But can the OPT still get saturated without an IP cap, as I speculated on in my earlier post I could only see a slightly offsetted working point for the IP tube (hence perhaps more distortion and earlier clipping) as the KT88 still have decoupling caps on their grids, no? The DC offset in the OPT is as I see it roughly dependent on the KT88 that they are both matched pairs and trimmed bias setting.
Cheers Michael
I need clarification on this. We need a cap on EVERY heater pin? 9 total?
And when you say "to the chassis", are you implying the ground line on the secondary of the power transformer is not acceptable for this and we need to create a separate line going to the chassis earth from the IEC power connector?
..Todd
True, but consider a warped record playing on a turntable if everything in the signal chain has frequency response down to say 1 Hz. In a solid state amp you will see your woofer cone moving in step with the record warp. In a tube amp you will hear the saturation induced distortion changing in step with the record warp. I have observed both.
In practice there should be a rumble filter somewhere in the signal chain (should be in the phono stage) to prevent this. If not your woofer cone would be in the front yard when you dropped the stylus in the groove. I must also state that I do not use an input cap in any of my amps and I have not had an issue. Resist the temptation to use oversized coupling caps, especially in an amplifier that uses GNFB.
I (and many DIYers) don't always build it and never touch it again. The bleeder is one of those parts that could fail and you will never know it. I design my PCB based amplifiers for the case where the builder may not always understand everything. The choke is external to the PCB and can be disconnected with a screwdriver. For these reasons I will always incorporate two bleeders in my designs. When safety is concerned I must err on the side of caution since I often don't know my end user. You guys are building your own amps and are free to do whatever you want.
Thinking in terms of this amp as a fully documented project (down the road), the input cap can be marked optional with the reasons for choice described in the documentation.
With the 200k resistor there, a .047uf cap will give us an F3 of about 17 Hz.
..Todd
Problem there is Miller from the next stage, 50pF or so, since the pentode (with feedback to its cathode) will have an extremely high effective plate resistance. With a source impedance in the megohms, you'll have a pole right where you don't want it somewhere in the 200-2kHz range. You'd need to add a cathode follower between input stage and phase splitter.
Well, not from the ones shorted together- they'll just need one. So for each monoblock, you'll have 8. They go right to the chassis, not to the ground wire. That's usually easy to do, since the tube sockets are generally mounted on the chassis, and you can either add a solder lug under one of the mounting screws or use the type of tube socket that has integral solder lugs.
If it were my amp, I'd go with just the 27k resistor and no CCS. If only to greatly simplify the P2P wiring.
Is everyone married to a CCS there? I'd be happy to toast it, if that's what everyone wants.
As an option, it's a bit awkward. ("Just remove this resistor and add this 6 component sub-circuit.")
..Todd
Is everyone married to a CCS there? I'd be happy to toast it, if that's what everyone wants.
As an option, it's a bit awkward. ("Just remove this resistor and add this 6 component sub-circuit.")
..Todd
Last edited:
I've only been thinking about tubes for a couple weeks now, I'm still reading the basics about how they work (the Crowhurst series book 2 currently). Once I've made it through that, I'll tackle [some of] the rest of the books on P. Millet's website.
I'll Google these books you mentioned. Why do I get the impression you've written them?
..Todd
I'll Google these books you mentioned. Why do I get the impression you've written them?
..Todd
I wish I had.
These were written by Morgan Jones, a former engineer at BBC. IMO, they are far and away the best currently-available books on the subject. I reviewed them on my website. Though it's no secret that Morgan and I are good friends and that we share many points of design philosophy, he wrote the most recent edition before we had met and I have no financial stake in the books' sales- I just honestly think that they're terrific and extremely helpful. In VA, Morgan does a very detailed and thorough analysis of the Mullard 5-20.
These were written by Morgan Jones, a former engineer at BBC. IMO, they are far and away the best currently-available books on the subject. I reviewed them on my website. Though it's no secret that Morgan and I are good friends and that we share many points of design philosophy, he wrote the most recent edition before we had met and I have no financial stake in the books' sales- I just honestly think that they're terrific and extremely helpful. In VA, Morgan does a very detailed and thorough analysis of the Mullard 5-20.
Wouldn't the set gain decide how big the Miller guy looks like on g1, though plate output impedance is in the MOhm region the gain is set by the external Rp and the gain goes down and so do the effective Miller capacitance seen on g1, gain times Cpg, right?
EDIT: checked little bit closer ECC81/12AT7 Philips datasheet, Cag 1,6 pF, with other words we would need a gain of 30 to come up to 50 pF.
