a 6H30 SRPP (gain = 12, Zout = 86ohm, maximum swing 177Vpk).
What do you think, and could you suggest other alternatives? |
|
|
| Giaime |
| quote: | Originally posted by mark_titano
Giaime, did you think about a cascode as voltage amp?
With an e88cc, 400V and a lod of 20K you can provide a gain of 43dB, with high bandwidt and high input impedence.
Tha output impedence should not be a problem cos the mosfet isolate the driver from the output stage.
Mark |
Interesting, Mark. I don't particulary like cascode circuits, however this could be an option. Zout is the same as the ECC83 SRPP, so slew rate will have similar limits, or am I wrong? |
|
|
| mark_titano |
No, I think that the output impedance is smaller with the cascode, say something like 14K.
But, the bandwidth is a lot improved and if you are able to implement it right, casdoded amps are great.
They have a really poor PSRR so regulated hum free pws is a must.
Mark |
|
|
| Giaime |
| quote: | Originally posted by mark_titano
No, I think that the output impedance is smaller with the cascode, say something like 14K.
But, the bandwidth is a lot improved and if you are able to implement it right, casdoded amps are great.
They have a really poor PSRR so regulated hum free pws is a must.
Mark |
Hello Mark, sorry to be ripetitive, but that ECC83 SRPP would give me 12k Zout, and a response up to 300kHz. This is simulated, of course, real world differs, but in what parameter do you find this big reduction in bandwidth and frequency response in respect with cascode?
I ask because I don't know :smash: |
|
|
| mark_titano |
A cascode bandwidth with an e88cc is over 1MHz. What you can note with such topologies is that you can provide huge swig with very low distortion and a really flat response in audio bandwidth, when the objectively high output Z is not a problem, even if you need to swing more than 150Vpp.
And…you can provide a really high gain, dependent only by the power supply headroom and Ra you can provide or choose.
All of this with tubes that I personally like more that an ecc83, that I’ve found comfortable only is some topologies, like the partial feedback approach and surely not in srpp.
Another personal opinion, I don’t like much all tubes mu follower and srpp ( I prefer hybrid mu follower) , except when the load is taken in account and tube chosen is something like the 6sn7 or the 6922.
However all is just a suggestion for further investigation ;)
Mark |
|
|
| Giaime |
Thank you very much! ;)
I'll investigate. |
|
|
| Giaime |
Updated design. What do you think?
Please note noise cancellation tecniques :D Yes I've studied a lot on Broskie's articles.
And a theory question: when calculating the resistor needed for a given feedback ratio, in the voltage divider formed by the resistor and the cathode resistor of the input tube, do I have to count the cathode resistor alone or I have to consider the total impedance seen at the cathode (formed by Rk || rk)? In that case, R16 should be halved...
And also, is there a better way to inject feedback in a cascode circuit than this? |
|
|
| kevinkr |
In regards to the Rk || rk question it depends on how big one is relative to the other, if the overall feedback node impedance is significantly lower than R3 by itself (and here I believe it is) then the value of R16 should be adjusted. This probably most easily accomplished at the testing phase of things as the internal cathode resistance is probably pretty small, and not precisely known.
Also you don't really need or want C1 in the feedback path, the values of R3 || R16 can be tweaked to give you the bias voltage you want on the cathode of the 12AT7. I suspect that C1 will introduce quite a lot of low frequency phase shift and also result in boosted response in the subsonic region where your transformer can't handle it, this is probably not an issue unless you play records. |
|
|
| Giaime |
Thank you very much Kevin for your reply.
But... rk isn't (Ra+ra)/(mu+1)? This leads to about 100ohm if I am not wrong, pratically halving the impedance seen at the cathode node.
You're right about C1. But how to mantain the bypassing of rk of the cascode (I must do this or the gain will be too low) and accept negative feedback?
Maybe the upper triode's grid could be a place to put error signal? However there the gain will be quite low, but maybe it will be enought. |
|
|
| tubelab.com |
Below is a copy of e-mail correspondence between myself (George) and Giaime. Since this is directly related to this thread and may help to answer a few common questions relating to PowerDrive I posted it here. I have a cookbook approach to PowerDrive web page coming, but it is not ready yet. I have added a few additional comments at the bottom.
Hello George!
I'm Giaime, from DIYaudio forums. You surely remember about me.
I would like first to thank you for your wonderful website and all the good ideas that you're spreading over the internet: I particulary liked the PowerDrive concept.
I already applied it, to an almost unusued tube: the 802. See the
results here:
http://www.giaime.altervista.org/802.html
Now, I designed a 211 single ended A2 amplifier. I'm not planning on building it, because of the high costs (and voltages) involved, however I would like an opinion from you: here it is
http://www.giaime.altervista.org/hivolt.html
I simulated the mosfet driver in a computer simulation program, using a diode in series with a 470ohm resistor to simulate the 211's grid going positive: there wasn't any distortion with the full 100Vpk drive to the 211 grid. I was running the mosfet at 21mA quiescent current, and I was puzzled, since I expected that the quiescent current would have been at least more than the maximum expected current in the 211 grid. So I made a little guess, that's explained in the article in that page: on
negative peaks of the source (so decreasing current in the mosfet) the source follower has only to drive the 200k resistor and the Miller capacitance of the 211's grid. On the contrary, on the positive peaks of the grid waveform current in the mosfet will increase and so necessary grid current will be available. Is this right?
Please feel free to make suggestions!
Hi Giaime:
I saw the post that you did on the diyAudio forums, and replied to it last night. I think that your mosfet drive ciucuit will work fine. I think that an ECC83 will have trouble driving the capacitance of the mosfet with enough bandwidth. I would suggest using a tube capable of higher current, as you suggested, a 12AT7. I would also consider a different fet. See the forum post for more details.
