VladimirK, I am very impressed that you are reaching for such a high bandwidth. Most of us 'hi enders' stop at 100K or so. It has always been controversial as to WHY such bandwidth is necessary, BUT it is partially answered by the example of the AD797 which has a gain-bandwidth approaching 100MHz. The 'challenge' of extending the gain-bandwidth must be worth it, even for audio frequencies. In my opinion, this is because ALL ELSE BEING EQUAL wider gain-bandwidth gives better measured performance than lower gain-bandwidth designs. A good example of this is the uA741 vs the 4558. The biggest change that I can note is the GB, and this makes this design 'tolerable' while the uA741 is now considered 'terrible'.
Now, once you have created this amp or preamp with such high bandwidth, what do you do with the extended bandwidth? This is the arguable debate, and it has been controversial for decades. Personally, I tend to put a 6dB/octave filter around 100K in my commercial designs, just to shunt out any RFI pickup that might get in.
Others of my acquaintance would shudder at less than 500KHz, and who am I to tell them they are wrong? I have even been criticized when I reduced the working bandwidth of a 250W/channel of a custom power amp for VMPS from 7MHz to 1MHz. Go figure, but Brian Cheney had pretty good ears, and I came to depend on them at times.
Now, once you have created this amp or preamp with such high bandwidth, what do you do with the extended bandwidth? This is the arguable debate, and it has been controversial for decades. Personally, I tend to put a 6dB/octave filter around 100K in my commercial designs, just to shunt out any RFI pickup that might get in.
Others of my acquaintance would shudder at less than 500KHz, and who am I to tell them they are wrong? I have even been criticized when I reduced the working bandwidth of a 250W/channel of a custom power amp for VMPS from 7MHz to 1MHz. Go figure, but Brian Cheney had pretty good ears, and I came to depend on them at times.
Hello, it's Me.
You could wrench V-grade J73/K146 or J72/K147 out of trash Japanese gear, untill have have enough well-matching pairs.
Twice the capacitance, also twice the transconductance of the monolithic pastics, but 3 times the Pd.
(but Hey, if that's too much trouble for ya, you can always buy NOS metal can V-grades from me, say $100/each or $500 for a 73/146 couple within 0.1mA Idss. Hold on, or how about 50 bucks for a Violet tall-hat bachelor party ?)
Hi John,
Thank you.
When the JFETs are biased at, say, 15mA and the MOSFETs are biased at, say 50mA, there will be less dV variations on the JFETs drain resistors, for a given dI. This makes the VDS less variable (more constant) with the JFETs varying current. Hence, distortion is attenuated, since distortion is caused by varying VDS.
It seems to me that this is yet additional aspect of the ingenuity of your circuit, which wasn't discussed in the thread.
Hi Christophe,
Thank you.
I'd rather leave the fights to others. I'm interested in electronics and in sound quality more than in fights.
Let's not let things get 'out of hand'. It is true that the 2SJ389/J109 V's are very hard to find commercially, but they are still in the world, today, and we still design with them (for special projects). Sooner or later, LSI will produce P channel 2SJ74 parts in enough quantity to meet this requirement. I have some here in my lab, and they work just fine.
My 'best' advice, if you want to build the Blowtorch circuit just for yourself, is to DOUBLE UP with the 2SK170/2SJ74's BL, and keep your input pot limited to 10K. Don't go for the 146/73 solution, because most of these parts have too low of current too. Trust me, I have many hundreds here, pre-sorted for a future project, and the current fallout is mediocre, and not practical, AND THEN, if you used them, you could not scale down to 1/2 the capacitance when the proper parts become available.
Just live with double capacitance at first, and later (if you want) upgrade to normal capacitance.
My 'best' advice, if you want to build the Blowtorch circuit just for yourself, is to DOUBLE UP with the 2SK170/2SJ74's BL, and keep your input pot limited to 10K. Don't go for the 146/73 solution, because most of these parts have too low of current too. Trust me, I have many hundreds here, pre-sorted for a future project, and the current fallout is mediocre, and not practical, AND THEN, if you used them, you could not scale down to 1/2 the capacitance when the proper parts become available.
Just live with double capacitance at first, and later (if you want) upgrade to normal capacitance.
Hello, it's Me.
You could wrench V-grade J73/K146 or J72/K147 out of trash Japanese gear, untill have have enough well-matching pairs.
Twice the capacitance, also twice the transconductance of the monolithic pastics, but 3 times the Pd.
(but Hey, if that's too much trouble for ya, you can always buy NOS metal can V-grades from me, say $100/each or $500 for a 73/146 couple within 0.1mA Idss. Hold on, or how about 50 bucks for a Violet tall-hat bachelor party ?)
Hi Jacco,
Thank you.
