Hi,
Thank you for going through the trouble to invest into this project, Kevin.
At three times the transconductance of the EC86 there's bound to be much more variance between samples so I'm curious about that.
OTOH it may allow us to use less triodes in //. Hopefully..They'd still need to be pretty closely matched though...........
Remains the question of linearity at Vg close to 0 and at what B+ and what topology to use best. I'm inclined to go for mu-followers maybe followed by WCFs to drive cable capacitance but I am just thinking out loud here....
From experience I also know that these kinds of high gain, high gm stages are notorious for picking up all kinds of crap so layout is very important.
Cheers,
Thank you for going through the trouble to invest into this project, Kevin.
At three times the transconductance of the EC86 there's bound to be much more variance between samples so I'm curious about that.
OTOH it may allow us to use less triodes in //. Hopefully..They'd still need to be pretty closely matched though...........
Remains the question of linearity at Vg close to 0 and at what B+ and what topology to use best. I'm inclined to go for mu-followers maybe followed by WCFs to drive cable capacitance but I am just thinking out loud here....
From experience I also know that these kinds of high gain, high gm stages are notorious for picking up all kinds of crap so layout is very important.
Cheers,
Definitely makes sense to me - my highest gain design to date is 54dB and the intent there was for output levels to meet or exceed redbook CD levels with most available cartridges and reasonable transformer choices where employed. Here we would be talking between 20 and 26dB of additional gain I would speculate.
I've not tried paralleling triodes in more recent designs, and found that even the 12AX7A which has puny transconductance and relatively good consistency does not parallel nicely. In those days (!) I used individual cathode resistors (unbypassed yet and wondered why those designs were a bit noisy ) to get them to work well together.
We're ambitious though..
Since we are going for gain sans transformers I think we can tolerate at least 1000pF of miller capacitance at the input, noting my previous comments about > 26nF at the cartridge with a 1:16 SUT and just 100pF of input and cable capacitance.. A mu follower is definitely viable given the relaxed input capacitance requirements. The cascode could potentially provide much more gain with similar overall noise performance so it is still possibly something to consider.
I avoided cascode front ends for decades because everyone insisted that they did not sound good, so far I have not found that to be the case with the fairly insane tubes I have used thus far, and they solve a lot of problems. I can easily get more than 50dB of gain out of a cascode with the right operating point, which largely results in inconsequential noise contributions in the equalizer network and second gain stage with some intelligent design choices...
A WCF on the output strikes me as a good idea (used in OTL headphone amp design!) should be able to drive long cables without issue. Even with the hybrid mu-follower in the MMIII I find some unfortunate cable interactions. (Output z is just 150 ohms) I think the 6N6P could be a good economical and yet high performing choice for a WCF, the 6H30 could be even better, but is probably overkill.
I've not tried paralleling triodes in more recent designs, and found that even the 12AX7A which has puny transconductance and relatively good consistency does not parallel nicely. In those days (!) I used individual cathode resistors (unbypassed yet and wondered why those designs were a bit noisy
) to get them to work well together. We're ambitious though..
Since we are going for gain sans transformers I think we can tolerate at least 1000pF of miller capacitance at the input, noting my previous comments about > 26nF at the cartridge with a 1:16 SUT and just 100pF of input and cable capacitance.. A mu follower is definitely viable given the relaxed input capacitance requirements. The cascode could potentially provide much more gain with similar overall noise performance so it is still possibly something to consider.
I avoided cascode front ends for decades because everyone insisted that they did not sound good, so far I have not found that to be the case with the fairly insane tubes I have used thus far, and they solve a lot of problems. I can easily get more than 50dB of gain out of a cascode with the right operating point, which largely results in inconsequential noise contributions in the equalizer network and second gain stage with some intelligent design choices...
A WCF on the output strikes me as a good idea (used in OTL headphone amp design!) should be able to drive long cables without issue. Even with the hybrid mu-follower in the MMIII I find some unfortunate cable interactions. (Output z is just 150 ohms) I think the 6N6P could be a good economical and yet high performing choice for a WCF, the 6H30 could be even better, but is probably overkill.
Hi,
Yes....Anyone having designed OTL amps will confirm this but short of having individual bias supplies for each triode you can only force them by doing away with bias altogether (and subsequently force current flow (bear with me)) etc.
Which is why there's no bias in my headamp and currents are force fed by (gosh) sand CCSs......)
As I'm sure you know, you can mimic a sand css by using highish rp valves as the upper part of the mu follower without much of a noise penalty. It being a bootstrap css etc.
I can easily understand people's reluctance against using triodes in // but here it is hard to escape it or noise will dominate........
Best,
Yes....Anyone having designed OTL amps will confirm this but short of having individual bias supplies for each triode you can only force them by doing away with bias altogether (and subsequently force current flow (bear with me)) etc.
