So, I have a repair in currently of a $5,500 headphone amp/linestage/phono stage, and I was surprised to see paralleled 12AX7's as the first stage in the headphone amp. Other than potentially a slight decrease in noise, is there any other reason to do this?
After measuring SNR of the linestage to be just a bit under 65dB, I can't imagine that phono stage power supply noise would be a concern.
After measuring SNR of the linestage to be just a bit under 65dB, I can't imagine that phono stage power supply noise would be a concern.
Let's see the schematic. Maybe the next stage needs a bit more drive than what's possible with half of a 12AX7?
A long time ago a SS variant of paralleling low Rb input transistors to reduce the noise was around, I believe D Self did the research (not sure) ; one can only recon that the author in this case has similar ideas to keep grid noise down, buit at a price! There are certainly far better tubes than the unscreened 12AX7 !! Like always in these impedance situations the venerable ECC88 and 6N23P equivalent front ends often scoop the trophy.
rich
rich
I would be quite tempted to de-couple the second triode and see how it sounds. 🙂
$5500 for phono/headphone/linestage? It had better have one darn good MC input. Maybe more than one? I still can't see how it can cost that much.. is the box shiny chrome? 🙂
$5500 for phono/headphone/linestage? It had better have one darn good MC input. Maybe more than one? I still can't see how it can cost that much.. is the box shiny chrome? 🙂
Last edited:
I used to parallel up to 4 sections of 12AX7A or 5751 (two tubes) in my very earliest designs 30 years ago - it did result in measurably less noise from the front end and could effectively drive a passive RIAA with reasonable value resistances which also helped with noise.
There was a terrible downside with the insane miller capacitance that resulted, however I was using HOMC cartridges at the time that laughed at large load capacitances.
Today a single D3A while plagued with a huge amount of miller capacitance offers better performance still. (not a small difference either - you'd need a lot 12AX7As in parallel to match its low rp and high transconductance.)
I've come to prefer cascodes for their low miller capacitance and high gain.
There was a terrible downside with the insane miller capacitance that resulted, however I was using HOMC cartridges at the time that laughed at large load capacitances.
Today a single D3A while plagued with a huge amount of miller capacitance offers better performance still. (not a small difference either - you'd need a lot 12AX7As in parallel to match its low rp and high transconductance.)
I've come to prefer cascodes for their low miller capacitance and high gain.
Yeah, it has a very, very nice box to house everything. No MC input though.
I found that the linestage used one 5867 paralleled for each channel running at 4.2 watts dissipation per plate, which is quite a bit above the specified maximum in the datasheet. The custom machined tube guards that sit almost against the glass of the tube can't be helping the heat situation. I dug into this after pulling the 5687's that had been in the linestage for some years, and I noticed that the plates were badly and unevenly discolored.
The Miller capacitance of a paralleled 12AX7 is nothing to scoff at, especially for a moving magnet cartridge!
I found that the linestage used one 5867 paralleled for each channel running at 4.2 watts dissipation per plate, which is quite a bit above the specified maximum in the datasheet. The custom machined tube guards that sit almost against the glass of the tube can't be helping the heat situation. I dug into this after pulling the 5687's that had been in the linestage for some years, and I noticed that the plates were badly and unevenly discolored.
The Miller capacitance of a paralleled 12AX7 is nothing to scoff at, especially for a moving magnet cartridge!
Dynaco PAS-4 did exactly this ( used both triodes in ecc83 in paralell at the
phono input stage). As a cartridge a fairly low impedance miller had less
effect.
Hi Peter,
Its useful to look at MM cartridge specs. Some MM cartridges specify quite low capacitive loading. Greater capacitive loading tends to push up the higher frequencies on such cartridges.
However in this example, the 12ax7 sections are not parallel for the phono, but for the line. I would never use 12ax7 for line stage, but hey.. high end guru's selling $5'500 worth of amplifier must know what they are doing. right? 😉
Ian
Its useful to look at MM cartridge specs. Some MM cartridges specify quite low capacitive loading. Greater capacitive loading tends to push up the higher frequencies on such cartridges.
However in this example, the 12ax7 sections are not parallel for the phono, but for the line. I would never use 12ax7 for line stage, but hey.. high end guru's selling $5'500 worth of amplifier must know what they are doing. right? 😉
Ian
Oh but Peter, they sell their fancy-named, chrome-plated super-inflated things for so much money! 🙄
I had this nice little amplifier that I had restored recently. I decided to sell it on a local online marketplace. I offered it for a fair price and made no sale.
THEN I did a new listing. I detailed how great it sounded with all this marketing psycho-dribble and nowhere-speak. Mostly it was a copy-paste job from the internet. AND just for fun, I doubled the price!
I almost immediately found a buyer.
Anyway, I don't do this often. It just doesn't feel good. Of course the buyer is extremely pleased. That gives me some consolation.
I had this nice little amplifier that I had restored recently. I decided to sell it on a local online marketplace. I offered it for a fair price and made no sale.
THEN I did a new listing. I detailed how great it sounded with all this marketing psycho-dribble and nowhere-speak. Mostly it was a copy-paste job from the internet. AND just for fun, I doubled the price!
I almost immediately found a buyer.