Yes the Fet must source all of the grid current on the positive going half cycle. R11 must discharge the miller capacitance on the negative half cycle. R13 (the 200K resistor)is probably not needed.
George
Hello George,
thank you very much for your kind and precise response. Sorry if I sent this mail to you, I did this because I wasn't sure to open a thread on the forum. Now I did it, and I'll be extremely grateful to you if you continue your priceless suggestions on the forum, when you have time. I know time is a big problem...
Ps, I've updated the design as you can see in the thread. Now, E88CC in cascode for the input stage. Please tell me if the idle current of the mosfet isn't enought: I have now 20mA and want to drive about 40-50mA into the 211 grid.
Hi:
I also used 20 mA as the idle current on the fet. This was set based on the total dissipation and the size of the heat sink. The fet can source as much grid current as the grid will draw. It is only limited by the source resistance of the +150 volt supply and the on resistance of the fet. I have used the same PowerDrive board with an 833A where I measured peak grid currents of over 100mA.
George
There are two distinct paths for current to flow in the PowerDrive (or any similar mosfet grid driver). On the positive going cycle the grid will draw current. As the grid is pushed further positive its dynamic impedance drops quickly. This current flows out of the +150 (in this case) power supply through the mosfet and into the grid of the tube. The grid current is limited by the RDSSon of the fet (on resistance, typically a few ohms) and the source impedance of the power supply (hopefully also a few ohms). This allows virtually unlimited grid current. Some tubes like the 845, 211, 833A, 811A and some 300B's remain linear well into this region. As the grid transitions into the negative (with respect to the cathode) region it ceases to draw current. There may be a small charge stored in the Miller capacitance of the tube (worse with high Mu tubes like 811A) if the transition is rapid. There should be a low impedance for this current to flow. In this case it is R11 and the current flows into the negative supply. 4.7k is sufficient to rapidly discharge the capacitance. I used 10 K. The negative supply must be sufficiently negative to ensure that the fet remains conducting as the grid reaches its most negative extreme. -150 volts is sufficient for a 211 or an 833A. It is not sufficient for an 845. I used -300 volts. The quiescent current of the fet is not terribly critical but it is set by the value of R11 (and the bias voltage). This resistor also sets the output impedance of the driver under conditions where no grid current is flowing. I wanted this to be low, which leads to a high idle current in the fet. The upper limit for fet current is determined by the dissipation capabilities of the fet and its heat sink. I have both fets, both resistors, and both driver CCS's on the same (large) heat sink. I settled on 20 mA as a happy medium between fet temperature, and output resistance. I tried a few different values and found little if any sonic differences above about 10 mA. |
|
|
| Giaime |
Thank you George of publishing our conversation, I strongly think that it could help people to desing PowerDrive circuits (however you could ask me the next time: not everyone like to have their own private mails published).
The question now is: negative feedback. Do I need it? I think yes. A2 operation can, if the driver isn't "ideal", deliver some distortion.
I simulated the output stage with SE amp CAD, and I can get up to 40W on 8ohm load, keeping total distortion under 5%. 1% distortion is about 5W.
Distortion here is essentialy produced by the output stage, like every good amplifier should. So I think that -6dB of negative feedback are needed, and I have the gain to allow that.
I know many "audiophile" designers would avoid nfb in a 211 SE amp, in fact if we want low distortion we wouldn't make a single ended amp, right? ;)
Instead of those semplifications, my idea (from an EE wannabe) is that we CAN design a low distortion 211 SE amp, and a little bit of nfb can further improve it. I've seen JUST TOO MANY 845 amps driven by an ECC82 sold at outrageous prices :dead:
The problem is: where could I put it? Is there a way to insert the error signal in the cascode input stage? |
|
|
| tubelab.com |
Sorry for not asking first. I read this line in your e-mail:
| quote: | | Now I did it, and I'll be extremely grateful to you if you continue your priceless suggestions on the forum |
and took it as a request to take this discussion to the forums. I cut out all but the technical stuff and pasted it here. I did this to help illustrate the PowerDrive concept. I get a lot of e-mail about PowerDrive. Yours was one of the clearest, easy to understand that I get, and it had already been brought to the forums.
I usually have small pieces of time during my work day when I look at the forums, I usually can't look at e-mail from work.
| quote: | | The question now is: negative feedback. Do I need it? I think yes. A2 operation can, if the driver isn't "ideal", deliver some distortion. |
I use no feedback in my 845 amp. I also simulated the amp in SE amp cad before building it. The measured results agree closely with the simulations. I used an 845 for my simulations and the initial amp design. I have tried 211's in the amp, but prefer the sound of the 845.
My amplifier used some real low cost output transformers that are no longer available. It sounds good and makes about 40 watts at 5% distortion. I have found that cathode feedback works well on most (but not all) output transformers to reduce output stage distortion without killing the sound. Cathode feedback increases the drive requirements and I don't have enough voltage on the driver to get any more drive, so I haven't tried it on this amp yet.
I don't have much experience with the cascode circuit that you show, so I didn't comment on it earlier. I would assume that you could apply feedback at the cathode of the input stage as shown in your schematic.
| quote: | | You're right about C1. But how to mantain the bypassing of rk of the cascode (I must do this or the gain will be too low) and accept negative feedback? |
Put a small unbypassed resistor in series with a larger, bypassed resistor, then inject the feedback across the small resistor. |
|
|
| kevinkr |
Hi Giaime,
The suggestion tubelab had about deriving the bias across the a larger bypassed capacitor in series with a smaller feedback resistor is a good one. Something on the order of 47 ohms seems appropriate.
Since the open loop gain is finite, the ratio of the resistors required to set the closed loop gain will be a little higher than the actual gain ratio, depending on the quality of the 12AT7 model and how well it models the nodal impedance at the cathode your simulation and real world might be a little different as well.
You can think of it this way for a non-inverting amplifier.
The open loop gain A is finite.