Which is why there's no bias in my headamp and currents are force fed by (gosh) sand CCSs......)
As I'm sure you know, you can mimic a sand css by using highish rp valves as the upper part of the mu follower without much of a noise penalty. It being a bootstrap css etc.
I can easily understand people's reluctance against using triodes in // but here it is hard to escape it or noise will dominate........
Best,
Hi,
Great to see you guys are at it again.
I know you might prefer triodes. However, I can't forget the results that Frank Blöhbaum showed in Linear Audio Volume 0. Here the D3a has the lowest noise, which in triode mode Frank measure 2.44 nV/RtHz. But the 6J9P is second best at 3.06 nV/RtHz.
I think that is a fine result from such a cheap valve and the possible spread should even out when a few is paralleled. Frank B. also mention it has low 1/f noise properties.
I seem to remember that Kevin already have some experience with this valve, so he might have some comments.
Mogens
Great to see you guys are at it again.
I know you might prefer triodes. However, I can't forget the results that Frank Blöhbaum showed in Linear Audio Volume 0. Here the D3a has the lowest noise, which in triode mode Frank measure 2.44 nV/RtHz. But the 6J9P is second best at 3.06 nV/RtHz.
I think that is a fine result from such a cheap valve and the possible spread should even out when a few is paralleled. Frank B. also mention it has low 1/f noise properties.
I seem to remember that Kevin already have some experience with this valve, so he might have some comments.
Mogens
Hi Piano3,
You might be right. I have no practical experience with this valve. I do have 10 pieces around here somewhere.
I did build an MC amplifier +20 years ago based on ECC88. These also need carefully selection. It was from Audio Express magazine. In the end I was never happy with it. I think I would take the easy way out and settle for a step-up transformer. But, where is the fun in that?
Mogens
You might be right. I have no practical experience with this valve. I do have 10 pieces around here somewhere.
I did build an MC amplifier +20 years ago based on ECC88. These also need carefully selection. It was from Audio Express magazine. In the end I was never happy with it. I think I would take the easy way out and settle for a step-up transformer. But, where is the fun in that?
Mogens
Hi,
@Mogens: Could you tell us more about that ECC88 headamp from AudioExpress? Topology etc....
Or...If it's online somewhere please post us a link.
And yes, he D3A is an extraordinary little tube but we'll have to settle for something along the line and I do understand Kevin's choice of the 6C45P.
IME I've always found x-formers somewhat lacking in transparency but I most certainly haven't tried all of them. Who could possibly afford that kind of luxury?
@Mogens: Could you tell us more about that ECC88 headamp from AudioExpress? Topology etc....
Or...If it's online somewhere please post us a link.
And yes, he D3A is an extraordinary little tube but we'll have to settle for something along the line and I do understand Kevin's choice of the 6C45P.
IME I've always found x-formers somewhat lacking in transparency but I most certainly haven't tried all of them. Who could possibly afford that kind of luxury?
I've only used mc SUTs once, for my father's phono amp; they were Sowter's and the result was very good. However, I've been singularly unimpressed by many transformer input mic amps (commercial units) with the result that I've never used transformers in any of my mic amps and never had any problems as a result of this. Admittedly a condenser mic amp is an order of magnitude less problematic than an mc phono stage.
There was a transformer free mc design published by Richard Sears, whom I believe Kevin knows, but that used an extremely exotic planar triode, WE 416c, if I remember correctly. Not only does that have very high transconductance but its design may possibly make it exhibit less of the non Gaussian types of noise which may lead to a disappointing result no matter how much gm can be mustered by parallel valves.
Kevin is breadboarding a head amp, but is there actually any compelling reason to attempt to keep to 2 stages rather than 3? I was looking at some little subminiature Russian valves that I have, 6S31B, they have a gm of about 15 when run at 30mA and 60V; at this voltage, providing current for 4 or 5 in parallel is no big deal and they have a mu of a bit less than 20.
There was a transformer free mc design published by Richard Sears, whom I believe Kevin knows, but that used an extremely exotic planar triode, WE 416c, if I remember correctly. Not only does that have very high transconductance but its design may possibly make it exhibit less of the non Gaussian types of noise which may lead to a disappointing result no matter how much gm can be mustered by parallel valves.
Kevin is breadboarding a head amp, but is there actually any compelling reason to attempt to keep to 2 stages rather than 3? I was looking at some little subminiature Russian valves that I have, 6S31B, they have a gm of about 15 when run at 30mA and 60V; at this voltage, providing current for 4 or 5 in parallel is no big deal and they have a mu of a bit less than 20.
I don't want to derail the discussion, but there are some other subminiature Soviet beasts out there. Have any of you ever considered any of these?