Anyway, I don't do this often. It just doesn't feel good. Of course the buyer is extremely pleased. That gives me some consolation.
Last edited:
Thanks for sharing this info.
It's an enlightening example. The good thing is that everybody is happy with the
end result.
It's an enlightening example. The good thing is that everybody is happy with the
end result.
The other thing that people often miss is the capacitance of the interconnect used, so in addition to high miller capacitance the cartridge also has to contend with the capacitance of the cable connecting it to the phono stage; the combination of the two with many MM cartridges is the kiss of death with HF peaking occurring inside the passband. Some induced iron and almost all HOMC are tolerant of surprisingly large amounts of input capacitance and hence better choices for use with such a phono stage.
My early phono stage designs incidentally all share this flaw (and some others) so I am by no means free of guilt in this regard. 😱 I know better now...😀
Hi Kevin
That is indeed true. I have recently been tinkering with the built-in phono stage of a commercial amplifier from the late 1980's. The RCA/Chinch connectors + shielded cable from input to valve pins alone are 56pF. In addition, the original schematic called for a 51pF styrene capacitor between input and ground. Of course it is using 12ax7.
Unbelievable stuff... The first thing I did was rip out the styrenes, and there is no need for shielding since the wire its in the chassis, so this capacitance has been drastically reduced. The results are audible.
Ian
That is indeed true. I have recently been tinkering with the built-in phono stage of a commercial amplifier from the late 1980's. The RCA/Chinch connectors + shielded cable from input to valve pins alone are 56pF. In addition, the original schematic called for a 51pF styrene capacitor between input and ground. Of course it is using 12ax7.
Unbelievable stuff... The first thing I did was rip out the styrenes, and there is no need for shielding since the wire its in the chassis, so this capacitance has been drastically reduced. The results are audible.
Ian
Last edited:
1. 12BZ7 is basically two 12AX7s in parallel. Double the gain, double the current, double the capacitance.
2. Paralleling devices (or using "wider" ones) reduces noise (statistically: as sqrt(N)), but reduces input impedance (i.e., more capacitance).
You may find it useful to build a cascode, so that miller effect is canceled out (and as a bonus, plate resistance, and therefore mu, is vastly increased).
You could use a pentode, but that would exhibit partition noise, about balancing the increase in gain.
A cascoded triode, or a pentode with CCS load (or a cascode with a CCS load!), should give the least noise and the most gain, for the most input impedance. (The CCS load cases will need shunt feedback to control gain and maintain linearity. This puts load back on the input, but the shunt resistors can be arranged to terminate the source impedance as needed, so this isn't a bug, it's a feature!)
Tim
2. Paralleling devices (or using "wider" ones) reduces noise (statistically: as sqrt(N)), but reduces input impedance (i.e., more capacitance).
You may find it useful to build a cascode, so that miller effect is canceled out (and as a bonus, plate resistance, and therefore mu, is vastly increased).
You could use a pentode, but that would exhibit partition noise, about balancing the increase in gain.
A cascoded triode, or a pentode with CCS load (or a cascode with a CCS load!), should give the least noise and the most gain, for the most input impedance. (The CCS load cases will need shunt feedback to control gain and maintain linearity. This puts load back on the input, but the shunt resistors can be arranged to terminate the source impedance as needed, so this isn't a bug, it's a feature!)
Tim
It will change the voicing considerably. It's the reason i avoid parallel tubes but maybe it worked for this particular design. Or maybe they just did not need the second triode due to layout considerations.
I have an 80s guitar amp (Yamaha T100C) that had a paralleled 12AX7 driving the effects loop send. That amp would only output around 80W into a dummy load, and troubleshooting pointed out it was that paralleled stage. So I did the trick of splitting the cathodes but leaving the plates and grids connected together, this immediately increased the output of the stage and the amp put out over 100W.
The above experience, plus others I've had in high end audio, makes me a bit leery of paralleling triodes unless the cathodes are split to their own resistors. In some hi fi builds I found this helped the sound (like in my nuvistor preamp). It might have been my imagination, but sometimes paralleled triodes sound like they are fighting each other, I guess unless they are well balanced.
OTOH I am building a M77 preamp using a Douk PCB, and that has paralleled 12AY7's for each section of the line stage (each gain stage and cathode follower are paralleled). No individual cathode resistor. And it sounds incredible. Old dogs can learn new tricks.
The above experience, plus others I've had in high end audio, makes me a bit leery of paralleling triodes unless the cathodes are split to their own resistors. In some hi fi builds I found this helped the sound (like in my nuvistor preamp). It might have been my imagination, but sometimes paralleled triodes sound like they are fighting each other, I guess unless they are well balanced.
OTOH I am building a M77 preamp using a Douk PCB, and that has paralleled 12AY7's for each section of the line stage (each gain stage and cathode follower are paralleled). No individual cathode resistor. And it sounds incredible. Old dogs can learn new tricks.
I'm guessing you used two 1k bias resistors, which is the same as if you'd used a 500R shared resistor. It wasn't splitting the cathodes that fixed the problem, it was warming up the bias.
http://rudn.nodevice.com/preview/big/352/352897-2.jpg
http://rudn.nodevice.com/preview/big/352/352897-2.jpg
Last edited:
- Status
- Not open for further replies.