The feedback factor which is also the reciprocal of the desired closed loop gain is B.
ACL is the actual closed loop gain so,
ACL = Vout/Vin = 1/B x [ 1/(1+1/AB)]
Note that 1/B might be expressed as the ideal gain provided that A is infinite or at least very large.
The term that follows takes into effect the less than infinite available open loop gain (OLG) and is ignored for most op-amp applications except in ATE and in our case where the open loop gain is finite.
Try the equation with progressively lower amounts of OLG and you will see exactly what I am talking about.
As to whether this is useful or not, I thought you might find a little feedback theory interesting. |
|
|
| Giaime |
Thank you very much George and Kevin, your help is much appreciated.
I'm not sure I understood the nfb connection you suggest. Is this correct?
I also have verified the difference between 1/B and ACL, as Kevin said, and I have to underline that this difference is negligible (0.005%) not only when A is very large, but when the feedback factor is low, so A*B is very large, and 1/AB tends to zero. In fact here I'm using -6dB of feedback only, giving an input sensitivity of 1.72Vrms for a power output of 40W on 8ohm (that I consider the extreme maximum).
Note that I have adjusted the cathode resistor of the first stage, because now rk is in parallel with the feedback resistor (and the OPT secondary, which is negligible) and so to mantain the same value it has to be increased (from 120 to 138ohm).
I have still doubts about the decreased open loop gain due to a portion of rk unbypassed. But: if there's a feedback signal there, won't the 47ohm resistor be effectively bypassed (AC ground)?
Also, but I think those things are meant to be tweaked in the real world, not on paper, we have still to consider the total impedance seen at the cathode of the cascode circuit. If I'm right and I remember well what Broskie said about cascode, the impedance seen at the anode of the lower tube is Ra/(mu+1), so about 300ohm. This, divided another time for (mu+1), gives about 8ohm at the lower triode's cathode. Is this effectively in parallel with Rk? |
|
|
| tubelab.com |
| quote: | | I'm not sure I understood the nfb connection you suggest. Is this correct? |
That is what I had in mind, except that I usually swap the positions of R17 and the R3 - C1 pair. This way the small amount of AC current that flows from the output through R16, R17 and C1 will not go through C1. With the current phobia about putting electrolytics in the cathode circuit of a voltage amplifier, why stress the capacitor any further? |
|
|
| Giaime |
| Yes, but that way the error signal will go through C1!!! That's worse in my opinion. :smash: |
|
|
| kevinkr |
I think tubelab is right on this one, there will be much less signal current flowing through that cap with his suggested orientation, and regardless of where you locate it is still part of the feedback network.
Another reason to do it the other way is it makes altering the operating point of that amplifier stage very easy as you can basically change the cathode resistor with no interaction with the rfb resistor.
Good low voltage blackgates are still available and are very good. I'm not a big fan of EL types either. In some cases though the circuit performance will be a lot better with them than without. Your other option is to use fixed bias on the grid of the lower tube and connect it directly to the feedback network or ground if global nfb is not used.
In theory at least the impedance at the feedback node will be small but not zero due to the limited loop gain, it may be that reducing the value of the 47 ohm resistor and scaling rfb may give you a little additional loop gain although I doubt it will make a significant difference, it might still be worth trying.
Like Tubelab I eschew the use of global negative feedback and I suspect substituting a lower mu driver tube or using an 845 (which many seem to prefer incidentally) would eliminate the need for nfb.
I tried varying amounts of nfb in my early se designs as well as an otl and pp amplifiers, and invariably came to the conclusion that the amplifier in question (even the OTL) sounded best with no feedback.
Should you need a lower source impedance I recommend you trade power output by choosing an operating point optimized for a higher primary Z - in many cases this will help dynamic linearity as well, while giving you a lower source impedance/higher damping factor.
This is just my opinion, YMMV..
:D |
|
|
| Giaime |
Thank you very much Kevin. In fact, I designed the input stage to accept 845 output tubes: a little tweak on bias, mosfet power supplies and OPT primary impedance can allow the 845 to be used, of course one must eliminate global NFB.
I'm very sad that this design will remain on paper for looooooong time :whazzat: :bawling:
Edit: an updated version of the schematic online soon on my site, at the usual address. I'm working on an 845 variant. |
|
|
| tubelab.com |
| quote: | | I tried varying amounts of nfb in my early se designs as well as an otl and pp amplifiers, and invariably came to the conclusion that the amplifier in question (even the OTL) sounded best with no feedback. |
I even built a few amplifiers P-P and SE with a variable global feedback control. I let several people borrow them and asked the users to tell me what setting they prefered on the knob (I didn't tell them what the knob did). All users that already owned tube amplifiers (and tube friendly speakers) preferred the zero feedback settings on all of the amplifiers, even the P-P amp. Several users that were into wall shaking home theater preffered feedback, often lots of feedback, especially on the KT-88 push pull amp. It is possible that some users preferred zero feedback because the amp will play louder, even though there is some distortion.
I have been experimenting with cathode feedback in the output stage only. I find that you can apply a relatively large amount of cathode feedback without destroying the sound. This is not always practical on some designs however, especially a P-P amp.
My latest amp (SE) design uses a CCS loaded 12AT7 driving a 6L6, KT-88, or an EL-34. It is switchable from triode to UL to pentode. There is also a switch to turn the cathode feedback on or off. With some output transformers I can select different amounts of cathode feedback (by using different secondary taps).
I am making up a good looking amp with no tubes visible from the outside of the amp. Some users may not realise that it is a tube amp. There are several switches (including the ones mentioned above) on this amp that make some changes (like tube or SS rectifier) and some that do nothing. The users will be asked to try each switch and rate what it does to the sound, and find the combination that they like best, in their home with their speakers, and their music. There is a spreadsheet for them to rate the sound of each switch and to list the particular music that was used. I have already got several people waiting to audition this amp (mostly engineers who will be critical), some own tube amps, most do not. I will be interested to see the results of this test. It will be a few months before it is complete.