1. 6S28B/6S29B, both look to be similar. S=19(+/-4), u=40 (+/-10)
http://klausmobile.narod.ru/td/data/_6s28bv.GIF
''GSTube.com''. Tubes, sockets etc. Parameters and characteristics 6C29b-B
2. 6S63N, RF nuvistor, S=12 (+/-2), u=75 (+/-15)
''GSTube.com''. Tubes, sockets etc. Parameters and characteristics 6C53H-B
Shunt or Folded Cascode Front End
JC Morrison did some work on a folded cascode front end for a phono stage using p-channel jfets and a pentode. You can use tubes instead of the jfets but you will need a pnp or p-channel jfet for your upper cascode element as Rod Coleman did. Coleman has a design on diyaudio.com with mc input. Both their results are excellent, perhaps give it a try and see what you think. A friend of mine has been very pleased with Coleman's circuit. I suppose the "only tube" purists can pass, but an open minded approach can often lead to surprising results.
JC Morrison:
http://www.labjc.com/wp-content/uploads/2011/10/hybrid-folded-cascode.gif
Shunt Cascode:
http://www.bartola.co.uk/valves/wp-content/uploads/2013/03/shunt-cascode-article-1.png
Rod Coleman's MC Preamp
http://www.diyaudio.com/forums/tubes-valves/261704-riaa-network-position-3.html post 24
JC Morrison did some work on a folded cascode front end for a phono stage using p-channel jfets and a pentode. You can use tubes instead of the jfets but you will need a pnp or p-channel jfet for your upper cascode element as Rod Coleman did. Coleman has a design on diyaudio.com with mc input. Both their results are excellent, perhaps give it a try and see what you think. A friend of mine has been very pleased with Coleman's circuit. I suppose the "only tube" purists can pass, but an open minded approach can often lead to surprising results.
JC Morrison:
http://www.labjc.com/wp-content/uploads/2011/10/hybrid-folded-cascode.gif
Shunt Cascode:
http://www.bartola.co.uk/valves/wp-content/uploads/2013/03/shunt-cascode-article-1.png
Rod Coleman's MC Preamp
http://www.diyaudio.com/forums/tubes-valves/261704-riaa-network-position-3.html post 24
Lots of things to investigate in the recent posts.
I am a big fan of the D3A but given the fact that it is now rapidly disappearing (at least temporarily, but perhaps permanently) I have decided to discontinue designing with and advocating this type. Prices for singles are $40 and up when you can find them, pairs typically over $100 now particularly on eBay. Some online vendors have small quantities left.. No point IMO in designing something no one else can build due to lack of parts.
The 6J9P from my practical experience is about 3dB noisier than a D3A, but I have also had some problems with 1/F noise depending on operating point which to minimize this should be < 10mA for this type. It's a nice tube in a lot of respects and is well made IMLE.
I will need to look at the recommended Russian subminis, some of which may be difficult to source.
I'm willing to go about as far as a CCS load or fet based follower, but I really don't like mixing sand and tubes in the signal path; the result just is not pleasing to my ear although the measurements are usually great. I suspect Frank would agree...
Currently we are thinking more of a dedicated mc phono stage rather than a head amp. I'm of the belief that getting as much gain out of the first stage as quietly as possible is the way to go, noise performance then is almost entirely dominated by the input stage and not much else.
I am a big fan of the D3A but given the fact that it is now rapidly disappearing (at least temporarily, but perhaps permanently) I have decided to discontinue designing with and advocating this type. Prices for singles are $40 and up when you can find them, pairs typically over $100 now particularly on eBay. Some online vendors have small quantities left.. No point IMO in designing something no one else can build due to lack of parts.
The 6J9P from my practical experience is about 3dB noisier than a D3A, but I have also had some problems with 1/F noise depending on operating point which to minimize this should be < 10mA for this type. It's a nice tube in a lot of respects and is well made IMLE.
I will need to look at the recommended Russian subminis, some of which may be difficult to source.
I'm willing to go about as far as a CCS load or fet based follower, but I really don't like mixing sand and tubes in the signal path; the result just is not pleasing to my ear although the measurements are usually great. I suspect Frank would agree...
Currently we are thinking more of a dedicated mc phono stage rather than a head amp. I'm of the belief that getting as much gain out of the first stage as quietly as possible is the way to go, noise performance then is almost entirely dominated by the input stage and not much else.
Richard and I have been friends for nearly 20yrs now, and I am familiar with the later version of the 416B/C phono stage. Some details here: 416 Triode Phono
The 416C currently runs about $350 each and my recollection is that not every tube is going to be suitable in this application; unfortunately it's a non starter due to cost.
Philosophically speaking I belong to the camp of less being more, and would very much like to avoid a three stage design; such not being the case a head amp in front of the Muscovite would be the preferred path. To try and get a sense as to whether acceptable SNR can be achieved I will try a head amp in front of the Muscovite. The ultimate goal would be a full blown MC pre-amp.. Good cable driving capability is important so a white CF may potentially be considered as part of the output section.