Your offer is very generous, $500 per matched pair is a bargain. Unfortunately I don't own a house which I could sell in order to take advantage of your generous offer.
Trustworthy sources currently charge $50 for a 2x200mW, in the Blue edition.
A non-matched 2x600mW NOS oldtimer likely goes for $100, in Blue.
A pair of very well matching Violet duallies with twice minimum Idss of an average Blue J109/K389 for 5 C-notes isn't that off the bend, imo.
Message was that to do the crime, you pay the fine.
(I take it an offer for my house, above 1.5M plus a bonus for my parts inventory, is not an option ?)
A non-matched 2x600mW NOS oldtimer likely goes for $100, in Blue.
A pair of very well matching Violet duallies with twice minimum Idss of an average Blue J109/K389 for 5 C-notes isn't that off the bend, imo.
Message was that to do the crime, you pay the fine.
(I take it an offer for my house, above 1.5M plus a bonus for my parts inventory, is not an option ?)
Reminds me of a Goalkeeper ammo chamber that dropped on me during a conversion job.
The Goalkeeper has it's 30mm shells stored under deck, ammo feed to the Gatling-U8 unit is through a rotating conveyor belt system.
At 70 rounds/second, you can imagine that the inside of the ammo chamber resembles an XXL blender.
No-one is allowed inside the chamber when the gun is ratling, and it makes quite a noise, the walls of the chamber are extra thick for safety and rigidity reasons.
When that hefty thing tilted and fell over on the yard floor, I was standing right in dead center of the sucker.
Which coincidentally was exactly where the door opening of the construction was, fortunately the just as brutal door was not mounted yet.
For that nostalgic Comedy Caper feeling, with an extremely pale-faced Buster Keaton.
I did not spend any special efforts for achieving MHz passband, simply I use special fast parts and simple schematics, and passband comes as a free addendum. But, I find it positive for sound quality, the less Ciss and Crss - the better. I try to attract attention to speaker back EMF, since it can not be handled by an amp with the speed of light, and cause more problems to GNFB amps.
There are plenty of simulations of various circuits, but I never seen even one, which would include real speaker model. To me, it would be much more useful, than inventing 0,0000001% THD model schematics.
As for reasonable limits for passband, I simply know some opinions of respectable amp builders, that from their listening experience, PB evolution 200kHz, 400kHz, 800kHz is accompanied by rise in sound quality. I have verified this conclusion by myself, but it became possible only after the moment, when I got PMC EB1i speakers at my listening room. I consider high-end audio as purely experimental field, since linear amplification theory serves for very basic issues only.
Last edited:
It has been my experience as well..... Wide bandwidth falls out from trying to make the audio amplifier more linear --- cascode it for lower distortion and you get wider BW. Use lower C devices,.... you get wider BW. etc. More gnfb... wider BW. Then you have to deal with that byproduct.... RFI etc. Gain here and lose there. Reduce here and add there.
Thx-RNMarsh
Thx-RNMarsh
I did not have any problems caused by RFI, never listened to RFI effects, no obseved anything by scope, therefore this issue is not undisputable for me. The guy in Moscow who builds 1MHz tube SE amps, also fights for maximum PB of the input stage at No NFB amps, and he did not mention any problems with RFI. In fact, I use sometimes input LP filters, but just to make GNFB design more stable. If design is dead stable without LP filter, I would not use it.
Last edited:
I agree Richard, intrinsically high bandwidth comes from careful device selection and cascoding. Whether this very high bandwidth is useful with typical 'quality' sources, is another debate altogether.
It must be understood, that many 'critics' here, think that 20KHz is more than enough, and even preferable to extended bandwidth. I certainly do not agree with this, and I believe that at least 100KHz should be passed in the electronics, even if the speakers most probably will not pass much more than 20KHz, just to keep the risetime of the electronics up to something reasonable.
However, every source that I know of, including master recordings, just does not top the 100KHz limit and is usually less than 1/2 that, or 50KHz, or so. This includes, the best practical microphones, 30 ips analog recording, and SACD.
Yet, some prefer even more working bandwidth in the electronics. Perhaps there is something we are missing. I'll keep an open mind about it, I hope that everyone else will.
It must be understood, that many 'critics' here, think that 20KHz is more than enough, and even preferable to extended bandwidth. I certainly do not agree with this, and I believe that at least 100KHz should be passed in the electronics, even if the speakers most probably will not pass much more than 20KHz, just to keep the risetime of the electronics up to something reasonable.
However, every source that I know of, including master recordings, just does not top the 100KHz limit and is usually less than 1/2 that, or 50KHz, or so. This includes, the best practical microphones, 30 ips analog recording, and SACD.
Yet, some prefer even more working bandwidth in the electronics. Perhaps there is something we are missing. I'll keep an open mind about it, I hope that everyone else will.
- Status
- Not open for further replies.