I have found that I like zero feedback with the amp in triode mode, but it wants feedback in UL depending on program material and volume level, and NEEDS feedback in pentode mode.
This is just one of those times where all the engineering, simulations, and performance testing can not relate to what people want to hear, and everyone hears something different, often based on what they are used to hearing. |
|
|
| Giaime |
That's an EXTREMELY interesting test George, please keep us informed.
The deepest listening test I've done on NFB on tube amps... is like your tests George, hooking a guitar to the amp and a speaker combo :D
I did this for the EL36 PP amp I'm working on. I liked it very much when cleanly driven hard by a guitar preamp, It's extremely loud even now that has no more than 5W. Obviously, I liked it better without feedback, even if I'm using only 4dB or so.
I have yet to audition it with music and hi-fi rig.
Ps. updated schematic of the 211 beast online. |
|
|
| Giaime |
A question for all 845 amp builders:
what is your favourite operating point, at what OPT primary impedance?
I ask because playing with load lines don't give me good results, efficiency is lower than the 211. I'm trying 100mA, 960V, 10k load and with 180Vpk input signal I have 34W, but I'm puzzled about the low Imax of the tube, only 120mA, and note that most of the time this limit is broken with this operating point.
I know, this is the continuous limit, not peak one... but... :xeye: |
|
|
| tubelab.com |
I use 1100 volts B+. The OPT is 10K ohms. I set the current at 70 ma when using vintage tubes to be safe. I have run the new production Chinese (Shuguang) tubes as high at 110 Ma (over the dissipation limit) with out any issues, although I usually run 80 to 90 mA. There is some sag in the power supply as the current is increased, and some voltage drop in the OPT so the tube sees about 1025 to 1050 volts.
I briefly connected this amp up to a BIG power supply that put out 1500 volts (over the line again). I set the current to 70 mA, using Shuguang tubes (I can get these for $30 each, so it was a relatively low cost risk). The amp positively ROCKED! I could play just about any music at any volume level without distortion.
The 845 has a lower Mu than the 211. This requires a lot of drive voltage. The higher the plate voltage, the more negative the grid must be. This just means that you can drive it harder and get more power. I swing the grid positive to +50 volts on peaks.
I played with the design in SE amp cad before I built it, and I remember seeing 40 watts in the simulations with the grid going positive. I can't remember the details since it was over three years ago. |
|
|
| Giaime |
Thank you George. So my design could accept 845 tubes, because:
- 10k OPT already in place
- B+ is almost the same
- current is similar so PSU can be quite similar
Things to be tweaked:
- mosfet power supplies: +-150V is not enought, it should be +150 -300 or more
- gain of the cascode input stage is enought if you remove nfb but output swing of the cascode must be tweaked, now it can't swing much more than 150Vpk, I need a little more.
However, I don't understand a thing. | quote: | | The 845 has a lower Mu than the 211. This requires a lot of drive voltage. The higher the plate voltage, the more negative the grid must be. This just means that you can drive it harder and get more power. |
Yes, I HAVE to drive it harder, but power is the SAME because of the lower mu! In fact 845 is even a tad less efficient than 211 at that operating point (regardless of the needed input swing), and 211 has better distortion figures. |
|
|
| mark_titano |
Giaime have you thought about the GM70 as natural 845/211 substitute?
Higher plate dissipation, higher mu and you can find it for cheap on ebay...even, due to lower rp, you can use a 6/7K opt…
http://frank.pocnet.net/sheets/084/g/GM70.pdf
You'll pay the higher performance losing a lot of power in the heaters but, as this is only an evaluation, I strongly recommend to take this tube in account. It is easier to drive and provide more power than 845 or 211 ;)
Mark |
|
|
| Giaime |
| Thank you very much Mark, you gave good food for thought :D |
|
|
| Giaime |
Some very rapid calculations, and rough guesses, lead to this following data:
8k load, 1kVak, -90Vgk, 120mA = 42.5W with Vin = 130Vpk
10k load, 1.2kVak, -120Vgk, 100mA = 40.5W with Vin = 140Vpk
6k load, 800Vak, -60Vgk, 150mA = 40.8W with Vin = 120Vpk |
|
|
| ilimzn |
| quote: | Originally posted by tubelab.com
My latest amp... There are several switches (including the ones mentioned above) on this amp that make some changes (like tube or SS rectifier) and some that do nothing. The users will be asked to try each switch and rate what it does to the sound, and find the combination that they like best, in their home with their speakers, and their music. |
Briliant idea about the switches that do nothing, I am looking forward to the results!
Regarding MOSFETs and nonlinear capacitances:
Cgd is indeed not a constant but it is not extremely nonlinear as long as you don't go into low Vgd conditions - how low, depends on the actual MOSFET but usually most of the problems are avoided by having a minimum Vgd of about 5-10V. In general, the nonlinearity is higher on lower max Vds rated MOSFETs.
As long as low Vgd is avoided, look for the lowest Cgd (reverse transfer capacitance).
For truly low capacitance, a self-cascoded follower can be used. In this design, there is another MOSFET 'on top' of the follower one, and it's gate is referenced to Vgd_minimum + Ggs_threshold over the output of the follower (This usually ends up being about 10-15V minimum). It is fairly easy to do with a high resistance divider from follower output to +V, with the resistor from follower output to cascode MOSFET (the top one) gate bypassed by a good low inductance capacitor. Relevant protection diodes and gate stopper resistors must be applied of course! This arrangement reduces the apparent Cgd of the follower MOSFET about as much as it's apparent Cgs is reduced due to follower operation - in essence, the top MOSFET bootstraps the Cgd of the bottom one, and the bottom MOSFET drives the Cgd of the top one, for which it has plenty of current available.