Cheers Michael
Last edited:
I am not. No problem with CCSs, but you specifically ask "there". The CCS only guarantees equal outputs from the phase inverter, not the KT88s. Then there are the OPT internal capacitances. So in practice ....
This type of circuit has been successfully built with a tail resistor often enough, and the resistance is constant with frequency (CCS is not necessarily). I cannot get to earlier measurements now, but the impedance on the combined cathodes is low, so a 27K is already comparatively high. Regarding unequal tube ageing, I have never had a problem with triodes in the same envelope in multiple builds.
But having said that, it will help to have a small balancing pot in the anode loads for final exact equalising on the KT88 anodes.. In the judgement of some that is equal extra trouble. So some personal preference creeps in ....
tubelab, all valid points there, I can fully understand that when you are providing PCBs for the DIY community it's good practice to add "too much" than "too little".
We'll shall hope the phono stage have some rumble filter as OPTs tend to saturate quite easy at very low frequencies, one R-core transformer I have been looking as a possible choice for my KT88 amp has an impressive 800+ H a-a but a few mA and it drops like rock down to 100 H, but R-cores are very special in that they do suffer the same overly sensitivity to DC components as do toroids, other cores that comes in 2 halves or so when mounted together still have micro-small gap and that makes a BIG difference believe it or not and that's why such transformers can take more DC components without dropping so much in inductance.
While on the matter DC and ULF as I noticed Johans advice for a more exact equalising on the KT88 anodes I would tend to say the current design with separate trimmers to g1 is very cumbersome to use, one can not make any quick trimming without doing it in small steps on both trimmers forth and back forth and back... I would really prefer one trimmer that set the rough and quick trimming on both KT88 and a small trimmer which is splitting up to each grid on the KT88 pair and makes the fine equalising Johan is talking about, but this is on the grid side and not on the anode as Joahn talked about.
Thinking here on the matter of adding some HP filter function in the first tube stage in an attempt to lower DC/ULF sensitivity I am pondering here for some practical places to add a pole and while there may be couple of places to do that but to mention one I was thinking one could add a R+C in parallel with the current Cathode resistor, the values will be chosen such that the single resistor sets the gain below cf while the R+C sets the final gain for the whole audio band and up, any drawback on that idea?
Cheers Michael
We'll shall hope the phono stage have some rumble filter as OPTs tend to saturate quite easy at very low frequencies, one R-core transformer I have been looking as a possible choice for my KT88 amp has an impressive 800+ H a-a but a few mA and it drops like rock down to 100 H, but R-cores are very special in that they do suffer the same overly sensitivity to DC components as do toroids, other cores that comes in 2 halves or so when mounted together still have micro-small gap and that makes a BIG difference believe it or not and that's why such transformers can take more DC components without dropping so much in inductance.
While on the matter DC and ULF as I noticed Johans advice for a more exact equalising on the KT88 anodes I would tend to say the current design with separate trimmers to g1 is very cumbersome to use, one can not make any quick trimming without doing it in small steps on both trimmers forth and back forth and back... I would really prefer one trimmer that set the rough and quick trimming on both KT88 and a small trimmer which is splitting up to each grid on the KT88 pair and makes the fine equalising Johan is talking about, but this is on the grid side and not on the anode as Joahn talked about.
Thinking here on the matter of adding some HP filter function in the first tube stage in an attempt to lower DC/ULF sensitivity I am pondering here for some practical places to add a pole and while there may be couple of places to do that but to mention one I was thinking one could add a R+C in parallel with the current Cathode resistor, the values will be chosen such that the single resistor sets the gain below cf while the R+C sets the final gain for the whole audio band and up, any drawback on that idea?
Cheers Michael
A 12AT7/ECC81 LTP yields a voltage gain of about 17X. CMiller is not much of an issue, when 'T7s are employed. Remember the type's VHF front end heritage.
If Dr. Miller's "pet" causes lack of sleep, definitely use the 6GK5 voltage amplifier. That type's high gm and low RP will swamp LTP capacitances.
A 150 KOhm grid to ground resistor in combination with a 68 nF. cap. sets F3 at a very nice 15.6 Hz. BTW, this is the place to use a PIO part, should that tickle your fancy. The I/P cap. is outside of the GNFB loop and attempts to control voicing here have good prospects of success.
If Dr. Miller's "pet" causes lack of sleep, definitely use the 6GK5 voltage amplifier. That type's high gm and low RP will swamp LTP capacitances.
A 150 KOhm grid to ground resistor in combination with a 68 nF. cap. sets F3 at a very nice 15.6 Hz. BTW, this is the place to use a PIO part, should that tickle your fancy. The I/P cap. is outside of the GNFB loop and attempts to control voicing here have good prospects of success.