It is possible to do this with a BJT-MOSFET combo (BJT for bottom, MOSFET for top device) but not truly suitable for A2 operation as the BJT will suddenly require current from the tube driver stage when grid current of the output stage becomes positive. Even though this load is reduced by Beta of the BJT, it is a much more abrupt change than what you get on MOSFETs, even though the input capacitance of such a combo can be made truly miniscule.
Finally, the Cgs of a MOSFET in a follower conenction is usually reduced proportional to it's Gm. For a given MOSFET architecture, Cgs is roughly proportional with Gm, which is again proportional to it's current handling (think of an incerase in Cgs, Gm, Idmax as paralleling MOSFETs). Unfortunately, so is Cgd - which is why, when chosing MOSFETs for follower applications, it is necessary to use the one with the lowest current handling that satisfies your needs (plus safety factor, of course), as this will give you the lowest Cgd. |
|
|
| tubelab.com |
| quote: | | mosfet power supplies: +-150V is not enought, it should be +150 -300 or more |
That is what I used in my amp. +150 and -300 volts. Easy voltages to make from a 120 volt transformer secondary, which is verry common here.
| quote: | | However, I don't understand a thing. |
I guess I wasn't clear when talking about the 845. I was trying to say that an increase in plate voltage requires an increase in bias voltage which allows an increase in drive, resulting in more power for almost any given tube. Just cranking up the plate voltage with no other changes only results in a hotter tube.
| quote: | | Briliant idea about the switches that do nothing, I am looking forward to the results! |
The "do nothing" switches are to rate the effectiveness of the tester. I have loaned my amps to a lot of people. That is where many of my sales come from. Sometimes a user will hook up the amp and play with it, but not really pay attention to what he is doing, then they will fill out the review sheet without thinking about it. Other times a person will borrow an amp and never even hook it up. Then they will tell you that it worked great. The "do nothing" switches, along with knowledge about what the other switches do, will allow me to find these users.
| quote: | | Cgd is indeed not a constant but it is not extremely nonlinear as long as you don't go into low Vgd conditions |
Unfortunately, all of the experts that I consulted work on GHz frequency logic and RF chips that work at 1.2 and 2.5 volts. A far cry from the hundreds of volts that we play with. That is why I simply ordered a few of every cheap fet that DigiKey sells that can take 600 volts or more and tested them. I tried to pick the lowest current fets that had the lowest capacitances associated with them. Several new parts have appeared since I ran that test, maybe its time to do it again. |
|
|
| Miles Prower |
| quote: | Originally posted by tubelab.com
I use 1100 volts B+. The OPT is 10K ohms. I set the current at 70 ma when using vintage tubes to be safe. I have run the new production Chinese (Shuguang) tubes as high at 110 Ma (over the dissipation limit) with out any issues, although I usually run 80 to 90 mA. There is some sag in the power supply as the current is increased, and some voltage drop in the OPT so the tube sees about 1025 to 1050 volts. |
This is what I've been looking for. However, you say that Zl= 10K. The Lundahl iron I've been considering was obviously wound for the Q-Point specified in the 845 spec sheet: Zl= 9.0K. What's the output xfmr you're referring to? That Lundahl iron looks pretty good, but price-wise: :bigeyes:
Here is a very preliminary design. This includes little more than basic topology. |
|
|
| tubelab.com |
The transformers that I used were purchased off of Ebay from Handwound Transformers. When he first started selling transformers he actually delivered product and the early ones worked pretty good. I got the pair for $120 including shipping. It only took 4 months to get them.
I got a pair of P-P transformers later on and they don't sound very good. After that he just took peoples money and sent them NOTHING! There are a few threads regarding Mr. Lucas, and it looks like I was one of the few people who actually got working transformers from him.
For a new design it looks like tha Hammond 1638SEA (10 K ohm) about $140 each or the Hammond 1629SEA (6.5 K ohm) at $99 each. If you find anything cheaper let us know.
I am still looking for an OPT for the "big one", an 833A SE amp. I have everything except the OPT. The Hammond 1642SE would work, but it is $260 each and no one has any in stock.
For a power supply consider a voltage doubler fed by a 480 volt "industrial control transformer". Maximum power for minimum money. I got a 750 VA transformer for $22 on Ebay. |
|
|
| Miles Prower |
| quote: | Originally posted by tubelab.com
[B]The transformers that I used were purchased off of Ebay from Handwound Transformers. |
Say no more. I caught that thread too.
| quote: |
For a new design it looks like tha Hammond 1638SEA (10 K ohm) about $140 each or the Hammond 1629SEA (6.5 K ohm) at $99 each. If you find anything cheaper let us know. |
Anything cheaper, nope, can't help there. :no:
| quote: |
For a power supply consider a voltage doubler fed by a 480 volt "industrial control transformer". Maximum power for minimum money. I got a 750 VA transformer for $22 on Ebay. |
Yuppers, considered that:
ICT --> X2 --> active regulator
Gives enough margin for the regulator to hit 1.1KV without dropout. Cheap enough to give each monoblock its own HV PS, and avoid cabling 1100V between a remote PS and the amp. |
|
|
| waltube |
Hello Giaime,
please consider to use the Gomes Totem pole with 12AX7, it has a gain of 70 and output impedance of 500 ohm; the V+ is about 480V .
In origin this was used to drive a plate of 211 push pull (double Tomem driven from a input unbal/bal. trasf.).
PLEASE leave the FET; it is a non sense with georgeus 211 (in my opinion).
Ciao
Walter Gentilucci |
|
|
| Giaime |
Ciao Walter,
glad to see you here.
Have you made measures about the setup you're suggesting? Are you sure it can drive a 211 in A2? We're talking of 50mA grid current here... ;)
(ps an useful suggestion from you... it has been long long long time from the last one :D ) |
|
|
| waltube |
In every case is not raccomanded to drive in A2 the grid of big power triodes.
With the Gomes circuit you can have a simply but very efficently way to drive the grid of 211.
With 1250 Vdc and -80 v of bias you have about 60 mA of current, with a load of 10k you will reach about 19w (of course if you use a good 211 not a chinese!). The dissipation of ther tube is about 75w. If you stay within this spec, the Gomes circuit works fine.
The main problem is not the grid current (you don't need!) but the quality of output trasf. It MUST be at the highest level otherwise you can design the best circuit in the world since the Roman empire but the sonic results (mainly in the low frequencies) wil be poor. This means that it is expensive; I worked on some of different circuit and the problem is always the same.
Ciao
Walter Gentilucci |
|
|
| Giaime |
I know, I know that. In fact in the article on my site I mention the problem of the output transformer.
Since I'm not planning on building it, no problem ;)
Why this bashing of A2 operation? What kind of problems? |
|
|
| Miles Prower |
| quote: | Originally posted by Giaime
Why this bashing of A2 operation? What kind of problems? |
I've heard rumours that grid circuit nonlinearities can compromise sonic quality. I suppose that can happen if your driver has too high Zo. Sounds like a good interstage xfmr or a MOSFET source follower will fix it. I like to use strong drivers that can charge up the input/stray capacitance and improve clipping behaviour on transients anyway.
Other than that, it's an audiophile rumour. Who knows? |
|
|
| Tubes4e4 |
Hi Giaime,
| quote: | Originally posted by Giaime
Why this bashing of A2 operation? What kind of problems? |
Maybe just unfounded fear of asymmetrical driver loading/operation when crossing the border to grid current. Nothing to worry about if done right.
I have auditioned two or three well engineered A2 op amps during the last years (but sadly all of them using the 813 tetrode) and I had not much to object.
Tom |
|
|
| tubelab.com |
| quote: | | Why this bashing of A2 operation? What kind of problems? |
Maybe this person never built (or heard) a GOOD A2 amp. I realize that it is real easy to build a bad A2 amp. And to build a good one you need an expensive IT transformer, or a FET. Most cathode followers will be asymetrical, and sound bad. Many poorly designed A2 amplifiers do not sound good. Some tubes like the 811A REQUIRE A2 operation unless operated at very high plate voltages. There have been several nice sounding 811A amplifiers built.
| quote: | | about 19w (of course if you use a good 211 not a chinese!). |
Yes about 19 watts is what you get from a 211 in A1. It is relatively easy to get 30 watts or more in A2 without violating the specs. If you don't turn up the volume, you never enter A2, but you have more headroom for transients. You can have clipped transients, or loud ones that MAY have some A2 coloration on them.
And by the way I have seen over 40 watts at 5% distortion using Chinese 211's. I certainly am not going to test NOS ones (I only have 2) at this power level.
Those are just my opinions. I have built several A2 amps. All of my DHT designs have A2 capability. It is useful on many tubes (including Chinese 300B's). Even if the tube goes non linear in the grid current region the resulting distortion is less objectional than clipping. |
|
|
| Giaime |
I must say I've made some experiments about Powerdrive, and in general low-Z drive for the output tubes.
I'm driving EL36 push pull in triode mode with a 5687 cathode follower, now it's AC coupled and I can strongly tell the difference in comparison with a differential amp made up with 12AT7. Less gain (obvious) but much more live sound, drums are less compressed.
Now I will try DC couple the cathode follower to the power tube's grid. I'm also doubtful, shall I use a mosfet instead of the cathode follower? |
|
|
| tubelab.com |
You have heard opinions on both sides. I don't know your particular circuit, but it may be possible to use either. If your circuit permits, try it both ways and let us know which you like best. The cathode (source) resistor can be much lower with a fet, due to its lower on resistance.
I did this a few years ago, and you already know which choice I made. It is also possible to put a pentode into the same circuit. Some say that a pentode makes a better cathode follower. I tried a 12BY7, but wasn't impressed. Perhaps a larger tube would have worked better. |
|
|
| Giaime |
I'll let you know. For now I've got to build up a stiffer B- supply: because for DC coupling of the driver to the output tubes it has to "sink" driver current other than providing negative bias to the output tubes.
What's the best way? I need about -100/120V. Half wave rectify the HV secondary or better use a transformer with a dedicated winding? |
|
|
| Tweeker |
| Dedicated transformers for everything is best, ignoring cost weight and space considerations... |
|
|
| waltube |
Regarding 211 is interesting to verify that the datasheet from RCA speaks, about class A operations, only for A1 (this means not grid current flows) and the max anode wattage is 75 and 1250 volts max for anode.
Is also explained other class operations, but is not for our purpose (class B and C). In every case the A2 condition is not mentioned.
I can agree if you go on A2 you can have more power but, in sinlge end, the main problem is (again) the transformer followed by the trouble to set the driver stage. Is difficult to understand that, in s.e., the problem is not (only) the final power we can obtain but the capability to drive the load at low frequencies (=give current to a real load ).
The problems at low frequencies are very important, a trasformer with 10 k primary, 4/8 ohm secondary, that is capable to handle about 75 bias (at 1250 Vdc max) and goes down, without problems, to 20 Hz, full power 0 dB, its dimension are equivalent to a main trasf. more than 300w.
I used a power trasf. for High voltage with a double C nucleus capable of 550w at 50 Hz, the dimensions are 13 x 13 cm, heght of 16 cm (enclosed). With the same nucleus on output trasf. (always 10 K) I reach the 20 w at 20 Hz, low THd, good square wave response ( very important to look, you can understand lots of things)) and a good response at high frequencies, the weight was about 9 kg ( about 8 for power). The cost are very high ( about 1200$/pair for output and 400$ for power) and are custom made.
After these experiences, also with 811A and 572 , I switch to a p-p of 300B with dedicated trasf., always double C, custom made, they sound much better , with a driver circuit very simple to set .
These are my opinions, they comes from a long experience on this field.
Happy to dicuss again.
Ciao
Walter Gentilucci |
|
|
| tubelab.com |
I agree that for any tube amplifier, especially a SE amplifier, that the single most important component is the output transformer. In my case I got the OPT's from a person that took a lot of peoples money and gave them NOTHING. The fact that I got anything from him is unusual, the fact that they are good transformers is really unusual. These transformers weigh over 10Kg each. I ran a frequency response test at 10 watts and got 22 Hz to 33KHz. I am sure that they are not capable of full power at 20Hz but the entire amplifier was built for about $400USD.
I got the transformers first, then started to design the best amplifier that I could around them that fit my budget. I do that with a lot of my designs. I started out to build the "Poor Man's Ongaku" form an article in an old Sound Practices magazine. I was not impressed with the sound that I got from my version of that circuit, so I started changing things. The amp that is on my 845SE web page is the result. It bears no resemblance to the original Ongaku design. I also wanted to use 845's (which can be operated in A2) and 838's (which MUST be operated in A2) because I had some NOS tubes.
True the 211 is not specified for A2 operation. A2 was not common in the 1930's when the original data sheet was written. The RCA data sheet that I have (dated 1936) does list class B modulator service with a plate dissipation of 100 watts. In later books 100 watts was listed for all classes of operation. The data listed for class B operation (1250 plate voltage case) shows a bias voltage of -100 volts, but a peak AF drive voltage of 205 volts (1/2 of 410) this is 105 volts positive on the grid. It also shows a drive POWER of 4 watts (1/2 of 8) dissipated in the GRID. Even in 1936 they were driving the grid positive. Look at the plate curves, they only go to - 125 volts, but show positive values to +225 volts. Plot a load line for a reasonable load (5 to 10 K) and you will see that the curves remain linear well into the positive grid region.
I am not stating that everyone should consider A2 operation. It does require careful design. I am saying that it should not be automatically discounted because it is difficult and uncommon. This started as an exercise in "theoretical" design. It is good engineering practice to consider all possible design solutions, then narrow down the choices based on the given constraints. The common constraints are budget, size, power dissipation, etc. We have all presented our opinions, but it is up to the designer to choose his particular design path.
I built this 845SE amplifier, I also built a push - pull 300B amplifier, and several other smaller SE amplifiers. They are all placed in a rack so that I can use any one of them. The 845SE does not get used as much as the others due to the amount of heat it generates. In the summer here the outside temperature reaches 36 degrees C most days. Even with air conditioning I can run that amp for about 1 hour before it gets too hot in here. It is also quite large. I usually listen to one of the small SE amps. When I want to get loud I will turn on the 300B P-P or the 845SE depending upon my music choices. |
|
|
| waltube |
This his only a discussion regarding different approach; of course the final goal is the quality of the sound (not only on lab).
I am still working on OPT to reach the best.
In my last design, p-p of 300B, I used a custom opt, as told before, in double C core (as power and choke).
I worked on low output impedance.
I have a switch on output to insert NO feedback, 6 dB of FB and 12 dB of FB.
Without FB the output imp. is less than 2 ohm, with 6 dB is about 0,8 ohm and with 12 dB is 0,45 ohm. This means that, at 10 v - 8ohm out, 1 KHz, if I switch to 4 ohm (with 12 dB FB), I have 9,8 volts. At 20 Hz it goes at 9,1 volts. Great performances because I can drive an hard loudpeaker without problems; in this period one amp is driving a B&W 801 II series. In this case if I switch on NO FB the sound is not clear, on bass, as with 12 FB inserted, this because a large woofer (30 cm diam. , very heavy ) need to be "controlled" by good Damping factor (otherwise you got gummy bass); on other situation with a two way Altec loud speaker (woofer 515B in bass reflex and horn 311-60 with 288H driver), 116 dB and the sound is superb without FB.
On this link:
http://www.multitask.it/FB/A-100Hz-ingusc.jpg
you can find a picture where in blu is the input and red the output
of a 100 Hz square wave, 6 Vp-p on 8 ohm, with 12 db FB.
The response at full power is 0db at 20 Hz, -3dB at 90KHz with no FB; 0dB at 15 Hz -3dB at 120 KHz at 6dB of FB; 0 dB at 12 Hz, -3dB at 230 KHz.
In my opinion is difficult to obtain the same performances with s.e. ( in general).
Ciao
Walter Gentilucci |
|
|
| snoopyma |
| quote: | Originally posted by tubelab.com
Been here built that! Look at the 845SE on my web page.
http://www.tubelab.com/845SE.htm
Look at the second schematic on that page.
I started out building an Ongaku clone, and after several itterations wound up with the amp that you see on that page. The power supply makes a BIG difference in the sound. I have done two power supply designs, and I am still not happy with it. I have a power supply from an old Motorola police radio base station that puts out 1500 volts at 1/2 AMP! It gives the best sound, but weighs 120 pounds and the transformer buzzes loudly.
I use mofet drive. I wanted to be able to drive 211's and 845's interchangeably which requires at least 200 volts peak to peak! I have since used this circuit to drive an 833A, which can draw grid current of over 100mA. You can't get this from a single stage. Even if you could, I don't think that an ECC83 has the slew rate capabilities to swing near this much voltage at say 50 KHz. For that you need current. I think you might get there with the 12AT7 at about 10 mA (or both sides in parallel at 20mA). I have been using them (CCS loaded) in my latest amps because 5842's have become scarce and expensive.
I have called the CCS, mosfet drive combination PowerDrive. I spent a couple of months experimenting with this when I first figured it out. I thought about the large capacitances associated with power mosfets (gate - source and reverse transfer) and didn't think that they apply in a source follower. There is however a significant gate to drain capacitance. This is not always specified. I work in a building full of CMOS IC designers, so I picked a few brains. I was told that this capacitance is not constant, and depends on the "channel depth" and "channel width". These are big words that are related to the voltage and current across the device. OK, you just can't look this up on a data sheet, it depends on the application. So, I ordered a bunch of high voltage mosfets and set up a test fixture. I drove them with a CCS loaded 5842 and measured the frequency response. I wasn't smart enough to document everything, but I did find that the Toshiba 2SK2700 has no problem hitting 500KHz at 100 volts P-P. I use this one in all PowerDrive designs. It can take 900 volts.
Just a suggestion, but use 1/2 of a 12AT7 CCS loaded for the first stage. A low Mu triode or triode connected pentode that can handle a plate voltage of 500 volts. I would try a 6BL7, 6BX7, or a triode wired EL-34. Use a CCS load and mosfet followers between each stage. In other words copy my design without the expensive tubes. It works with 211's AND 845's only by adjusting the bias. |
Hi TubeLab,
I read your Powerdrive design with great interests. I want to build a single-ended power amplifier using 805. I want to use your Powerdrive to drive the 805 to A2. Will it be feasible and sounding good? If yes, when will your Powerdrive be available for sale?
I want to try the IXYS 10M45 CCS chips that you recommend for plate load on these forums and I will buy some from the Digikey. Can the chips be used to replace the cathode resistor of a self-biased cathode follower (e.g. 6EM7) easily? If yes, can you elaborate?
Thanks.
Regards,
T.C. MA |
|
|
| widowmaker |
Sorry for my late reply, but I just take a look at your thread...
There are two bad "things" in your schematic:
p.1. MOSFET follower
p.2. N feedback (more harmful than p.1)
Good driver stage for russian GM70 and 211 direct heating triode is loftin-white schematic with russian 6N6P double triode, Tungsol/RCA 5687 or similar double or two single triodes... |
|
|
| Giaime |
| I think we would all hear your opinions on those two things, possibly supported by scientific proofs and measured data. |
|
|
| Miles Prower |
| quote: | Originally posted by widowmaker
Sorry for my late reply, but I just take a look at your thread...
There are two bad "things" in your schematic:
p.1. MOSFET follower
p.2. N feedback (more harmful than p.1)
Good driver stage for russian GM70 and 211 direct heating triode is loftin-white schematic with russian 6N6P double triode, Tungsol/RCA 5687 or similar double or two single triodes... |
I assume you mean my preliminary design for an 845 SET, since mine was the only one that included gNFB. On your two points, I gotta disagree.
Followers (cathode/source/emitter) are the least offensive type of stage. With the Source Follower, you avoid the problems of Cgs nonlinearity since the gate and source are at nearly the same potential. Unlike the Emitter Follower, the input impedance is not dependent on the load at the output. It is better than a Cathode Follower since the MOSFET has much higher gain, therefore, a much lower Zo. Low Zo is desireable for driving the finals into Class A2. The lower, the better. MOSFETs do that very well. I'm not a "sandophobe", and will use SS when/where it does the job at hand the best.
As for p2, rejecting gNFB is just plain nutz. Now, it is quite true that the technique has gotten a bad rep due to its misuse by lazy designers who try to compensate for a bad fundamental design by simply pouring it on in a misguided attempt to sweep those fundamental design flaws under the carpet. Unforch, they also sweep away a lot of the details of the program material. This is not a good thing.
However, there is no doubt that gNFB can make a good design even better. Now, you might argue that, since the 845 has an r(p)= 1700R, that it isn't necessary. However, you can improve speaker damping by getting that Zo down with NFB. Isn't poor bass performance a big complaint against SETs? Why put up with that? Also, active device nonlinearities and the nonlinearities of magnetics are unavoidable. gNFB can compensate. There is no good reason not to use it.
Yes, I know that it's "heresey" in a SET. But heresey be damned! That's how I intend to design 'em. I want something that sounds really good, and that's what gNFB does best: make a good design sound even better. |
|
|
| poobah |
Get 'em Miles... get 'em!
:cool: |
|
|
| Giaime |
Dear Miles,
I appreciate much your response and that's exactly my opinion on gNFB and the source follower. However, I think he was referring on MY amp that included global negative feedback as well: take a look |
|
|
| Miles Prower |
| quote: | Originally posted by Giaime
Dear Miles,
I appreciate much your response and that's exactly my opinion on gNFB and the source follower. However, I think he was referring on MY amp that included global negative feedback as well: take a look |
It would be a good thing if more folks learned how to use the quote feature. ;)
Except for the hybrid Kimmel, it looks very similar. Great minds think alike. :cool: |
|
|
| Giaime |
| quote: | Originally posted by Miles Prower
Great minds think alike. :cool: |
Please don't say that, you'll embarass me :innocent: |
|
|
| tubelab.com |
| quote: | | I read your Powerdrive design with great interests. I want to build a single-ended power amplifier using 805. I want to use your Powerdrive to drive the 805 to A2. Will it be feasible and sounding good? If yes, when will your Powerdrive be available for sale? |
As I have stated on my web site, the PowerDrive circuit can be adapted to many different designs. Because of this each one is slightly different from the others. This is usually because of the different voltage and power dissipation requirements. This means that several different PC boards would be required to cover all of the different cases. The economics of PC board fabrication, and the time required to prototype and test each different layout, do not allow for me to offer these PC boards at this time.
The PowerDrive circuit is simple enough to be constructed on perf board or even careful point to point wiring. This has been done by other DIY'ers.
I have not tried the PowerDrive with an 805, since I do not have any 805 tubes. It should work (and sound great) based on my experiences with similar transmitting tubes.
I have not tried using the CCS in a cathode follower circuit, although others have suggested it here, and possibly even tried it. |
|
|
|
